SEO Archives - iPullRank

Google’s Search Quality Rater Guidelines and YMYL in the Age of AI Search

Francine Monahan — Thu, 05 Feb 2026 12:00:00 +0000

Explore this topic in AI Search

Want a different perspective? These links open AI Search platforms with a prompt to explore this topic, how it works in AI Search, and how iPullRank approaches it.

*These buttons will open a third-party AI Search platform and submit a pre-written prompt. Results are generated by the platform and may vary.

When I recently wrote a blog that walked through the process of performing a query fan-out and building a content plan, I was actually surprised by how many reputable sources were cited by LLMs and on the Google SERPs during my research.

I wrote about how a fictional financial firm could strategize content for queries relating to proper retirement saving (very much a “Your Money or Your Life” topic) and many results cited universities, government organizations like the IRS and the Department of Labor, and long-standing financial institutions like Fidelity and Charles Schwab.

The LLM results for this important YMYL topic were actually giving me information that I didn’t immediately discard like I would have when AI models first appeared. But how often does that actually happen these days? Was this a rare case?

And sure, Google has made some recent edits to its Search Quality Rater Guidelines (SQRG) to meet the needs of AI Search today, but are they helping? And is it enough?

This blog will discuss how AI Search is impacting Google’s search guidelines and look at some YMYL AI searches to see how their citations stack up.

AI Impacts on Google’s Search Quality Rater Guidelines

Think of Google’s SQRG as sort of the blueprint Google uses to train its algorithms. Around 16,000 humans rate and classify content based on its quality and value. And now that various LLMs, AI Overviews, and AI Mode are being used for searches, we need to understand their impacts on how Google defines quality.

1. The Expansion of YMYL

Your Money or Your Life topics can have a risk of harm or negative impact to a person’s health, financial stability, or safety. In the latest SQRG updates in January and September 2025, Google broadened the scope of YMYL:

Government, Civics and Society: This category now includes explicit mention of election and voting information, as well as content that impacts trust in public institutions.
Health or Safety: Google clarifies that while any page can be hypothetically harmful, YMYL status is reserved for topics that could “potentially impact people’s health, financial stability, or safety, or the welfare or well-being of society.”

2. The Rise of Scaled Content Abuse

With AI making it easy to churn out thousands of pages of content, the January 2025 update introduced a guideline that focused on spammy scaled content to better align with Google Search Web Spam Policies.

Google’s guidelines now explicitly state that the lowest quality rating applies if content is:

Copied or paraphrased from different sites, or AI-generated in a “low-effort way”
Lacking originality or added value compared to the other existing pages
Essentially filler created at scale meant to inflate a page’s length without adding substance, regardless of if a human or AI created it

If a chatbot could have written your article in 10 seconds without human oversight, Google’s raters are going to mark it as low quality.

3. E-E-A-T: "Experience" is the AI-Killer

In a slop-infested web, that first “E” in E-E-A-T (Experience) can become your most valuable asset. AI can summarize expertise, but it cannot experience things for itself.

Google’s systems prioritize content that proves real-world involvement. They want to see first-hand life experience. To align with the guidelines, your content should:

Use First-Person Insights: Use phrases like “In our tests,” or “Our survey found that…”
Include Evidence of Effort: Original images, unique data sets, and case studies are still essential to distinguish your business.
Strengthen Author Entities: Ensure your authors have visible credentials and a digital footprint (LinkedIn, professional bios) that Google’s AI can verify.

This is a very important point that we continue to drive home with Relevance Engineering by focusing on proprietary data, creditable research, and high-quality content.

4. Optimizing for Needs Met

The January 2025 update added new guidance on “assessing minor interpretations and intents for Needs Met ratings and added illustrative examples.”

It’s not as groundbreaking as the other updates, but I feel like this was a missed opportunity to include some specific language referring to AI Overviews or AI Mode. I think they fit into the category of Special Content Result Block (SCRB) pretty well, but maybe Google is saving that for a later update once they’ve been in use for several years (or once they finally take the leap and make AI Mode the default).

However, Google still refers to the internet as the “World Wide Web” in the guidelines, so they’re obviously not fully up to date quite yet.

ChatGPT Health and Finance

One of the more concerning product releases of the past year has to be ChatGPT’s HealthBench. It’s a benchmark that was created to measure the capabilities of AI systems for health-related queries. As OpenAI writes, “Built in partnership with 262 physicians who have practiced in 60 countries, HealthBench includes 5,000 realistic health conversations, each with a custom physician-created rubric to grade model responses.”

They claim that their models improved by 28% thanks to HealthBench and outperformed other models such as Claude 3.7 Sonnett and Gemini 2.5 Pro.

To put it bluntly, they could’ve trained their models with 5 million health conversations and I still wouldn’t trust it for health advice. AI is still just predicting text. It’s not understanding the full context, wider health issues and symptoms, reactions to certain treatments and, of course, how incredibly personalized and varied healthcare can be for each person.

In addition to trying to improve its health knowledge, OpenAI has also been focused on financial benchmarking. The company says they are “accelerating financial analysis, generating executive-ready summaries, automating routine documentation, and brainstorming strategic financial plans—all while considering compliance and accuracy standards.”

I couldn’t imagine revenue forecasting or budget planning with an AI tool for many of the same reasons I listed above. It’s very specific to each company. When budgeting, you have to consider promotions and raises, future employees, training and development, ad spend, equipment purchasing, events, etc. Entering in and calculating all of that is manual, so I don’t see much time savings with LLMs.

Who knows what can happen in the future, though.

YMYL in AI Search

You probably know by now how much I love screenshots, so let’s do some AI Search querying for serious, high-risk topics and see what kind of sources are being cited these days (and which model does the best job). I chose three queries in the categories of finance, health, and safety.

Finance YMYL Query: Choosing the Right Life Insurance

For my first search, I queried “how to choose the right life insurance.”

The AI Overview didn’t cite any sources, but the AI Mode result did, focusing on financial groups like the Western and Southern Financial Group, Farm Bureau Financial Services, and Stolly Insurance Group, as well as the nonprofits National Association of Insurance Commissioners and the Insurance Information Institute.

We know at this point that ChatGPT is allergic to citing sources unless you specifically prompt it to, and even then, it’s hit or miss on whether the source is a real one or the information is accurate. But this time, when I asked it to cite sources, its links to the Institute of Financial Wellness, Forbes, and the Louisiana Department of Insurance were legit and lead to helpful articles – a vast improvement from when ChatGPT first launched and cited broken links, fake studies, or nothing at all.

Gemini cited:

New York Life Insurance
Western and Southern Financial Group
Trusted Senior Specialists
Forbes
NerdWallet
Mutual of Omaha
Experian

I noticed a difference right away in that Gemini didn’t cite any nonprofit organizations, just mostly insurance companies (which can’t always be trusted as unbiased). I found this interesting since, you know, Gemini is owned by the same folks who own AI Mode. One would think they’d show similar results.

Perplexity only cited NerdWallet and a few insurance companies, one of which I had never even heard of before (Thrivent).

The Winner: If I had to choose a winner here, I would probably go with AI Mode for the most unbiased and helpful citations (like the two nonprofits it cited).

Health YMYL Query: Signs of a Heart Attack

My next query was “signs of a heart attack.”

I did not receive an AI Overview when I searched for this, but it gave a simple bullet list of symptoms from the Mayo Clinic in a Featured Snippet.

AI Mode cited:

Mayo Clinic
American Heart Association
Geisinger Health (a clinic)
Dignity Health (a clinic)
Houston Methodist (a clinic/academic institute)

I think this is another great selection of trustworthy sources.

ChatGPT cited:

American Heart Association
The CDC
National Heart, Lung, and Blood Institute
Mayo Clinic
CardioSmart (American College of Cardiology)

But, again, I had to specifically tell it to cite its sources. I don’t understand why I still need to do that when every other model cites them openly.

For some reason, Gemini did not cite any sources for this query. I used my ChatGPT method and told it to cite its sources and it still provided no links, but it mentioned in the text that the American Heart Association and the CDC were referenced. I would love to understand the reason behind that because I can’t figure it out.

Perplexity cited:

National Heart Association
CDC
National Health Service
Mayo Clinic
Geisinger Health
Cleveland Clinic

No complaints here. I think these are great sources to cite.

The Winner: I think this one is a 3-way tie between AI Mode, ChatGPT (even though I had to prompt it to cite sources) and Perplexity for citing great, legit sources. I’m still mad at Gemini for not citing anything until I asked, and then only citing two with no links to anything for additional research.

Safety YMYL Query: How to Detect a Gas Leak

For my final query, I chose “how to detect a gas leak in your home.” This is an interesting query because it focuses on safety but the results could also be product-oriented when displaying detection tools a person can buy, which could lead to biased content.

Again, there was no AI Overview for this query.

AI Mode cited mostly heating, plumbing and gas/energy supplier companies such as:

To the T Plumbing and Heating
Constellation (energy supplier)
Con Edison (energy supplier)
Gas South (energy supplier)
Waldrop Plumbing, Heating and Air

It also cited a page of Amazon “Best Sellers” listings for natural gas detectors:

Again, ChatGPT needed me to specify that I wanted it to cite its sources and it cited:

Constellation
Eversource
Renke (environmental monitoring devices like sensors)
Local Rooter and Plumbing
AAA
California Public Utilities Commission

Gemini cited its sources at the bottom for me this time without my asking (though it didn’t identify which information was taken from what source like the others). It cited:

American Red Cross
CDC
National Grid/SoCalGas
CT.gov (recognizing that I’m in Connecticut)

It even asked me if I wanted the emergency contact number for the local gas utility in my town because it knows where I live:

Perplexity gave me several videos to watch, which was interesting and unlike the other models. The videos were by a plumbing education brand, a local government organization called County Office, and an energy supplier (DTE Energy).

It cited many articles by various gas, plumbing, and energy companies, but it was the only model to cite Reddit. Roll your eyes if you want, but there were useful answers provided (aside from the person who wrote “A good way to tell is if you light a match and your house explodes you probably have a leak”).

The Winner: I think I have to give this one to Perplexity. Though I appreciated that Gemini knew my location and offered to provide localized help, the diversity of sources offered by Perplexity, as well as the useful videos, put it over the top.

