What Google Actually Says About Ranking in AI Overviews
You can’t buy your way into AI Overviews. Google has now said so in writing, three different ways.
Between two official documents, one defensive blog post, and a rewritten spam policy, Google spent 2025 and 2026 answering the question everyone kept asking. Here’s every statement that matters, what the data says back, and what still actually works.
The Short Version, If You Read Nothing Else
- Google’s official position, published May 15, 2026: there are no special requirements, no AI markup, and no secret schema for AI Overviews or AI Mode. The features run on the same index and ranking systems as regular search.
- Google named the tactics that don’t work, in its own documentation: llms.txt files, AI-specific formatting, Markdown page versions, and manufactured brand mentions.
- Selection happens through query fan-out. The model splits your question into dozens of background searches, then builds one answer from the results. Passage relevance beats page-level rankings.
- Google claims total organic clicks stayed “relatively stable” and click quality went up. Pew Research measured users clicking links 8% of the time when an AI summary appears, versus 15% without one. Both things are on the record.
- Since May 15, 2026, manipulating AI answers is spam by policy. The legitimate path runs through the same work as always: earned links, real mentions, and content worth citing.
Four Documents Hold Everything Google Has Said
For two years after AI Overviews launched in May 2024, Google’s guidance lived in scattered conference answers and podcast clips.
Then it consolidated. Everything official now traces back to four sources, and every claim in this article links to one of them.
AI Features and Your Website
Google’s first documentation page on how AI Overviews and AI Mode work, how inclusion is controlled, and how traffic gets reported.
The Generative AI Guide
The official AI search guide, announced by John Mueller. Includes the mythbusting section that names tactics you can skip.
The Traffic Defense
Search head Liz Reid’s blog post claiming stable clicks and higher click quality. The most contested document of the four.
The Spam Boundary
The same day the guide dropped, Google rewrote its spam policy to cover AI answers. The carrot and the stick shipped together.
Two Years From Launch Stumbles to Written Doctrine
AI Overviews rolled out to all US users in May 2024, and Liz Reid pitched the upside at launch: links inside AI Overviews, she said, pick up more clicks than the same pages would as standard listings.
Then week one happened. Viral wrong answers, including the infamous suggestion to put glue on pizza, forced Google to dial back how often summaries triggered and tighten the query types that got them. The feature spent the rest of 2024 quietly rebuilding trust.
2025 was the expansion year.
More countries, more languages, more query categories, and AI Mode arriving as its own conversational tab before spreading fast through the year. Publishers ran the numbers the whole way, and the gap between Google’s reassurances and publisher dashboards became the industry’s loudest running argument. Reid’s August 2025 traffic post exists because that argument got loud enough to require an official answer. It answered the question about direction while sidestepping the one about magnitude, and the industry noticed.
2026 is when Google switched from ad-hoc answers to written doctrine.
February brought a redesigned citation display with hover previews on desktop, making sources more visible inside AI answers. May 15 delivered the official guide and the spam policy line in a single day. And by late June, the first spam update under the new wording had already run. Two years of improvisation, replaced by a rulebook.
The Eight Things Google Actually Says
Strip away the commentary industry and Google’s position fits in eight statements.
Each one below is sourced to an official document or a named Google employee. No inference, no reading between lines.
There Are No Special Requirements
Google’s documentation states it flat: “There are no additional requirements to appear in AI Overviews or AI Mode.” No application, no toggle, no separate eligibility. If a page can rank and can be shown as a snippet, it can be cited.
It All Runs on the Same Ranking Systems
The May 2026 guide opens by asking whether SEO still matters for AI search and answers itself: “In short, yes!” AI Overviews and AI Mode pull from the same index, the same crawl, and the same quality systems as the blue links. A page that can’t rank can’t ground an answer either.
You Don’t Need llms.txt or AI Markup
The guide’s mythbusting section kills the checklist an entire vendor category was selling. No llms.txt file, no AI-specific text formatting, no Markdown versions of your pages. Google says it may crawl those files like any other file, but assigns them no special meaning for AI features.
No Schema Gets You Into AI Answers
Structured data is still worth using for rich results, and that’s where Google stops. There’s no schema.org type that gets you into AI Overviews, per the same guide. Two years of “add FAQ markup for AI visibility” advice, contradicted in one paragraph.
Selection Happens Through Query Fan-Out
Google’s VP of Product for Search, Robby Stein, described the mechanics in plain terms: the model appends dozens of related background queries to yours, “and it’ll start Googling basically.” The official guide later defined the term in writing. More on why this changes your content strategy below.
Buying Mentions Won’t Get You Cited
Google acknowledges AI features surface what blogs, videos, and forums say about brands. Then it warns that seeking inauthentic mentions “isn’t as helpful as it might seem,” because ranking systems weigh quality and spam systems filter the rest. Real coverage counts. Rented coverage gets filtered.
Manipulating AI Answers Is Spam, By Policy
Since May 15, 2026, Google’s spam policy names “attempting to manipulate generative AI responses” as a violation, right alongside ranking manipulation. Paid citation schemes and fake mention networks moved from gray area to enforcement target. The June 2026 spam update was the first pass under the wider wording, and enforcement research shows why Google needed the rule: a single planted comment can steer what an AI system recommends.
