TL;DR
- In May 2026, Google Search Central published its first dedicated guidance page for AI Overviews and AI Mode, ending a year of scattered I/O talks and blog posts.
- AI Overviews and AI Mode draw from the same Google index regular Search uses. There is no separate AI crawler to court.
- The new mechanic is query fan-out — Gemini breaks one prompt into several sub-searches, picks a source per sub-search, and stitches an answer.
- AI Mode is treated as a peer surface to AI Overviews. They run on the same AI model and cite different sources for the same query, so audit both.
- The quiet warning is the same one Google has carried for years — content hidden inside JavaScript stays invisible to AI Overviews, AI Mode, and the older indexer alike.
In May 2026, Google quietly added a new page to its Search Central developer docs. It is the first official guidance for how AI Overviews and AI Mode pick the pages they cite.
The page sits at developers.google.com/search/docs/appearance/ai-features. Google’s framing is calm. There is no separate AI playbook. The same content that earns regular search visibility is what AI Overviews and AI Mode draw from.
The guidance names a few new mechanics — query fan-out, two distinct AI surfaces, structured data for AI shopping. But the core message is that the old hygiene still does the heavy lifting.
That single page changes something for small sites.
For most of 2025 the rules for AI search visibility lived in scattered places. A blog post here. An I/O talk there. A long thread on a developer forum.
Site owners who wanted a clear answer were stuck stitching their own version together.
Now there is one map. It is short. It is calm. And it says quieter things than the loud AI-search hot takes have been saying.
Why did Google publish a 2026 AI search guidance page at all?
Pressure from site owners who could not find a clear answer in any one place. Until May 2026 the rules for being cited in AI Overviews and AI Mode lived in scattered blog posts, I/O talks, and developer threads.
In May 2026, Google’s Search Central published the AI features page as the first single source. Google offers it as a reference, not as a sales pitch.
The framing is restrained. There is no claim that AI search has replaced regular search. The page treats AI Overviews and AI Mode as new surfaces on top of the same index, not as a parallel universe small sites need a separate strategy for.
That restraint matters. It pushes back against a year of AI-search content that has been quietly hostile to small businesses. A small site cannot rewrite its whole architecture every quarter. The official map says it does not have to.
What does the guidance say about how AI Overviews picks your page?
AI Overviews uses the same index that powers regular Google Search. There is no separate AI crawler.
A page already indexed for the related query is in the running. The new piece is query fan-out. Gemini breaks one prompt into several smaller searches, picks a source per sub-search, and stitches the answer together.
This is the part most site owners miss.
A question like "how do I speed up my WordPress site" is not one search anymore. It is six or eight. Each sub-search picks its own source. The final AI Overview is a collage stitched together from different pages.
Your page might be a great answer to the head question and still not be cited. The sub-question that would have surfaced you may never have been run. The fix is to write each section as a clean answer to a smaller, more specific question — the kind a sub-search would actually fire.
The way AI engines find content is the surface change. The fan-out is the deeper one. Old SEO trained writers to think one page, one keyword. The new pattern wants one page made of several short, separable answers.
What does the guidance say about AI Mode, the separate tab?
AI Mode is a distinct product from AI Overviews. It has its own tab in Search. It has its own retrieval behavior.
The guidance treats it as a peer to AI Overviews, not a subset. That is new framing. For a long stretch of 2025, AI Mode was discussed as a beta variant or a power-user feature. The May 2026 page promotes it to a first-class surface.
Both run on Gemini. Both draw from the same index. And yet they pick different sources for the same query.
That mismatch is uncomfortable. Same company. Same machine-learning model.
Same web. Different lists of cited pages.
The practical lesson is simple. If you audit your AI visibility on only one of the two, you are missing half of Google’s AI surface. A page can be quoted in AI Overviews and ignored in AI Mode for the same prompt the same week. A clean audit checks both.
Where does the new guidance break from the old SEO playbook?
The break is the fan-out.
Old SEO optimized one page for one head term. The new pattern asks every page to be a clean answer to the smaller sub-questions the engine actually runs. That favors short answer blocks, named entities, and structured data over keyword density.