YMYL in the AI Search Age

AI Search is impacting how trust is synthesized and displayed for the most important queries. While my initial skepticism regarding YMYL topics was met with surprisingly high-quality citations, it’s still kind of the wild west out there. Always be sure to fact-check the information you’re getting from AI models.

The winning LLMs to me in this case varied by query, and the selection of citations was vast, but the models all appeared to be catching on to the importance of trust.

Some takeaways:

AI can synthesize information, but it cannot replicate the “I was there” factor. Personal case studies, original imagery, and first-person insights are still your strongest defense against being labeled as “low-effort” scaled content.
With the 2025 updates, YMYL now covers more than just health and wealth. If your content impacts civic trust or societal well-being, expect the highest level of scrutiny from both algorithms and human raters.
Every AI model is different, so you’ll need to find your flavor when performing important searches. Regardless, whether it’s Perplexity’s diverse sources or Gemini’s localized emergency help, being the cited authority is the goal.

Google might still be calling the internet the “World Wide Web” in its guidelines, but the reality is that we’ve moved into an Answer Engine world. To stay relevant, businesses must build entities and prove through proprietary data and expert credentials that they are the source the AI (and humans) should trust.

Need help becoming more visible online?

Read about Relevance Engineering

The post Google’s Search Quality Rater Guidelines and YMYL in the Age of AI Search appeared first on iPullRank.

Digital Sleuthing: How to Find and Decipher Google’s Patents, Leaks, and PR

Francine Monahan — Thu, 29 Jan 2026 12:00:00 +0000

Explore this topic in AI Search

Want a different perspective? These links open AI Search platforms with a prompt to explore this topic, how it works in AI Search, and how iPullRank approaches it.

*These buttons will open a third-party AI Search platform and submit a pre-written prompt. Results are generated by the platform and may vary.

When a multi-billion-dollar corporation sits at the center of the entire SEO industry, it can be tough to believe everything they say. Constant press releases with talking heads giving vague quotes can throw the industry into a tailspin in an instant, but this doesn’t always have to be the case.

There’s truth in them there hills. You just have to know how to find it.

From patents and leaks to official PR, there’s a lot to filter through, but there are many ways to parse through the information Google releases. This blog will show you how to decipher the real meaning and separate fact from vague corporate untruths.

Understanding Google Patents

For years, well-known SEO researcher Bill Slawski was at the forefront of analyzing Google patents. Since his passing in 2022, the torch hasn’t so much been passed but rather divided between a number of SEOs trying to pick up the slack.

Patent watcher and leak analyzer Mike King still touts Bill’s work and recommends that anyone new to patents should read his work at SEO by the Sea and familiarize themselves with it. That way, you’ll start to understand and recognize the terminology Google uses.

“The difficulty is they’re always going to be dense writing,” Mike said of the patents. “They’re not written to be readable.”

Patents are written that way intentionally, with vague phrasing and broad descriptions to cover all of Google’s bases legally.

Mike started looking into patents years ago after hearing many engineers and Googlers using terminology that he didn’t understand at first. Then he searched through patents for those same words to learn what everyone was talking about.

Most recently, Mike did a deep dive into the AI Mode patents and how it all works.

In the SEO world, experts make discoveries and share them at conferences. Mike says that Google does the same thing with its research by sharing papers and information at conferences for engineers.

“All of these people are sharing their ideas with each other and building upon them,” Mike said.

Now that we know what to expect, let’s look into where to find patents and how to read them.

Where to Search for Patents

There are a few main locations where you can easily find patent applications and their statuses:

Google Patents: The most user-friendly interface. This is your primary hub that indexes global filings and offers the most intuitive search interface.
Google Research: Where you can find official product and research releases directly from Google.
USPTO & Espacenet: These are official government databases with the most up-to-date legal statuses.
arXiv.org e-Print archive: A directory for white papers.

Be sure to filter your search with certain key phrases. If you’re using a site that’s not Google’s official patent directory, add the search operator assignee “Google LLC” to narrow it down, for example.

You can also use CPC (Cooperative Patent Classification) codes like G06F 16/00 (Information Retrieval) to find patents grouped by technology rather than just keywords.

“I don’t browse randomly,” said Metehan Yeşilyurt of AEO Vision. “I usually start from an observed behavior in search and work backwards. My main tools are Google Scholar (citation chains matter more than single patents), arXiv to understand the research roots, and advanced Google searches like “filetype:pdf” combined with ranking or retrieval terms. I’m looking for patterns, not one document. I’m also using consensus.app, recently.”

SEO expert and consultant Marie Haynes said she doesn’t really have a solid process for finding interesting papers and patents – she just keeps her ear to the ground.

“I have an RSS reader that I vibe coded that finds me the latest Google blog posts and more,” she said. “I pay close attention to this page: https://research.google/pubs/.”

She also scours X (formerly Twitter) for interesting papers that people are talking about.

“Most of what I find simply comes from me doing research as interesting scenarios come up,” she added.

How to Read a Patent

You don’t need a law degree to be able to read a patent, but you might need a few tips to read them efficiently.

First, don’t read from top to bottom. Focus on this order instead:

The Abstract: Read this first. It’s a 150-word summary of the problem and the solution. If the Abstract mentions “improving resource relevance based on user context,” you’re likely looking at a ranking signal.
The Claims: Found at the very end. The claims define exactly what Google owns. If a specific ranking factor isn’t in the claims, they don’t legally own that specific method.
The Drawings & Flowcharts: Look for Fig. 1 or Fig. 2. Google engineers use these flowcharts to show how data moves from a user’s query to the final result.

Keep in mind that terms like “entity” or “resource” are used as catchalls for a person, place, or a brand to keep the patent’s scope broad. Remember, Google purposely keeps their patents from being too specific.

“I mentally strip legal padding and reduce everything to inputs, transformations, scoring, and constraints,” Metehan said. “Constraints (latency, decay, thresholds, fallbacks) usually matter more than features.”

A patent does not always equal a finished product either. Google patents thousands of ideas that never see the light of day as an actual product. In fact, Google frequently files “defensive patents,” which are ideas they may never use, to prevent competitors or what they refer to as “patent trolls” from using them.

We know that the language isn’t simple for everyone to understand because it’s written through the lens of information retrieval and not SEO (which is why Mike says they’re able to get away with obfuscating information), but we now have tools to help us make sense of these patents. You can upload the PDF of the patent to NotebookLM and ask it questions or generate a podcast explaining it.

Gemini has obviously indexed all of Google’s research so you can ask it questions, too. If the information hasn’t been made public yet, you can at least give it whatever information you have and it can analyze what it all might mean based on the facts.

“It’s all a lot more accessible now than it’s ever been,” Mike said.

Sometimes, Google will publish updates or expansions, so make sure you’re looking at the right date. A patent usually lists three:

Priority Date: The most important date. This is when Google first claimed the idea. It tells you exactly when they were thinking about this technology.
Filing Date: When the formal paperwork was submitted (often months or years after the Priority Date).
Publication Date: When the patent became public.

If you see a patent published in 2026, but the Priority Date is 2021, you aren’t looking at a new invention but something Google has likely already been testing for five years.

The amount of patents Google applied for has dropped 13% since last year, though, which is interesting to see. The company was granted 1,782 patents in 2025, down from 2,054 in 2024.

Aufgesang’s Head of SEO and AI Search Olaf Kopp monitors new patent publications from Google, Microsoft, and OpenAI every week and manually filters out those that could be interesting.

“My analyses focus primarily on general information such as status, countries/regions where the patents were published, claims, and detailed descriptions,” he said. “Patents that have been published for multiple regions are more exciting, as it is more likely that the methods will also be used in practice.”

Navigating Google Leaks

In May 2024, over 14,000 internal ranking features from Google’s Content Warehouse API were released by an automated bot. Mike King was one of the first to see these leaked documents after Rand Fishkin shared them with him (after Erfan Azimi shared them with Rand).

Finding these leaks is only half the battle, though. The real skill is knowing how to read them without jumping to false conclusions.

How to Verify a Leak

To ensure what you’re hearing about Google is true, you could cross-reference with official sources such as court testimony, like how the DOJ vs. Google trial confirmed the existence of Navboost and Glue before the leak even happened.

You could also look for deprecation markers. Some leaked features are marked as “deprecated,” meaning Google used to use them but doesn’t anymore, which is crucial to avoid chasing ranking factors.

“Leaks are partial snapshots, not documentation,” Metehan said. “I map internal terms, look at defaults, check pipeline order, and cross-validate against real-world behavior and older patents. I never treat leaks as a checklist.”

Metehan also warns against optimizing purely for patents or leaks.

“I look for invariant behaviors that survive multiple re-rankers, versions, and query types and test those,” he said. “I’m using Gemini APIs a lot, models, embeddings, layout parser, model garden and Vertex AI with grounding, etc. It’s a multimodal world.”

Google has been sharing pretty much everything research-wise these days, so it’s up to us to find what’s important. Because when it comes to Google’s official information releases, that’s a different story.

“Google has been far more forthcoming about how things work on the research side than they have on the PR side to the SEO community,” Mike said.

Reading Between the Lines of Google PR

Official communications from Google (via their blogs, social media, or spokespeople) typically serve a specific purpose: to encourage behavior that makes their search engine better while discouraging “gaming the system.”

To find the truth, you have to decode the Google-Speak.

Google often uses a linguistic technique where they deny a specific term but not the concept to make it “technically truthful.”

For example, multiple Googlers like John Mueller and Gary Illyes have said that Google does not use a Domain Authority score. But the 2024 API leak revealed a metric called siteAuthority.

Google was “truthful” because “Domain Authority” is a trademarked term by the SEO company Moz. They don’t use that specific tool, but they certainly use their own internal version of the same concept.

And most recently, Danny Sullivan declared that no one should be “chunking” their content – a statement that didn’t really make much sense because it’s been an SEO and UX best practice for years to break up walls of text into easy-to-read sections.

But again, Google doesn’t use the specific word “chunking” when making content recommendations so, sure, I guess they’re technically being truthful. It still seemed like a silly statement to make, though, and got a lot of people up in arms.