AEO and GEO Are, In Google’s Words, Still SEO
The guide defines both acronyms, then dissolves them. From Google’s perspective, working on generative AI visibility is working on the search experience, which makes it plain SEO. Semrush’s read on the guide landed the same way: content that can’t rank in traditional search won’t perform in AI answers either.
Put the eight together and the shape is hard to miss.
Statements one through five describe the door: same index, same systems, open to any page good enough to rank. Six through eight describe the lock: no bought mentions, no manipulation, no separate discipline to shortcut through. Google published the door and the lock on the same day, which tells you the company saw exactly where the GEO vendor market was heading and decided to get ahead of it in writing.
How a Page Actually Gets Picked
Google’s own guide uses a lawn-care example. Someone asks how to fix a lawn full of weeds, and the model quietly runs related searches on herbicides, chemical-free removal, and weed prevention.
Your page competes in every one of those background searches, and the person only ever typed one of them. Here’s the pipeline in three moves.
Fan Out
The model splits the question into concurrent sub-queries, sometimes dozens, covering angles the person never typed. Stein compared it to the system running its own Google searches in the background.
Retrieve
Each sub-query pulls candidates from the regular Search index using the regular ranking and quality systems. This is the step where all your existing SEO either pays off or doesn’t.
Synthesize and Cite
A language model merges the retrieved passages into one answer and picks which sources to link. Passages win citations, which means a page can get cited without ranking first for the head term.
The strategy shift hides in move one. Search Engine Land’s fan-out guide and SEJ’s breakdown of Stein’s comments both land on the same conclusion.
Ranking for one phrasing no longer guarantees visibility. The pages that keep showing up in AI answers cover the sub-questions around a topic instead of the head term alone.
There’s even a paper trail on the architecture.
SEJ noted a December Google patent describing thematic search: sub-queries generated from inferred themes, results grouped by topic, summaries compiled across documents. Google hasn’t confirmed implementation, but the description matches Stein’s account almost line for line. Either way, the practical takeaway holds: topical depth beats keyword repetition, on every AI surface at once.
Where Google’s Story and the Data Disagree
Everything above is Google describing its own systems, and there’s little reason to doubt it.
The traffic claims are different. Here, Google’s statements and independent measurement point in opposite directions, and honesty requires showing both.
Total organic click volume to websites has stayed “relatively stable” year over year, per Liz Reid’s August 2025 post.
Pew Research tracked real browsing from 900 US adults. With an AI summary on the page, users clicked a result 8% of the time, versus 15% without one. Nearly half the clicks, gone.
Click quality is up. People who do click stick around instead of bouncing back, and AI Overviews show more links on the page than before.
The post shared no numbers, which tech press flagged immediately. Meanwhile publishers keep reporting the floor dropping, like Mail Online losing more than half its Google traffic on affected queries.
AI features create opportunities for more types of sites to appear, helping people discover content they wouldn’t have found before.
Partly true. Citation studies keep finding pages cited from outside the organic top ten. But Pew also found AI summaries appeared on 18% of tracked searches and rising, so the discovery upside rides on a shrinking pool of clicks.
Share of tracked visits where users clicked a link inside the AI summary itself, per Pew. Being cited is brand visibility and entity credit. It is not, on its own, a traffic strategy. Plan for both.
Four Things Google Still Won’t Tell You
Reading what Google says means noticing what it skips.
The documentation is genuinely useful, and it goes quiet in the same four places every time. Plan around these blind spots instead of waiting for them to close.
No AI Traffic Breakdown
Search Console blends AI Overviews clicks into the Web search type with everything else. You can’t isolate what AI features send you, which conveniently makes Google’s traffic claims hard to check from the outside.
No Citation Selection Criteria
Google explains retrieval, then stops. How the model picks three citations from twenty candidate passages stays undocumented, which is why citation tracking became its own tool category.
No Trigger-Rate Data
Which queries get an AI Overview, how often, and in which verticals comes entirely from third-party trackers. Google has never published its own numbers, even while disputing everyone else’s.
No Word on Other AI Platforms
The guidance covers Google’s features only. ChatGPT, Perplexity, and Gemini’s standalone app weight sources their own way, so Google’s mythbusting settles the debate for exactly one platform.
None of this changes the playbook.
It changes the measurement. Teams that treat AI visibility as its own reporting line, tracked query by query, are the ones catching shifts before their clients ask about them.
What Google Tells You to Do Instead
The guide isn’t all mythbusting. It carries a positive list, and it’s short.
Six things, all of them boring, all of them straight from Google’s own pages.
Stay Crawlable and Indexable
Robots.txt, CDN rules, and rendering can all silently block AI features. If Google can’t fetch and index the page, nothing downstream matters.
Write Non-Commodity Content
Google’s framing splits content into commodity and non-commodity. Generic summaries of what already ranks add no citation value. First-hand experience, original data, and a real point of view do.
Cover the Sub-Questions
Fan-out means one query becomes dozens. Pages that answer the surrounding questions, comparisons, costs, alternatives, and how-tos, win more of those background searches.