The guidance is gentle about saying this. It does not announce a paradigm shift. It just describes what happens after the prompt is sent.
But the implication is sharp. If your homepage answers one question and your service page answers one question, you have two answers on a site that needs ten. Each H2 has to start carrying its own answer to a real sub-query.
The other break is what the guidance does not say. There is no list of new tags. No new schema you must add. No new file you must publish.
AI search is not a separate technical project. It is a content-shape project. That is the biggest reframe a small site can take from the page.
Which mistake does the guidance quietly warn small sites about?
Hiding content inside JavaScript.
The page repeats what Google has said for years. Crawlers fetch raw HTML. They skip client-rendered content. AI Overviews and AI Mode inherit the same limit.
A site that loads its paragraphs through a page builder’s script is invisible to both surfaces. The page shows up perfectly in a browser. It looks fine on a phone. But the crawler that feeds the AI surfaces sees an empty shell.
This catches WordPress sites more than people expect. Some block builders render the page server-side. Others wait until JavaScript runs in the browser, which means the AI crawler never sees the content.
The guidance does not name specific products. It just names the rule.
The fix is to check what the raw HTML actually contains. Open the page in a browser, view source, and search for one of your own sentences. If you do not find it, the AI surfaces will not find it either. The way you check AI citation starts here, with the underlying view-source step.
Trust matters here too. Most AI-search advice in 2026 is louder than it is useful. The way to tell whether the AEO advice is worth trusting is to compare it against this kind of plain Google page.
The guidance is dry. That is the point.
Other questions worth answering
How long until refreshed content appears in answer-engine citations?
Two to four weeks is the rough practitioner heuristic in 2026. No verified Google source pins this down precisely. Most accounts suggest the gap between a refresh and any citation movement falls in that range.
Treat the number as an estimate, not a measurement. Re-check the same prompts week by week and let your own log build the curve.
Does Bing’s index help ChatGPT find your content the way Gemini does?
Largely yes. ChatGPT pulls from Bing’s index, OpenAI’s own crawler, and partner feeds. Gemini’s path is different because AI Overviews and AI Mode both sit on the Google index. So content earning Bing visibility does not automatically reach Gemini, and vice versa.
A clean answer-engine audit usually checks both indexes against the same prompts in 2026.
Where does the disagreement between 20% and 48% prevalence numbers actually come from?
Short answer: different datasets explain the gap. 108 million queries from Brand Radar in 2026 anchor the lower figure near 20% prevalence on US Google searches. Aggregator pieces citing 48% usually rely on smaller third-party scrapers with different query-mix assumptions.
Neither number is dishonest. They just measure different slices of the web. Anchor on the lower verified figure when you talk to clients.
What kind of structured data matters most for product listings inside Gemini’s outputs?
Product schema with price, availability, reviews, and identifier fields like GTIN. The May 2026 Google Search Central reference at developers.google.com/search/docs/appearance/ai-features names this as the input Gemini uses for shopping experiences.
Add the basic Product schema first. Add offer details only when stock and price are real. Add the review block only when reviews are real and approved by your reviewer.
How should you put the guidance to work on your site?
Read the page once. Slowly.
Pick the three or four sentences that name a check you can actually run. Indexing in Search Console. Structured data on the product or service template. A static-HTML render of the homepage.
Run those checks. Then re-test the same query in AI Overviews and AI Mode a week later.
You will not see a dramatic change. AI search visibility moves slowly. But the audit gives you a baseline you did not have before. Next month’s audit can move that baseline up.
The honest framing is this. Brand Radar’s 108M-query dataset puts AI Overviews on roughly 20% of US Google searches in 2026, not the 48% some aggregator pieces claim. That is still a lot.
But it is not yet most of search. Treat AI optimization as a layer on top of regular SEO, not as a replacement for it.
If you read the guidance and still feel a little uncertain about which checks to run first on your own site, that is normal. Most small business owners do. You can contact me here, and we can read the page together against your homepage. I do not charge for this kind of call, and there is no pitch waiting at the end.
I will tell you which two or three sentences on the Google page actually apply to your site, and which ones you can safely skip. That is usually enough to get a small site moving without a stack of new tools.