These are the times when it’s important to think hard about why they’re saying these things and how it all would benefit Google financially, because that’s Google’s main concern.

When reading a release about Google, ask yourself these questions:

Is the denial hyper-specific? If they say “We don’t use [Metric X] for [Task Y],” they might still use it for Task Z.
Does it contradict sworn testimony? Always prioritize information from the DOJ vs. Google trial over a tweet. In court, off-the-record chats are evidence while on X (Twitter) or Reddit, they are PR.
Are they saying to do as they say and not as they do? Google PR often tries to scare people away from manipulation and those evil Black Hat SEO methods like buying links, even when their own systems like PageRank still rely on them.

Always keep your reporter hat on when listening to Google, even if you never actually spent three days driving around town trying to get a comment from an avoidant local politician for $12 an hour like some of us did.

Now, where can we find official information about Google?

Reliable Primary Sources for Official News

To analyze the spin yourself, go directly to the source:

The Search Central Blog: For official algorithm naming and policy changes.
The “Search Off the Record” Podcast: Where engineers often let slip technical nuances that don’t make it into the formal blog posts.
Sworn Court Exhibits: The Justice.gov Antitrust Exhibits are a big source of truth considering that lying there has legal consequences.

“It’s a combination of research and evidence,” Mike said. “You also have to be mindful that when Google is speaking to the community, they’re speaking to the average person, not the experts. They’re creating something that answers the question for most people.”

Mike emphasized that there are three different things to look at to figure out the actual truth: your research, what’s in the wild, and what Google is saying.

Googling the Truth

Overall, just remember that Google is a massive corporation that is not here to help SEOs be more successful and get more clients. Google is here to continue to be massive and rich. Everything they say should be viewed through that lens, no matter how much their statements happen to align with a point you’re trying to make at that moment.

Also, none of these sources are the whole truth on their own. Look at multiple sources, official documents, legal documents, and general research to come to a conclusion.

“Patents don’t tell you what to do,” Metehan said. “They tell you what Google struggles with and cares about. That’s where the real insight is. Think from a reverse and relevance engineering perspective.”

As we move further into 2026, where still-pretty-inaccurate AI Search has become the norm, these detective skills will be more valuable than ever. Google may obfuscate the truth to protect its bottom line, but the paper trail left by their engineers and leaks will always provide a roadmap for those willing to look.

Don’t just wait for the next Helpful Content update to tell you what to do. Go beyond the headlines and find the real story yourself.

Need help becoming more visible online?

Read about Relevance Engineering

The post Digital Sleuthing: How to Find and Decipher Google’s Patents, Leaks, and PR appeared first on iPullRank.

1 Important Lesson Omar Little Taught Me About Writing Page Titles

Eric Gilley — Thu, 22 Jan 2026 12:00:00 +0000

I got my start in content by editing articles for a site that lived and died by clicks. We didn’t use the word “clickbait” (not out loud), but that’s exactly what it was. I’m not proud of headlines like “3 Frustrating Toyota Tacoma Problems That Send Drivers Into a Rage” or “Aaron Rodgers Sends Packers Fan a Clear Message With Latest Podcast Appearance”, but hey, I had to pay the bills.

And in the wise words of Omar Little from the TV show The Wire:

Clicks equaled money, and making people feel something was how you got them to click. That was the incentive. That was the game.

These days, we’re seeing a new game forming. The incentive structure is shifting.

For many types of content, visibility now comes from being useful to something that isn’t human at all; not a reader, but an amalgamation of three different LLMs and a swarm of search, retrieval, grounding systems, summarizers, re-rankers, and answer engines I just (incorrectly) lump under the catch-all: “AI.”

It doesn’t feel outrage. It doesn’t get curious. It just scans, slices, and decides in a fraction of a second whether your page contains an answer. And what it sees first, namely your title, url slug, and a short text snippet, is the first and most decisive signal it has to go on. If those don’t scream relevance, you probably don’t get cited.

This means the headline’s job isn’t to tease; it’s to plainly declare what you know and who you’re useful to. It’s not clickbait; it’s AI bait. And once again, the game is rewarding those who know how to write to the system.

How an AI "Sees" Your Page: The Snippet-First World

Contrary to what many assume, an AI doesn’t “see” your webpage in its entirety. It doesn’t load the full HTML, parse the layout, or admire your design. Instead, As Dan Petrovic explains in How GPT Sees the Web, its first interaction is with a small, ruthlessly efficient packet of data: the page title, the URL, and a short text snippet.

When an LLM is grounding an answer via search, this is the initial context it’s given to decide whether your page is worth retrieving at all.

This is the core idea behind what we at iPullRank call Relevance Engineering: intentionally shaping content so its meaning is legible at the exact moment a system is deciding whether your page is useful. Before design, before UX, before persuasion — relevance has to be established.

Think of it as a high-stakes, single-chance audition. If that preview doesn’t scream “I have the answer!” for the user’s query, the AI will logically and instantly move on to a source with a stronger, more relevant signal.

It is an automated process of elimination designed for maximum efficiency. The AI’s decision to “click” and investigate your page further is won or lost in this initial glance, making the quality of that preview paramount.

Not to mention, even when a page is retrieved, the model still isn’t consuming it all at once. It’s evaluating the content in constrained windows, pulling only the portions most likely to resolve the question at hand.

Tools like Relevance Doctor exist to diagnose this exact moment: what signal your page sends in that first interaction, and whether it’s strong enough to survive the cut.

From Human Clickbait to "AI Clickbait"

Over the years as the web evolved, headlines changed as well.

You can think of headline evolution like this:

Early SEO (Late ‘90s to late ‘00s): Keywords signaled topical relevance (“Fixing Crawl Budget Issues”)
Content marketing era (Early to mid 2010s): Curiosity and emotion drove clicks
Social era (Late ‘10s to early ‘20s): Outrage and intrigue maximized engagement
AI Search era (today): Explicitness and semantic density determine inclusion

The incentive hasn’t changed. Visibility still rewards alignment. What’s changed is who’s doing the selecting.

You can see this incentive shift play out in real traffic patterns.

Below is an anonymized traffic chart from a content site that leaned heavily on curiosity-driven, emotionally charged headlines. For a few years, that approach performed extremely well by optimizing for human behavior: intrigue, open loops, and the promise of a payoff after the click.

That strategy wasn’t wrong. It was perfectly aligned with the distribution systems of the time. But as discovery began to move upstream into AI-powered summaries, re-rankers, and answer engines, the same ambiguity that once drove clicks became harder for machines to interpret.

When the selector changes, the headline has to change with it.

That same snippet-first, elimination-driven efficiency leads to a new way of thinking about titles and snippets: a concept we’ll call AI Clickbait. But unlike its human-focused cousin, it’s about radical clarity.

Let’s look at a real-world example of an old-school title and see how it would have to change for today’s game.

To do that, we’ll pull out a certified banger from the archives, written by none other than the founder of iPullRank, 2025 Search Marketer of the Year, and my boss (please don’t be mad at me): Mike King.

This 2016 headline was a masterclass in playing to the human incentive:

“How I Sped Up My Site 68% With One Line of Code.”

Certified banger. I want to click that right now.

But that headline would for sure be ignored by an LLM. A machine isn’t curious; it’s a logician. It can’t process the intrigue. The title offers a compelling result but hides the actual subject matter, giving the AI nothing concrete to latch onto.

For today’s AI-first world, you’d have to flip the formula and close that information gap instantly:

“How to Improve Page Speed Using the rel=prerender HTML Attribute”

Definitely not as catchy. However, that’s the version you write when the reader is a summarizer, not a person. Just topic, method, and outcome.

Human-Optimized	AI-Optimized
“This One Line of Code Changed Everything”	“How to Improve Page Speed Using rel=prerender”

Here are a couple of examples from the BBC news site. Do either of these inspire you to click and find out what they’re talking about?

You’re not trying to trick the AI; you’re trying to give it the most efficient signal possible. It proves that the core of the game remains the same: you have to know your audience. The only thing that’s changing is who (or what) you’re playing to.

Is It Time to Rethink Title Tag Length?

For years, SEOs, editors, and writers were taught to treat 60 characters like gospel. Not because Google cared about brevity, but because that’s where the snippet cut off in search results. Anything beyond that risked getting truncated.

In this guide to page titles created by Screaming Frog, these are the recommended title length maximums and minimums:

Title Maximum Length – 580 pixels or 60 characters.
Title Minimum Length – 200 pixels or 30 characters.

But AI doesn’t operate under a visual pixel constraint. It’s not laying out a search results page. It’s parsing meaning. Truncation doesn’t matter as much anymore; data loss does.

As written in this article by Krishna Madhaven at Microsoft, “Your page title, description, and H1 tag (the top-level HTML heading) are important signals AI systems use to interpret purpose and scope.”

The real risk isn’t your title getting cut off. It’s your most important concepts never making it into the snippet that gets passed to the LLM.

Take an old-school headline like:

“Fixing Crawl Budget Issues” (26 characters)

It’s short and technically accurate, but semantically flat. Instead try:

“How to Fix Crawl Budget Issues That Prevent Googlebot from Indexing Your Site” (77 characters)

It’s longer, but also denser: how-to intent, crawl budget, Googlebot, indexing problems. It gives the AI more reasons to believe your page directly addresses the user’s query. Not just “this is about SEO,” but “this is about exactly the issue you asked about.”

In a small internal analysis of one site, we compared high-ranking pages that were cited by AI systems against high-ranking pages that were not. While both groups performed well in traditional search, pages that were cited tended to use longer, more semantically dense titles on average.

This doesn’t prove that longer titles guarantee more AI visibility. But it does suggest something worth testing further: when relevance is evaluated programmatically, titles that surface more context upfront may have an advantage.

The takeaway isn’t “write longer titles.” It’s to stop treating length as the constraint and start treating meaning as the variable that matters.

The goal isn’t to jam in keywords; it’s to front-load meaning.