Use Snippet Controls, Not Hacks
The only real inclusion levers are the ones that always existed: nosnippet, max-snippet, and noindex. They limit AI features the same way they limit regular snippets.
Measure in Search Console
AI Overviews and AI Mode traffic reports inside the Performance report under the Web search type. No separate AI breakdown exists yet, which Google acknowledges and site owners keep requesting.
Ignore the Tool Promises
The guide states no third-party tool has access to Google’s internal ranking or AI systems. Use tools for workflow. Check their advice against the documentation.
The guide runs wider than this list, for what it’s worth.
It covers local visibility, e-commerce feeds, and the first official mentions of agentic experiences, where AI acts on a user’s behalf. But every branch loops back to the same root: demonstrable expertise, clean technical foundations, and content only you could have written. Nothing in the document rewards a shortcut.
Why This All Points Back to Earned Links and Real Mentions
Read the eight statements again and notice what Google left standing.
Every shortcut got named and killed. What survived is the work that feeds the retrieval step: pages worth ranking, on sites worth trusting, referenced by sources the index already respects.
There’s a wrinkle worth naming before anyone celebrates.
Google spent 2023 and 2024 talking links down. Gary Illyes said Google needs “very few links” to rank pages and disputed their top-three status outright. Both statements were about counting, though. Retrieval still has to choose which of ten similar pages to trust, and earned references remain how a site proves it belongs in that answer, which is exactly why marketers keep circling high-authority placements as the AI era’s tempting shortcut. Tempting, and now policed.
That’s a links-and-mentions story whether Google frames it that way or not.
Manual outreach to relevant publications still earns the placements that carry authority into retrieval. Guest posts with a genuine editorial bar put your expertise on the exact pages fan-out queries keep pulling. And AI-era backlinks on trusted industry sources double as the citations AI systems lean on when naming brands.
The mention warning cuts the other way too. Google flagged inauthentic mentions, not mentions.
Genuine brand mentions in editorially real content, and honest inclusions in third-party listicles people actually read, are precisely how AI systems learn a brand belongs in an answer. The difference between those and a fake mention network is the difference Google’s spam systems now exist to detect.
One caution from three years of covering spam enforcement: anchor discipline and site vetting matter more now, not less.
A citation from a burned domain is worth nothing on any surface. The fundamentals of link quality got promoted, not replaced.
We Were Already Doing What Google Just Wrote Down
Here’s the part that made May 15 an easy day for us. Google’s sanctioned playbook, earned placements on relevant sites, real editorial standards, no manufactured signals, is the model our 150+ agency partners already run through our white label link building and AI SEO services.
Every domain gets pre-approved before outreach starts, every placement clears a human relevance check, and every deliverable ships under your agency’s brand within 25 days.
No AI hacks to unwind, because we never sold any. The agencies that spent 2025 buying AI placement packages are auditing them now. Ours are reporting citation wins.
Vetted publishers in our network, reached by 150+ human outreach specialists. The same link building work that ranks pages in blue links now earns the citations AI answers draw from. One program, every surface.
Quick Answers, Straight From the Documentation
Do I need an llms.txt file to show up in AI Overviews?
No. Google says it doesn’t process the file in any special way for AI features. It can crawl it like any text file, and that’s the extent of it.
Does schema markup help me rank in AI Overviews?
Not specifically. Google recommends structured data for rich results and stops there. No schema type is required for, or gives an edge in, AI features.
Can I pay to get my brand cited in AI answers?
Not safely. Since May 15, 2026, manipulating generative AI responses is a named spam violation. Paid citation schemes and fake mention networks carry the same risk as bought links, including loss of visibility across search, AI Overviews, and AI Mode at once. The June 2026 enforcement pass arrived barely five weeks after the policy did.
Is AI Mode replacing the regular results page?
No. AI Mode is a separate conversational tab, and Google says AI Overviews are designed for queries where a summary adds something beyond standard results. Both draw on the same index and ranking systems, which is the practical point: one program covers every surface.
How do I track traffic from AI Overviews?
In Search Console’s Performance report, under the Web search type, blended with regular results. Google hasn’t shipped a separate AI Overviews filter, so most teams pair Search Console with query-level tracking of terms that trigger AI answers.
Earn the Citations AI Answers Actually Trust
Pre-approved domains, human outreach, and white label delivery in 25 days. The playbook Google just endorsed is the one we never stopped running.
Deepan Paul
AuthorDeepan Paul is a SEO Lead with four years of experience helping brands recover, scale, and sustain organic growth across global B2B, B2C, and D2C markets. He is recognized as a ranking revival expert, specializing in diagnosing traffic drops, fixing indexing and technical issues, and restoring lost search visibility. He has managed international clients and led cross-functional teams, aligning SEO strategies with core business goals. His expertise spans technical SEO, content strategy, indexing optimization, and building scalable growth systems that adapt to constant algorithm changes. Beyond execution, Deepan is also an SEO trainer and guest speaker, mentoring professionals and contributing insights to leading digital marketing publications. His approach is focused on sustainable, system-driven SEO that delivers long-term results rather than short-term gains.
