- Google's guide on generative AI features, as reported by Search Engine Journal on 15 May 2026, says optimizing for these features is still SEO, with no special markup, llms.txt file or content chunking required.
- A ranking page can be left out because it is not snippet eligible, because it does not answer the sub questions the feature asks, because it repeats what other pages say, or because another page states the answer more clearly.
- Check in this order: indexing and snippets, direct answers, non commodity content, supporting facts and entities, page experience, and the right query set. Then measure with repeated tests, not one search.
Why can a page one result be missing from an AI Overview?
A ranking and an AI Overview citation are two different outcomes. A ranking says your page is one of the results for a query. An AI Overview or an AI Mode answer is built to answer a question, often by combining information from several pages and showing links to some of them. Google has not published a rule that says page one equals included, and we have seen no source that says so. A page can rank well and still be left out because another page answers the specific question more clearly or because the feature asks a different question from the one you ranked for.
Everything we say about Google's own position comes from trade coverage. Search Engine Journal reported on 15 May 2026 that Google published a guide called Optimizing your website for generative AI features on Google Search. We could not open Google's page from our research environment, so treat the summary as reported and read the original before you quote it.
What does Google's reported guidance say?
| Topic | As reported | What it means for you |
|---|---|---|
| Is it a new discipline | Optimizing for generative AI search is optimizing for the search experience, and so is still SEO | Do not abandon normal SEO. It is the base |
| Structured data | Not required for generative AI search. No special schema.org markup needed. Still recommended as part of normal SEO | Use markup that matches visible content. Do not expect it to secure inclusion |
| llms.txt | Not needed | Skip it for Google. See our view in the llms.txt post |
| Chunking | Not needed. Systems understand multiple topics on a page | Do not split good pages into fragments |
| Rewriting for AI | Not needed. Systems understand synonyms and meaning | Write for readers, clearly |
| Inauthentic brand mentions | Not as helpful as they might seem | Do not buy mentions |
| What to focus on | Non commodity content with unique insight, plus standard technical SEO | Say something others do not, and keep the basics healthy |
Check 1: Is the page indexed and snippet eligible?
The reported guide lists indexing and snippet eligibility among the basics. A page that is indexed but blocked from snippets by a robots meta tag, such as nosnippet, or a data attribute may be harder to use in an answer. Check the page source for robots meta tags and the response headers for directives. Inspect the page in Google Search Console. If a plugin or theme adds nosnippet or a restrictive max snippet value across the site, remove it unless you intend it. Wix, for example, documents a nosnippet setting that keeps a page's content from being used as a direct input to AI overviews, so a site owner can switch it on without realising what it does.
Check 2: Does the page answer the question directly?
Take the query you rank for and write the question behind it. Does your page answer that question in the first two sentences under a clear heading? Pages that rank because of authority or links can still bury the answer in paragraph nine. A feature that is trying to answer a question will favour a page where the answer is easy to lift. Rewrite so that each heading is a question a buyer would ask, and the answer follows immediately. See how to write content AI engines cite.
Check 3: Does the page cover the sub questions?
AI features can break a question into several related searches before answering. We explain the idea in query fan out explained. Your page may rank for the main query, while the sub questions about cost, timing, risks and alternatives are answered by other pages. List the sub questions a buyer would ask next, check whether your page answers each, and add what is missing. This is also the cheapest route to content with information gain.
Check 4: Is your content a commodity?
The report says Google points to non commodity content, meaning content with unique insight beyond common knowledge. If your page is a rewrite of the top five results, an AI feature has little reason to prefer it. Add something others skip: a test you ran, a real case, a decision rule, a screenshot, a number with its method. Our post on information gain for AI search sets out how.
Check 5: Are your facts clear and consistent?
If the page names your business, services, prices or location, are those facts the same on your other pages and profiles? An AI feature that finds conflicts is less likely to quote any of them. See entity SEO and brand consistency.
Check 6: Is the page fast and usable?
The report lists page experience and semantic HTML among the standard basics. A page that is slow, covered in pop ups or hard to read on a phone is a poor candidate for a link in an answer. Fix these as you would for ordinary SEO.
Check 7: Are you measuring the right queries?
Many rankings are for short, vague queries that do not trigger an AI feature at all. The queries where AI features appear tend to be longer, specific, question shaped searches. Check Search Console for your long, question shaped queries and test those in AI Mode and AI Overviews. Record whether you are cited and how the answer reads. Do it several times, because answers vary. The measurement method is in why AI answers change between runs.
What if you are cited in Google but not in ChatGPT?
That is a different problem with different causes. Crawler access for OpenAI's bots, for example, is separate from Googlebot. OpenAI's overview, read in October 2026, says OAI-SearchBot is used for ChatGPT search features and obeys robots.txt. If you rank in Google but are missing in ChatGPT, start with the eight causes to check first.
What does a worked check look like?
Take an invented example. A Dubai accounting firm ranks fourth for a query about VAT registration help. In the AI answer for the same question, three other sites are linked. The firm checks its page. Indexing is fine and no nosnippet is set. The answer to the question sits in the seventh paragraph under a heading that says Our approach. The page never says how long registration takes or what documents are needed, which are the two sub questions the feature seems to address. The page is also close to a copy of the government guidance in wording. The fixes are plain: a direct answer in the first two sentences, a section on timing and documents with the date checked and a link to the official source, and one original element, such as a checklist the firm uses with clients. None of this guarantees a link in the answer, but it addresses each plausible reason for being left out, and the firm can retest with five runs of the same prompt before and after.
Notice what the check did not involve: no special markup, no extra file and no rewriting for a machine. It was ordinary editing aimed at the reader's next question.
What should you do next?
Pick your three best ranking pages. For each, run the seven checks and write down which fail. Fix the cheapest first: snippet settings, direct answers and missing sub questions. Retest in a few weeks with a fixed prompt set. If you want the checks and fixes done for you across Google and four other AI engines, the GEO Pilot is a fixed $999 and delivers in 30 days. For a free first look, use the AI Visibility Checker.