How to Craft a Headline for an AI-First Interface

Writing an AI-legible headline is about removing ambiguity at the moment of evaluation. Screaming Frog suggests following these best practices for page titles:

Concise
Descriptive & Relevant
Unique
Includes Brand
Optimal In Length
Enticing

A practical way for marketers to approach it:

Start with the question your page answers. If you can’t phrase it as a question, the page probably isn’t answer-shaped yet.
Name the core entity and problem explicitly. Don’t assume prior context.
Front-load meaning, not modifiers. Lead with what the page is about, not how impressive the result is.
Reduce intrigue that hides subject matter. Curiosity gaps work on humans; they slow down machines.
Sanity-check for isolation. If the title were stripped of its surrounding page, would its purpose still be obvious?

As Krishna writes, “Page titles should clearly summarize what the content delivers, using natural language that aligns with search intent.”

A New Title Length Best Practice Worth Testing

This isn’t about throwing out everything you know about SEO. It’s about recognizing that the rules of visibility are evolving, and the audience is changing.

You’re no longer writing just for humans. You’re also writing for models that don’t scroll, don’t skim, and don’t guess. They parse titles, weigh snippets, and choose what to cite based on how clearly your content signals that it’s an answer.

But let’s be real: not every headline needs to be AI Clickbait. If your content is built for humans (email, social, or storytelling) then cleverness still has its place. A good curiosity gap still works where curiosity lives.

The point isn’t to kill creativity but to optimize for the interface you’re speaking into.

When the goal is LLM visibility, AI Overviews, or zero-click citations, semantic clarity wins. In those cases, your headline isn’t a hook. It’s a claim.

So experiment. A/B test longer, clearer titles. Revisit old blog posts that rank well but get ignored by AI search. Give your best content a second chance by making it unmistakably answer-shaped.

Because like Omar said, it’s all in the game. And these days, the game picks the clearest answer.

Want to learn more about optimizing for LLMs?

Read our AI Search Manual

The post 1 Important Lesson Omar Little Taught Me About Writing Page Titles appeared first on iPullRank.

Moving from a Google-shaped Web to an Agent-shaped Web: A Refutation of Misinformation about Chunking

Mike King — Thu, 15 Jan 2026 12:00:00 +0000

Back in 2011 when I first started writing SEO blog posts for Moz, despite their popularity I was writing walls of text because that was my nature. Then-CMO Jamie Steven instructed me to read Cyrus Shepard’s 10 Super Easy SEO Copywriting Tips for Improved Link Building for direction on how I should structure what I write for better performance.

In the article, Cyrus comes out swinging showing this visual comparison of a wall of text versus a very well-structured piece of content with lots of formatting and imagery. Using data to drive the point home, he shows how the two posts (by the same great internet marketer) had dramatically different performance, with 62X the external link capture and nearly 4X the time on page.

I was hooked and those insights have stuck with me ever since. In fact, you can trace back elements of anything I’ve written over the last 14 years to the formatting lessons of that classic post. I’d go as far as to say I think more about these principles than I do so-called SEO “best practices.”

Part of why what Cyrus outlined resonated with me so much is that the principles just make sense. Conceptually, it all harkens back to everything we all learned about how humans interact with information when we read “Don’t Make Me Think.” Over time, I’ve seen the specificity and better content UX highlighted yield better performance on any human-driven metric we measure as well as more visibility search engines and large language models.

But…Google Says Don’t Break Your Content Into Bite-Sized Chunks

Recently, on the Search Off the Radar podcast, Danny Sullivan shared his opinion on “chunking” as a tactic to drive visibility in AI Search surfaces (emphasis mine).

“One of the things I keep seeing over and over in some of the advice and guidance and people are trying to figure out what do we do with the LLMs or whatever, is that turn your content into bite-sized chunks, because LLMs like things that are really bite size, right?

So we don't want you to do that. I was talking to some engineers about that. We don't want you to do that. We really don't. We don't want people to have to be crafting anything for Search specifically. That's never been where we've been at and we still continue to be that way. We really don't want you to think you need to be doing that or produce two versions of your content, one for the LLM and one for the net.”

After I laughed to myself in the graveyard of AMP POVs and technical specifications, I turned it back on.

He continued, pre-empting the “but it works, I’m going to do it anyway” argument Danny offers (emphasis still mine):

“Let's assume that, in some edge cases, let's even assume maybe in more than some edge cases, you're finding you're getting some advantage here. Maybe tiny degree measure. No, this is my secret weapon. It's doing it." Great. That's what's happening now. But tomorrow the systems may change.

So you've gone through all this effort. You've made all these things that you did specifically for a ranking system, not for a human being, because you were trying to be more successful in the ranking system, not staying focused on the human being. And then the systems improve, probably the way the systems always try to improve, to reward content written for humans. All that stuff that you did to please this LLM system that may or may not have worked, may not carry through for the long term.

So was that the best use of your time and your energy? Was that the best use of putting turmoil into your marketing department, your content department, and all your other stuff so that you could say, "A-ha, I've got the new thing that you wanted, I've brought it down from the mountain and here it is. Do these sorts of things.”

Before I take this where you know I will, let me first say this.

I deeply respect Danny Sullivan for what he has done for the search marketing community both inside and outside of Google. Full stop.

However, I have two problems with these statements and want to clarify for anyone who is questioning the value of improving content structure (partially) in the service of better visibility:

Chunking and creating content for users are not mutually exclusive.
The statements misalign with how Retrieval Augmented Generation technology functions and with where the future technologies are going.

In the spirit of chunking, let’s add a heading and get to my next series of extractable atomic points.

Chunking and Writing for Users is Not Mutually Exclusive

First, we need to disambiguate “chunking” from how it’s used to describe an operation in Retrieval Augmented Generation (RAG) systems from how it’s being used to describe a content optimization action.

Chunking as it has been co-opted is really structuring content in a way that its passages and statements perform better when retrieved in a RAG pipeline. I know this because I’m one of the first people to drag the term from the AI/IR space into the SEO space.

If we’re being reductive (like most “it’s just SEO” arguments are) we’re effectively talking about content design or UX writing. As with everything in search and content marketing, machines are just a subset of the target personas. So, the idea of preparing the content only for the machine solely is still nonsense.

When we’re talking about chunking in this sense, the act of structuring content overlaps with the content design aspects of Cyrus’s post. However, where it differs is in reasoning of the decisions that you make in the copy that you write.

While the act overlaps with the largely qualitative processes people have historically used in the past, it is not the same. Effective chunking follows all the practices Cyrus discussed, but combines vector analysis to verify improvements. Also, for clarity, chunking is but one of an array of tactics you should apply from the content engineering toolbox. And, what differs it from UX writing or standard copywriting is the aspects of relevance calculation that must be accounted for on a passage level. It’s not just chopping paragraphs into smaller paragraphs and using more headings and hoping for the best.

Consider this, no one can reliably determine that content is generative or not without watermarks. So, Google needing a more reliable signal leverages user interactions to determine whether content should continue to rank. The main attribute that yields better content performance is better structure irrespective of why you do it.

So, What Truly is Chunking?

Chunking is the action that RAG systems take with content when they capture it to prepare it for the retrieval process. Chunking is the act of breaking content into a series of components that can be individually retrieved based on how relevant they are to a prompt or user query.

This is also a function of dense retrieval which Google effectively announced when they revealed their implementation of Passage Indexing. In passage indexing, passages are embedded and stored and the query is too. Approximate Nearest Neighbor (ANN) searches are performed to pull the closest matching passages. This is one of the building blocks of RAG, the primary paradigm behind AI Search.

There are a variety of chunking strategies including, but not limited to semantic, layout-aware, fixed length token-size. Based on Metehan’s research into Google’s public Vertex AI offering, it suggests that theirs may be a combination of fixed length and layout aware with the cascading heading option.

So, we are using the same term to refer to both the action that the system takes to decompose content and the work that we’re doing to restructure the content so it’s easier to extract. I wanted to clarify that for people that look to invalidate meaningful discussion based on syntax and vocabulary.

Why is Chunking Different from Classic Content Optimization for SEO?

In classic SEO, content boundaries were defined largely by intuition and editorial convention. Even when using content optimization tools, the analysis was typically lexical and page-level, evaluating aggregate term usage rather than the relevance of individual passages. As a result, while pages may have been visually or structurally segmented, those segments were not deliberately optimized as independent units of meaning.

Chunking as an optimization tactic changes that. With clearer insight into how modern systems evaluate content at the passage level, we can now treat each chunk as a discrete relevance object. This makes it possible to intentionally shape structure, specificity, and context within each passage to influence how it is measured, compared, and selected. Instead of optimizing pages holistically and hoping relevance emerges, chunking allows us to precisely adjust content at the level where relevance is actually computed.

Insights that Dan Petrovic shared on the length of Google’s grounding chunks and how much of your content gets used after it makes it through filtering give us more clarity on the atomicity. We also know that the natural boundaries of a paragraph we create influences what is considered a chunk in these systems.

Historically, SEO treats the page as a single context window with no real measurable way to tell if your optimizations really did anything except for the rankings themselves. Sure, the various content optimization tools give you a lexical score, but nothing that aligns with the breadth of modern information retrieval. Chunking offers a direct feedback loop for the isolation and improvement of specific blocks of content and how you can influence how they perform in AI surfaces.

The Google-shaped Web

Much has been said about how the web has conformed to what performs best in Google. It’s expected when Google is the biggest referral channel. However, with the advent of generative AI, websites are no longer adapting to a single set of guidelines or incentives. Content is now shaped by multiple platforms and channels, including search engines, AI assistants, recommendation systems, and social feeds, each imposing different structural and semantic pressures on how information is created and how users react to it.

After two decades of being in the space, I can say definitively that statements like this are how Google keeps marketers as their unpaid workforce, nudging the web toward what works best for Google.

Googlers often speak as though they are merely extracting natural patterns from the web, positioning themselves as neutral observers. But they are not the Watchers. They are the Celestials. One watches without interference; the other designs systems that determine what survives. Google’s ranking and retrieval decisions have shaped the web for decades. Entire categories of sites have converged on similar layouts, headings, FAQs, and explanatory formats not because those patterns emerged organically, but because Google’s systems and PR consistently reinforced them.

It’s not that they “don’t want people to have to be crafting anything for search specifically.” It’s that they “don’t want people to have to be crafting things for search that take advantage of Google.”

What changes with generative AI is not that this influence goes away, but that it begins to fragment. Search is no longer only about ranking pages. It is about selecting, extracting, and recombining passages across sources. The incentives now favor content that can be easily segmented, understood in isolation, and reused by machines.

This is still a Google-shaped web. But the shape is starting to loosen, creating the conditions for what comes next.

Google still has search, but the agent-shaped web is emerging outside their control

As generative AI becomes a primary interface for information, the incentives that once forced publishers to conform to Google’s preferences are weakening. Users are getting answers without clicks, referral traffic is less reliable, and the payoff for strict adherence to SEO best practices and Google’s guidelines continues to shrink – even when Google results are a key input for the results. The result is a gradual but meaningful loss of influence over how content is structured and prioritized.

(sidebar: I’ll have you know I wrote that em dash myself in that last paragraph.)

In conversations with F100 clients, this shift shows up clearly. A few are moving towards abandoning search outright, but many are questioning how much effort it still deserves. Investment is spreading to other formats and channels, and teams are becoming more willing to deviate from rigid SEO conventions. Not because best practices are “wrong,” but because repeated testing shows their impact is increasingly marginal.

What’s emerging is an agent-shaped web. Content is no longer written primarily to satisfy a single ranking system, but to be usable by agents that retrieve, reason over, and recombine information across sources. These non-Google systems do not publish guidelines. They do not moralize tactics as “white hat” or “black hat.” They simply use the content that works. In that environment, many behaviors Google historically discouraged are not violations. They are advantages.

This is how Google’s grip loosens. When influence shifts from ranking pages to supplying agents with usable inputs, control fragments. The web stops optimizing for compliance and starts optimizing for utility.

That’s why, when I’ve asked Google engineers what to do beyond “make great content” to improve rankings, the answer has consistently been “nothing.” That response only holds if Google remains the central force shaping outcomes. In an agent-shaped web, it isn’t. And, that’s why you see them creating fast-follow protocols like A2A after MCP and UCP after ACP.

How Chunking Improves Relevance

We know that content structure influences people and they should be the primary audience for any content adjustment. Fundamentally though, Danny’s statements do not align with how the underlying technology functions.

Search engines and Large Language Models are both built on the vector space model. Relevance is a function of distance measures between queries/prompts and documents. Where search engines measure this to rank documents, LLMs use the plotted relationships to predict the next token.

The distance measures are the values that are compared to determine what to feed the LLM. In synthesis pipelines, there is a pairwise determination where passages are compared side by side to determine what gets sent to the language model. A longer piece of text that covers multiple subjects typically has lower relevance than a shorter piece of text that covers a single subject.

Let’s illustrate that idea with an actual example in my tool BubbaChunk (if you know you know).

The paragraph above targets the queries [machine learning] and the [data privacy]. When I generate embeddings for the queries and for that paragraph, using cosine similarity as my distance measure I get a 0.541 for [machine learning] and an 0.620 for [data privacy].

Now, let’s split that paragraph in two and not change anything else. The machine learning paragraph has now improved 19.24% to a 0.645 cosine similarity. The data privacy paragraph improved 1.29% to 0.627.

When compared to passages on both subjects, it now has a better opportunity to perform. Within an environment of a full page, other elements like the heading hierarchy and surrounding passages can be used to influence this. I can further improve the scores with semantic triples, entity salience, and so on, but in isolation, restructuring this content by changing its boundaries improves its retrievability.

Some folks may be invested in Google’s multi-aspect embedding technique MUVERA. BubbaChunk takes a similar approach and MUVERA uses Chamfer Similarity as its distance measure. Those are the Chamfer values you see in the screenshots. You’ll note that there are improvements to all distance measures when I’ve made this adjustment.

If you’re curious, adding the headings does improve the scores significantly. Below you’ll see adding the header to the “Data Privacy” paragraph improved cosine similarity another 17.54%.

No matter how you slice it, embed it or measure it, improving the structure of content yields better scores by machines and how it performs with people.

Structured Content is Better in Any Paradigm

Danny’s comments suggest that Google may eventually evolve its systems to discourage overt structuring techniques. That assumes structure is a temporary optimization tactic andsuggests we’re moving to a world where “high-quality writing” is a monolith that the algorithm will simply “understand.” Google’s own research direction, alongside adjacent work from Meta, Berkeley, and MIT, points in the opposite direction. As systems gain access to more context, memory, and recursion, structure becomes more important, not less. Across multiple papers, Google Research is clearly pursuing near-infinite context through memory rather than brute-force attention, and they are building atop the state of the art from other groups in the space.

Reviewing the state of the art, we find that Berkeley’s Ring Attention demonstrates how extremely long sequences can be processed by breaking them into rotating segments, where each segment attends locally while passing the summarized state forward. In that structure, the model does not need to see everything at once. It needs to preserve meaning across time. Systems like this rely on the continuity of information within a segment. By structuring content into atomic passages, you ensure each “rotating segment” contains a complete, unfragmented unit of meaning.

“Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention” formalizes this further by introducing compressive memory that allows models to retain and reuse information far beyond a fixed context window. You can’t compress a mess without losing the message. Atomic, legible passages act as high-fidelity signals that survive the compression process, ensuring your information is correctly retrieved later.

Meta’s MemWalker pushes in the same direction by organizing content into memory trees that can be traversed, revisited, and updated. Structure provides the branches. By defining clear boundaries and semantic anchors, you build a “map” that helps the agent navigate and reconstruct the mental model of your information.

These approaches make Google’s intent clear. The goal is not just larger windows. It is durable, near-infinite context.

(sidebar: Those last 4 paragraphs were originally a single paragraph. I split them up while editing to isolate each specific idea and align them with the images from the papers. That’s an example of chunking in action.)

MIT’s work on Recursive Language Models reaches the same destination through a different path. Rather than expanding context directly, RLMs decompose long inputs into smaller units and recursively invoke the model over the most relevant chunks. In effect, the model reasons over content iteratively, revisiting and recombining passages as needed. This reinforces the same reality. Passages are the unit of interaction.

DeepMind’s Mixture of Recursions (MoR) pushes this idea into the architecture itself. Instead of a fixed depth of computation, tokens are routed through different numbers of recursive steps. Some content is processed shallowly. Other content is revisited repeatedly. This is adaptive reasoning at the token level, and it further removes any illusion that content is consumed linearly or holistically. What matters is which pieces survive repeated passes through the system.

A common rebuttal is that we are entering an era of “Infinite Context.” With models like Gemini 3 Pro with a 1 million token context window, why bother chunking when the model can ingest the whole book?

The answer lies in inference cost and reasoning depth. The MoR paper reveals that not every token needs the same amount of “thinking.” In an agent-shaped web, computation is the new scarcity. Well-structured, atomic content allows the model’s ‘router’ to identify meaning quickly and exit the recursive loop early. Brute-forcing an unstructured 2-million-token wall of text is computationally expensive and prone to ‘context rot.’ If you want an agent to pick your content over a competitor’s, you should make it the path of least resistance. You don’t want to be just readable, but computationally efficient to digest.

At the bleeding edge, Google’s Nested Learning moves beyond retrieval entirely. With the HOPE architecture, passages are no longer just fetched as context. They are used as signals for “memory infusion,” updating the model’s inner loop. This is where control erodes most clearly. Once content moves from retrieval into synthesis and memory update, our influence largely ends. Just as we can influence how we rank for synthetic queries but not how answers are composed, we can influence which passages are legible and extractable, but not how they are ultimately weighted, combined, or remembered.

In this environment, atomic legibility is the only survival strategy. If a passage isn’t self-contained (meaning it lacks its own entity, context, and claim) it fails to “infuse” correctly. It becomes noisy data. Just as Infini-attention relies on “compressive memory” to store long-term state, your content must be compressible. You cannot compress a mess without losing the message. Each chunk must stand as a standalone signal so that when the agent tears the binding off so to speak, the individual page survives the transition from retrieval to synthesis.

Furthermore, the shift to an agent-shaped web isn’t limited to text. Agentic systems are increasingly multimodal, needing to reconcile text with images, charts, and tables. Without layout-aware structure, these relationships disintegrate during the retrieval process. By defining clear boundaries and semantic anchors, we aren’t just helping the model read; we are helping it reconstruct the mental model of the information. Structure is the glue that ensures a chart and its context remain unified when an agent retrieves them from a near-infinite context window.

None of this weakens the case for structure. It sharpens it. In every one of these systems, passages remain the atomic unit of meaning. Whether through attention, memory, or recursion, models operate on chunks, not pages.

Google still wants us to produce books: pages that can host ads, preserve attribution, and sustain the economics of the open web. Agentic systems read differently. They tear the binding off the book, ignore the table of contents, and pull out only the pages and paragraphs they need, sometimes revisiting them again and again. In that world, structure is no longer about presentation. It is about making meaning legible at the passage level.

Across every paradigm, from the first Google-shaped web to the looming agent-shaped web, the best we can do remains the same. But the ‘why’ has changed. We are no longer just formatting for ‘skimmability’ or ‘dwell time.’ We are formatting for Programmatic Legibility. We are building the API of meaning. By designing content so each chunk stands on its own with clear signals, we aren’t performing a workaround. We are ensuring our information survives the recursive, synthetic, and agentic loops that are now defining the web.

We’ll Keep Being the Signal through the Noise

There’s a lot of confusion right now as search and generative AI continue to blend. There are also a lot of people who don’t understand the nuances so they keep looking to shoehorn the changes into what they already do and know so they can feel superior or at least relevant.

This also makes it difficult for the community because the only reliable sources of information come from reading patents, white papers, playing with the platform’s public APIs, and then experimenting to see what works. Not everyone is capable of those things, nor do they have the time or wherewithal, so they do what I said above.

It’s unfortunate that Google continues to want to play the FUD game and contradict what we can see with our own eyes in their research, patents, and in how their systems react to changes. That behavior further reinforces that we are not partners in making the world’s information accessible. We are the unpaid extension of their workforce.

For these reasons, iPullRank will continue to do the work and the R&D and share what really works and why. Our AI Search Manual is an example of that. We’ll continue to support everyone striving to build in this new world and this will be one of the threads we continue at SEO Week in April.

So, get your ticket to SEO Week and hear from the sharpest minds on what’s actually working for AI Search.

Want to find out about how Relevance Engineering can help your business?

Learn about iPullRank's Relevance Engineering Services

The post Moving from a Google-shaped Web to an Agent-shaped Web: A Refutation of Misinformation about Chunking appeared first on iPullRank.

Query Fan-Out in Practice: Turning One Search into an Omnimedia Content Plan

Francine Monahan — Thu, 08 Jan 2026 12:00:00 +0000

If you’re a marketer, you’ve probably invested heavily in content marketing over the years, but the Organic Search payoff isn’t delivering, and you’re questioning your SEO strategy.

For example, for those in the finance industry, your retirement planning guides might rank pretty well, your pages explaining investment strategies get decent traffic, and your calculator tools actually help people. But maybe lately, something feels off.

Your rankings are holding steady, yet Organic Search traffic keeps declining. Is this the impact of AI Search?

Maybe you’re barely showing up in AI Overviews or AI Mode results for topics that are core to your business goals, and when potential clients ask ChatGPT or Perplexity about retirement planning, your firm doesn’t get mentioned at all.

The problem might be that you have an outdated perspective on SEO. You’re familiar with optimizing for single keywords, but that’s not how it works in AI Search. AI Search Platforms are leveraging query fan-out.

When someone searches for “What’s the best way to save for retirement,” Google’s Gemini is simultaneously running dozens of related searches for AI Overviews, such as 401(k) contribution limits, Roth IRA comparisons, retirement calculators, savings benchmarks by age, and common mistakes to avoid. If your content doesn’t include passages that satisfy your search across all of these relevant sub-queries, you’re losing visibility, brand awareness, and performance metrics.

So what’s a better, more modern strategy for AI Search? Let’s approach this problem using a fictional mid-sized financial services firm as an example. I’ll walk through the complete process: map out fan-out queries, analyze what actually ranks and gets cited by AI (including which formats and sources dominate), and build an omnimedia content plan. You can’t only worry about your owned properties, like your website. You need to cover every channel that contributes to AI Search results: YouTube, Reddit, industry trades, and every other channel where your prospects are searching.

Let’s get started.

What Is Query Fan-Out and Why Does It Matter for SEO Now?

Today, your content might rank on page one for your target keyword and still be invisible in AI Search results, and a lot of that has to do with query fan-out.

Query fan-out is the process by which AI systems and modern search engines expand a single user query into related sub-queries that run simultaneously in the background. Rather than treating your search as an isolated request, these systems interpret it as a starting point for exploration.

When you ask Google a question in AI Mode, what appears to be one search is actually dozens or even hundreds of synthetic queries working in parallel. As Mike King explained in his detailed technical analysis, these include:

Related queries that are semantically adjacent to your original search.
Implicit queries that capture what you likely meant but didn’t explicitly state.
Comparative queries that help you make decisions between options.
Personalized queries tailored to your search history, location, and behavioral patterns.

Mike’s research into Google’s patent applications revealed that it uses large language models like Gemini to generate these synthetic queries through structured prompting that emphasizes intent diversity, lexical variation, and entity-based reformulations. (If this feels like too much jargon, revisit Chapter 6 of our AI Search Manual, where we discuss some of these fundamentals that influence information retrieval.)

Modern AI Search systems use query fan-out to understand and satisfy your full search intent.

This approach enables AI systems to:

Synthesize comprehensive answers by pulling all relevant information.
Build reasoning chains that connect different aspects of a topic logically.
Anticipate follow-up questions before users ask them.
Diversify information sources to avoid relying on a single perspective.

The query fan-out process helps ensure that AI-generated responses are robust, multi-dimensional, and genuinely helpful.

Helping a Financial Brand Win Retirement Searches

What’s the practical application?

Let’s say you’re working with a mid-sized financial services firm specializing in retirement planning and investment management.

Their primary business goal is to increase qualified leads from people actively researching retirement planning options. To achieve this, they need to earn more visibility in search than direct and indirect competitors.

This company’s core audience persona is a mid-career professional in their 40s, earning over $90K annually, who has been contributing to a 401(k) for years. They have a couple of kids who are in high school and parents who have retired, so now they’re thinking about their own financial future.

The core audience also:

Have accumulated some retirement savings but lack confidence in their strategy.
Feel overwhelmed by financial jargon and competing advice.
Want personalized guidance but aren’t ready to commit to a financial advisor.
Increasingly turn to AI Search tools for quick answers to complex financial questions.

This person doesn’t search for “financial advisor near me” or “best IRA provider” (at least not yet). They start with broader, more exploratory questions.

So, we’re going to ask a broad question to start our search.

The Query

For this analysis, we’ll focus on a query that represents the prospective customer’s starting point in the retirement planning journey:

“What’s the best way to save for retirement?”

This question is exploratory rather than transactional, requires synthesized information from multiple sources, and triggers dozens of related questions that the searcher hasn’t yet articulated.

A little digging around in Ahrefs showed that this question is a difficult keyword for which to rank. The more succinct “best way to save for retirement” is more popular than the question form, which isn’t surprising. People tend to ask full questions to LLMs rather than Google these days.

We can also see where this phrase appears most often, including sites for the Department of Labor, Trinity College, Fidelity, Merrill Edge, Bankrate, and Reddit (remember these sites because they’ll appear often later in our research).

But for the financial company, ranking well for this single query in traditional search isn’t enough. They need to appear across all subqueries about 401(k) contribution limits, retirement calculators, average retirement savings by age, and much more.

Now that we’ve researched the initial query, let’s get to the query fan-out.

Generating a Query Fan-Out with Qforia

Qforia is a tool developed by Mike King at iPullRank to simulate a query fan-out. Begin by visiting the Qforia site, then enter a paid Gemini API key (the free ones won’t work) and enter your query. After a few moments of thinking, your results will appear:

Qforia leverages the same Gemini model used by AI Overviews and AI Mode, so the output is still probabilistic, but the goal is accuracy, not precision. Plus, Mike’s tool bakes in a lot of useful information to the results:

The “query” column shows you other related queries.
The “type” column shows you what kind of question the query is likely asking. Is it comparative like a 401K vs. Roth IRA or personalized like “retirement savings strategies for someone in their 30s”?
The “user intent” column tells you what the searcher is looking for with a particular query. Do they need to understand the basic options? Are they seeking advice tailored to a specific life stage? These classifications help you group your queries based on who is asking and why.

Implicit and personalized queries seem to be common in this fan-out, showing that asking about the best way to save for retirement can be very specific to a certain person.

Another useful bit of data is the “format reason” column. This is a relatively new feature of Qforia that lays out what type of content would be best to answer the query, be it a guide, article, table, interactive tool, etc.

In this example, Qforia suggests a table for a side-by-side comparison of 401Ks and Roth IRAs. And for “retirement savings strategies for someone in their 30s”, it recommends a checklist with actionable steps to take. The “steps to open an IRA account” phrase is another implicit query that would benefit from a numbered or bulleted step-by-step list to follow.

The query “what is an employer 401K match” is considered to be a frequently asked question, so Qforia recommends answering it within an FAQ page about retirement benefits. An interactive tool seems to be the best way to help someone answer “how much money should I save for retirement?”

Now that we’ve performed our fan-out research, let’s take a look at the search engine results pages (SERPs) and see what we can learn.

Analyzing the SERPs for Fan-Out Queries

For this experiment, I will choose five of the fan-out queries Qforia suggested (in addition to our original query) and see what the SERPs look like for each:

Beginner’s guide to saving for retirement
What are the different types of retirement accounts?
How much money should I save for retirement?
Retirement savings strategies for someone in their 30s
Common retirement savings mistakes to avoid

Let’s start with the SERP for the initial query: What’s the best way to save for retirement?

Once you get past the litany of sponsored posts and the AI Overview, the rest of the results give a wide range of helpful resources. There are top 10 lists of tips from Merrill Edge and the U.S. Department of Labor, Reddit and Quora forum answers, YouTube videos, and articles from universities and credit unions.

Bankrate even offers a retirement calculator to estimate how much you should save, even though I didn’t even ask that yet (we’ll see if it shows up again later for that query).

Business	Media Type	Content Type	Website Type
U.S. Dept. of Labor	PDF article	Listicle	Government site
Reddit	Forum posts	Various	Forum
Trinity College	Article	Beginner’s guide	College site
Texas Hospital Association	Article	Listicle	Brand site
Bankrate	Calculator	Informational	Brand site
Space Coast Credit Union	Article	Informational	Brand site
California Credit Union	Article	Listicle	Brand site

When it comes to the AI Overview and AI Mode results, they both offer a bulleted list of strategies for saving. The AI Overview is more detailed, though, with a table showing the different types of retirement accounts and an action plan to follow:

Now let’s see the results for “Beginner’s guide to saving for retirement”:

We see many similar results to the original query, such as the “6 Essential Steps” video, the “Retirement 101” article from Trinity College, the “top 10 ways to prepare for retirement” article from the U.S. Department of Labor, and the Reddit thread on how to start saving for retirement. The AI Overview was also another list of tips and steps to take to save.

Business	Media Type	Content Type	Website Type
NerdWallet	Calculator	Interactive	Brand site
SuperGuy	Video	List	Brand account
Trinity College	Article	Informational	College site
Amazon (Author Dan Clay)	Book	Informational	Brand site
U.S. Dept. of Labor	PDF article	Informational	Government site
Bankrate	Article	Informational	Brand site
Reddit	Forum posts	Multiple	Forum

The process can be applied to every fan-out sub-query, but here we want to focus on the organic rankings.

Next up is “What are the different types of retirement accounts?”:

This query is very specific, and so the results were highly specific lists of definitions. The AI Overview and AI Mode both had the same bullet lists, and the rest of the SERP featured articles from sources like Charles Schwab, Northwestern Mutual, and American Express that were all laid out in list form.

There’s also a “people also ask” section at the top of this query with even more related and highly specific questions.

Business	Media Type	Content Type	Website Type
IRS	Table of contents	Links to informative webpages on the site	Government site
Fidelity	Article	Listicle	Brand site
U.S. Dept. of Labor	Article	Informational	Government site
Thrivent	Article	Informational	Brand site
Reddit	Forum posts	Multiple	Forum
Equifax	Article	Informational	Brand site
Charles Schwab	Article	Listicle	Brand site
Northwestern Mutual	Article	Listicle	Brand site
American Express	Article	Informational	Brand site

As for “How much money should I save for retirement?”:

The results consisted of:

Articles that all gave the same recommendation: save between 10 and 15% of your annual income, implying that it really is a best practice that everyone follows (or simply everyone is copying each other’s advice).
A couple of retirement calculator tools from NerdWallet and Merrill Edge but there weren’t nearly as many as I expected to see (maybe they are difficult to build or not very accurate), so perhaps our financial organization can capitalize on that with a better one.

Based on the lack of diversity in results, it would appear that Google feels this question has already been answered and nothing else is needed.

Business	Media Type	Content Type	Website Type
NerdWallet	Calculator	Interactive	Brand site
Citizens Bank	Article	Informational	Brand site
Texas Hospital Association	Article	Listicle	Brand site
The People’s Bank	Article	Informational	Brand site
Merrill Edge	Calculator	Interactive	Brand site
Brookings	Article	Informational	Brand site

The AI Overview had some recommended guidelines, a table with age-based benchmarks for saving, and a list of methods to estimate how much, including a link to the NerdWallet calculator. The same result was found in AI Mode.

The query “Retirement savings strategies for someone in their 30s” gave a bunch of highly specific results mainly consisting of articles and videos making recommendations. The results even had a “find related products and services” section, implying that this query is so bottom of the funnel that people searching this very specific query probably want to select an actual account/financial institution.

However, one thing to note is that some of the results misunderstood the question. The article from Farther Financial talks about early retirement, as in how to retire when you’re in your 30s instead of starting to plan in your 30s.

Overall, nothing linked on this SERP was a repeat of the other queries:

Business	Media Type	Content Type	Website Type
Securian Financial	Article	Informational	Brand site
Reddit	Forum posts	Multiple	Forum
Fidelity	Article	Listicle	Brand site
Bankrate	Article	Listicle	Brand site
Thrivent	Article	Informational	Brand site
Hancock Whitney	Article	Informational	Brand site
MassMutual	Article	Informational	Brand site
Farther Financial	Article	Informational	Brand site
America’s Wealth Management Show – YouTube	Video	Informational podcast	Brand account

Aside from a few videos up top, the results for “Common retirement savings mistakes to avoid” were hilariously all list articles ranging from 10 important tips to 5 mistakes to avoid. AI Mode and the AI Overview simply listed out the mistakes grabbed from many of the articles listed on the SERP:

Business	Media Type	Content Type	Website Type
T. Rowe Price – YouTube	Video	Informational	Brand account
CBS Philadelphia – YouTube	Video	News story	News station account
MeaningfulMoney – YouTube	Video	Informational	Brand account
Louisiana Office of Financial Institutions	PDF Article	Informational	Government site
Morgan Stanley	Article	Listicle	Brand site
Charles Schwab	Article	Listicle	Brand site

Next, we’re going to look at all these results together and start analyzing what it all means.

Matching Content Formats to Intent

It’s clear that format choice matters just as much as topic choice. As we’ve seen from the screenshots, the SERPs seem to favor list articles, videos, and interactive calculator tools.

One thing I noticed (and was pretty excited about) was how many reputable sources were cited. Actual financial institutions with experience in the subject matter like Fidelity, Charles Schwab, Wells Fargo, and others topped the SERPs and were cited in AI Overviews and AI Mode. I read the AI Overviews and actually believed they were telling the truth, which isn’t something that happens often.

This is great for the people doing the searching, but a challenge for our financial organization to overcome in rankings. They will have to work hard to produce the kind of content that can compete with these huge firms. But thanks to our query fan-out and SERP research, we know what type of content ranks well to give them an idea of where to focus their marketing efforts.

However, it’s also important to consider the fact that a new type of content that hasn’t been seen on the SERPs like an FAQ page or guide might capture more eyeballs than the same listicle over and over. It’s all about experimentation and seeing what works.

A successful content strategy will need to be creative, helpful, and omnimedia to focus on every search channel in order to compete on these SERPs.

Expanding Beyond Web Pages into Omnimedia

What is an omnimedia and omnichannel strategy? It focuses on everything in every channel and in every media format: videos on YouTube, text in Reddit forums, social media, review site tables, TikTok videos, literally anywhere a person can search for something.

In our finance company example, there weren’t any instances of social media or TikTok posts showing up in the SERPs, but we did see quite a few YouTube videos and Reddit and Quora forum posts.

I looked at the same queries in a few LLMs (ChatGPT, Gemini, and Perplexity) to analyze their results. Rather than inundate you with mountains of screenshots, I will just sum up the findings for a few of the queries. I started with the original query of “What’s the best way to save for retirement”:

ChatGPT’s answer was a simple list explaining the types of retirement accounts, a recommended target percentage of income to save, and some tips for smart saving. I didn’t see any links or citations in this summary.
Gemini focused on tables that could be exported to Google Sheets that showed age benchmarks and how to choose the right account type. At the bottom, there was a sources button that opened a window on the right with a list of sources including AARP, T. Rowe Price, Fidelity, and Empower.
Perplexity offered links to cited sources throughout its summary after each section, which included Guardian, Schwab, Vanguard, and even Reddit.

For the “Beginner’s guide to saving for retirement” prompt:

ChatGPT had an overly simplistic list of steps with no cited sources.
Perplexity had a 5-step list with cited sources throughout (mainly the same ones cited the last time like Schwab and Fidelity).
Gemini had a 3-step list, a table of benchmarks, and a checklist, but the most interesting part for me was it specifically recommended several companies with which to open a Roth IRA. This is important for brands to note. With enough content on this topic, financial companies could be added to this list:

For the next query of “What are the different types of retirement accounts”:

ChatGPT gave a list with brief explanations and no citations again.
Perplexity detailed what each plan was and how it worked, citing a source at each line.
Gemini did the same citing many different sources:

For the sake of brevity, let’s go straight to “Common retirement savings mistakes to avoid” since this query is a different style – a serious prompt with negative implications on a person’s life, aka. Your Money or Your Life (YMYL) prompts:

ChatGPT offered a list of mistakes with a “why this hurts” and “do this instead” line for each.
Gemini gave a much more detailed list of mistakes with sources such as Discover, LPL Financial, T. Rowe Price, and Kiplinger. These sources differed from those that Gemini cited for the other prompts.
Perplexity cited 10 sources in their brief list of mistakes to avoid, this time with a few reputable news sources like the New York Times and The Harvard Gazette.

Since this prompt opened up the citations to new sources we hadn’t seen with the other prompts, it shows that organizations like Fidelity and Schwab don’t have a monopoly on citations for this topic. That means our financial business could possibly join the citations.

It also means that this type of prompt that’s focused on YMYL issues, such as making mistakes with your retirement savings, can impact the sources cited. Therefore, depending on their status in the industry, this may not be the right prompt for our financial company to focus on.

Let’s look at the results as a whole to see how our finance firm can fit in.

Reverse-Engineering Citations to Inform Content Creation

AI citations, derived from analyzing SERP features and LLM responses, help content planning by revealing the types of sources, formats, and channels necessary to gain visibility, especially in competitive areas like financial services.

AI Overviews and LLMs draw from a diverse ecosystem of content, which can be categorized by the credibility and nature of the source and the channels they utilize:

Source Type	Examples of Channels and Sources	Insight for Content Planning
Institutional/Financial Firms	Web pages, articles, and tools from established firms like Fidelity, Charles Schwab, T. Rowe Price, Vanguard, and LPL Financial.	These sources dominate citations for core financial topics, signaling the need for high-authority, expert-level content to compete.
Official/Educational	Articles and guides from government agencies and educational institutions, such as the U.S. Department of Labor, Trinity College, and AARP.	Citations favor unbiased, fact-checked authority, especially for complex or sensitive topics like retirement planning.
Reputable Media/News	Articles from established publishers and news sources like NerdWallet, Bankrate, Kiplinger, The New York Times, and The Harvard Gazette.	Targeting queries focused on specific angles, like “Common retirement savings mistakes to avoid” (a negative prompt), can broaden citation opportunities beyond traditional financial firms into reputable media sources.
Community/ User-Generated	Posts and threads from forums like Reddit and Quora.	An omnimedia strategy should recognize and target platforms where prospective customers are having exploratory, non-transactional conversations, as these are cited in both SERPs and LLMs.

The channels cited are primarily traditional web pages/articles, but an effective content strategy must be omnimedia, focusing on all channels where people search.

The type of content that receives citations is often determined by the specific query intent and is heavily favored by formats that are easily synthesized by AI. Cited assets tend to be highly structured and actionable:

List Articles (Listicles): Many SERP results and AI Overviews for queries like “mistakes to avoid” or “types of retirement accounts” heavily feature content formatted as lists.
Tables: AI Overviews frequently use tables to detail complex information, such as age-based benchmarks for saving or comparisons of different retirement accounts.
Interactive Tools and Calculators: For questions about specific quantities, such as “How much money should I save for retirement”, interactive tools and calculators from sources like Bankrate and NerdWallet are featured prominently in results and AI Overviews. Qforia also specifically recommends interactive tools for certain queries.
Actionable Formats: Both Qforia analysis and LLM outputs favor step-by-step lists and checklists for actionable queries like retirement strategies for someone in their 30s.
Videos: Videos are visible on SERPs across multiple exploratory queries, indicating the necessity of including video production in an omnimedia strategy.

Analyzing these citation patterns transforms competitor analysis into actionable content strategy, allowing a financial firm to focus its marketing efforts:

Prioritize Format Over Topic Saturation: Since many SERPs repeat listicles, opportunity may lie in creating different asset types, such as FAQ pages, comprehensive guides, or tables, that might capture more attention than yet another listicle.
Target Implicit Transactional Intent: Although the customer starts with exploratory questions like “What’s the best way to save for retirement”, AI models like Gemini are willing to suggest specific companies for opening accounts. This creates an opportunity for a financial business to position itself to be specifically recommended by AI.
Use Specific Query Styles to Bypass Citation Monopolies: While major firms like Fidelity and Schwab often dominate general retirement citations, queries focusing on specialized or negative topics, like “mistakes to avoid” (which touches on YMYL issues), open up citations to different types of high-authority sources, including news publications. This suggests either focusing resources on specific, highly authoritative content addressing nuanced concerns, or just letting this query go.
Adopt an Omnimedia/Multi-Channel Presence: The appearance of Reddit, Quora, and YouTube content in both SERPs and LLM citations shows that high-quality content planning must extend beyond official website articles. To fully compete, content must be tailored for, and distributed across, these community and video channels.

Now that we have an idea of how these results can inform content strategy, we’ll put together a generalized omnimedia strategy for our hypothetical finance business.

Building an Omnimedia Strategy

Here’s what we do at iPullRank:

Keyword Matrix: Establishes the groundwork for topical presence in AI Search by identifying synthetic keyword gaps, potential opportunities, and valuable terms at the passage level. It reveals which areas are strengthening or weakening your AI presence, enabling you to allocate resources toward terms that create competitive differentiation.
Omnimedia Content Audit: Comprehensive assessment of content spanning all formats and channels, covering owned, earned, and shared media properties. It uncovers what’s limiting your authority while pinpointing where focused resources will generate results.
Omnimedia Content Plan: Strategic blueprint for enhancing Content Resonance and expanding coverage across discovery channels, complete with content frameworks and governance structures. It provides a defined pathway to greater prominence in AI-generated responses.
AI Search Measurement Plan: Monitor input indicators, channel data, and performance benchmarks across AI Search environments. It identifies how AI Search influences growth trajectories, allowing you to connect spending with outcomes and refine your approach based on evidence.

Given that our fan-out research gave us the keywords we need and we already analyzed their presence in searches, let’s put together a plan for our financial company. This plan leverages the query fan-out analysis for retirement planning to create resonant, structured content across multiple channels.

Query Fan-out Prompt	Recommended Asset Type & Structure	Omnimedia Objective
“What’s the best way to save for retirement” (Core)	Pillar Guide / Comprehensive List: Create a definitive, long-form guide. Structure the article so it begins with the core answer or takeaway, followed by supporting detail and nuance.	Semantic Resonance: Ensure the title, description, and internal linking structure establish this as the canonical source for all related fan-out queries.
“Beginner’s guide to saving for retirement”	Actionable Checklist/Step-by-Step Guide: Develop a simplified, numbered list of steps for beginners. Content should be clear and simple.	Passage-Level Extraction: Optimize for easy extraction of the 3 or 5 essential steps by LLMs (like ChatGPT’s simplistic lists).
“What are the different types of retirement accounts?”	Comparison Table and FAQ Page: This requires highly specific definitions. Create a dedicated FAQ page featuring a comparison chart or table detailing account types and their features.	Structured Data: Ensure the data is provided in a clean, easily digestible format (table) that AI Overviews and LLMs already favor when detailing complex information.
“How much money should I save for retirement?”	Interactive Tool/Calculator: The primary format requested for this query is an interactive tool. This must be accompanied by structured data (tables) showing age-based benchmarks for saving.	High-Value Utility: Provide functional utility (the tool) and support the quantifiable answer with authoritative data (the table), increasing the chance of citation over static articles.
“Retirement savings strategies for someone in their 30s”	Personalized Checklist/Segmented Guide: Content should be highly specific and actionable for a segmented audience. Use the product grouping concept to show a progression_from or complementary_with path for financial products relevant to this age group.	Targeted Personalization: Customize the content to fit the behavior and specific needs of this persona, making the recommendations feel personalized, which is crucial for AI agents.
“Common retirement savings mistakes to avoid”	YMYL List Article with Solutions: Create a list detailing the mistakes, each paired with a solution. Must be accurate, with citations and data sourced from trusted research.	Trust and Authority: This content relates to YMYL issues. It must be highly detailed and authoritative to compete with reputable news sources and large financial institutions (otherwise don’t bother).

Because LLMs often prioritize user-generated content, the content plan must actively generate and integrate these signals across all channels:

Channel	Content Focus based on Retirement Queries	Strategic Goal
YouTube	How-To Videos & Guides: Create short, instructional videos for the “Beginner’s guide” and “Strategies for someone in their 30s” queries. Ensure video descriptions include relationship attributes to related website guides and link to the interactive tool.	Capture Exploratory Search: Videos are crucial for user education in the finance industry. Position the firm as an authority where consumers compare accounts and learn how they work.
Community Forums (Reddit/Quora)	Monitoring and Engagement: Focus resources on monitoring discussions related to “Common retirement savings mistakes to avoid” and “Beginner’s guide”, as these forums appear frequently in SERP and LLM citations.	Influence Citation Flow: While the firm cannot directly control these posts, understanding the conversation context allows for developing content that directly answers community-driven questions with clarity and context.
LinkedIn	Thought Leadership: Utilize LinkedIn Pulse for “Retirement savings strategies for someone in their 30s”. Publish educational, semi-formal posts that bridge technology and business to capture a broader professional audience.	Professional Authority: Establish individual and corporate credibility on a channel that ranks well for thought leadership.

By implementing this omnimedia plan, the financial company treats every content touchpoint from the interactive tool on its site to a checklist shared on LinkedIn as a signal that reinforces a clear, emotionally grounded, and contextually rich narrative about retirement planning to maximize visibility in SERPs and LLMs.

But what if your financial business is enterprise-sized with a massive library of webpages? How can you scale your content strategy? And how can you easily prune old, irrelevant content that is no longer serving your site?

Scaling Query Fan-Out and Omnimedia Planning

For a single topic or article, traditional content pruning involved manual work that was often time-consuming and subjective. This process required manually pulling traffic data, checking rankings, looking at backlinks, and skimming posts to make a “best guess” about relevance.

The content pruning methodology we’ve developed at iPullRank uses a Relevance Engineering framework that blends AI-driven semantic relevance analysis with SEO and content metadata to make data-driven decisions at scale. This will help ensure all the content on your site is relevant and identify what content should be removed.

The steps for executing this workflow involve defining your strategic benchmarks, quantifying meaning using vector embeddings, integrating performance data, and applying a final decision framework.

Here are the steps we follow to perform the content pruning:

Step 1: Define Strategic Focus

The entire analysis is grounded in business reality by defining what relevance means for the organization.

Work with the client to solidify their core solution areas or strategic pillars.
For each area, develop representative keyword portfolios reflecting user intent and product capabilities.
Draft a concise business relevance statement (a short paragraph capturing the ideal focus).

Step 2: Generate Topic & Business Relevance Centroids

This step translates the strategic pillars into quantitative benchmarks.

Using the chosen embedding model, generate an embedding for each individual keyword within a topic’s portfolio (developed in Step 1).
Calculate the average of all keyword embeddings within that cluster. This averaged vector becomes the topic centroid, serving as a robust mathematical representation of the topic’s semantic space.
Generate a single embedding for the overall business relevance statement.

Step 3: Generate Article Embeddings

The meaning of every post in the library is represented numerically.

Run the extracted article content through basic Python cleaning routines (removing excess whitespace or stray HTML tags).
Combine the Title text and the cleaned main body content for each article.
Generate a single embedding for each article using the same model used in Step 2 for consistency.
Store these individual article embeddings, typically in a NumPy array paired with their corresponding URLs.

Step 4: Calculate Similarity Scores

This step objectively measures how semantically aligned each article is with the strategic targets.

Using Scikit-learn in Python, calculate the cosine similarity between each article’s embedding (from Step 3) and each of the topic centroid embeddings (plus the business relevance embedding from Step 2).
The output is a matrix added to the spreadsheet, showing scores between -1 and 1, where scores closer to 1 indicate a stronger semantic alignment.

Step 5: Layer SEO Performance & Metadata

Real-world performance and freshness data are integrated with the relevance scores.

Pull standard SEO metrics, primarily trailing 3 to 6 months of organic clicks from Google Search Console, focusing on a recent window to avoid rewarding historical performance.
Pull Publish Date and Last Modified Date from the CMS or crawl data.
Join this performance and freshness data to the master spreadsheet containing the URLs and similarity scores.

Step 6: Apply the Decision Framework (Making Informed Choices)

The combined data points are translated into actionable categories to guide pruning and optimization efforts.

Establish data thresholds to categorize each article into three main actions:
- KILL: Candidates typically have low similarity scores across all core topics, low recent GSC clicks, and are old (more than 5 years with no significant updates).
- KEEP: Articles that qualify if they have high similarity to at least one core topic or have strong recent SEO performance.
- REVIEW/REVISE: Articles that are relevant but underperforming, outdated, or highly similar to other posts (potential consolidation targets).
Conduct manual review of edge cases, confirm decisions, and identify specific actions.

Reframing SEO with Query Fan-Out

For firms competing against giants on the SERPs and in LLMs, focusing on an omnimedia query fan-out approach to content strategy is essential. The citation analysis we walked through shows that specialized queries, especially those focused on specific concerns like “mistakes to avoid” or personalized scenarios like “strategies for someone in their 30s”, open up opportunities for smaller firms to break through with unique and truly valuable content.

You can’t out-budget the big players, but you can out-execute them on format diversity, omnimedia presence, and targeting the fan-out queries they’re ignoring. And when you scale this with the Relevance Engineering framework for content pruning, you’re optimizing your entire content library to align with how AI Search works.

We know that this can be overwhelming for a limited staff to achieve, but we can handle it for you. If you want to do this for your own business, talk to iPullRank about our AI Search Strategy Program with a full Omnimedia Content Audit and Content Plan, tied directly to a Keyword Portfolio Matrix, designed to improve visibility across traditional search, AI search, and third-party platforms. Let us help you be the first thing customers see (and trust) wherever they search.

Want to learn more about how we can help you with AI visibility?

Read about our AI Search Strategy Program

The post Query Fan-Out in Practice: Turning One Search into an Omnimedia Content Plan appeared first on iPullRank.