Google AI Overviews Audit: Why You Are Not Showing

You cannot submit a page to Google AI Overviews or pay to appear in them. To show up, a page must be indexed and eligible for a snippet, match a question that triggers an overview, and offer something more useful and original than the other pages Google could pull from. That makes the fastest route a structured audit: check eligibility, find which of your queries trigger overviews, compare your page with the ones being cited, fix the gaps and measure the result.

This post is the audit and fix plan. It is deliberately different from our earlier guide on how to get cited in Google AI Overviews, which explains what Google says and how to structure a page. Here we take Google's newer May 2026 guidance on generative AI features and turn it into a diagnostic you can run on any page that is not appearing. It sits within our wider answer engine optimisation work.

How do AI Overviews choose which pages to show?

Google's documentation describes two techniques behind its generative AI features, and understanding them explains most of what follows.

• Retrieval-augmented generation. The system uses Google's core ranking systems to retrieve relevant, current pages from the Search index, reads them and writes a response with clickable links to supporting pages. Your page has to be retrievable before it can be quoted.
• Query fan-out. The model generates several related searches at once to gather more information. A single question can therefore pull in pages that rank for different, related queries.

The practical consequences are simple. Pages that already rank well and are clearly relevant are the pool of candidates, and a page that covers the related sub-questions around a topic has more chances to be chosen. Google also calls this work "still SEO", so a good audit starts with fundamentals.

Step 1: Build a query list worth auditing

Choose 20 to 30 queries where an AI Overview could realistically help your business. Prefer question-style queries and informational searches near a buying decision, such as definitions, "how do I" and "what is the difference between" questions. For each query, note the page on your site that should answer it. If you have no page, that is already a finding.

Then search each query in a clean browser window, with your target country and language, and record four things in a sheet: whether an AI Overview appeared, whether your domain is linked in it, which domains are linked, and where your page ranks in the normal results. Repeat the check on another day, because overviews do not appear consistently for every search.

Step 2: Audit technical eligibility

Google states that to appear in generative AI features a page must be indexed and eligible to be shown in Search with a snippet. Its guide also says a site must be included in the Search generative AI features setting in Search Console to be eligible, so confirm that for your property. Then check each page on your list:

• Indexed? Use URL Inspection in Search Console. "Discovered, currently not indexed" and "Crawled, currently not indexed" both mean the page is not in the pool yet.
• Snippet allowed? Look for nosnippet in robots meta tags or headers, data-nosnippet on the paragraphs that hold your answer, and a very low max-snippet value. These controls limit how your text can be used.
• Crawlable? Make sure robots.txt does not block Googlebot from the page or its key resources, and that the answer text is in the rendered HTML rather than loaded only after interaction.
• Canonical and duplicates? If another URL is the canonical, that URL is the one considered. Duplicate pages dilute signals.
• Page experience? Check mobile layout, speed and whether the main content is easy to tell apart from banners and pop-ups.

Fix technical blockers first. No amount of rewriting helps a page Google is not allowed to show.

Step 3: Run the originality test

Google's guide says that unique, non-commodity content that is helpful and people-first will likely influence generative AI visibility more than any other tip it lists. Commodity content is based on common knowledge that could come from anyone and adds little for readers. Score each page honestly:

1. First-hand input. Does the page include something only you know, such as your own data, process, test or experience?
2. Specificity. Does it give concrete numbers, steps and conditions, or only general statements?
3. Distinct view. If someone asked a general-purpose AI to write the page, would the result say the same thing?
4. Completeness. Could a reader act on it without opening another site?
5. Freshness. Is the date visible, and are the facts still correct?

A page that scores poorly on the first three is a rewrite candidate. Add what only you can add. For example, replace "regularly review your results" with the exact cadence you use and what you do when a number moves.

Step 4: Check coverage of related questions

Because of query fan-out, think about the neighbouring questions a model might ask on the way to answering your target query. List the follow-up questions a careful reader would have: definitions, steps, costs, risks, alternatives, how to check it worked. Then check whether your page answers them in clearly headed sections.

Take care here. Google warns that creating separate content for every variation of how people search, or for fan-out queries, primarily to manipulate rankings or AI responses violates its scaled content abuse spam policy, and that it is ineffective anyway. The right response to a coverage gap is to make one page more complete, not to publish ten near-duplicates. Create a new page only when a sub-question deserves its own full answer for real readers.

Step 5: Structure the page for humans

You do not need to cut content into tiny chunks or write in a special style. Google says it can understand pages with several topics and show the relevant piece. Ordinary good structure is enough:

• A heading that states the question or topic plainly.
• An opening that answers it in a sentence or two, then a fuller explanation.
• Short paragraphs, descriptive subheadings and lists where steps or options really are lists.
• Images or video where they help, with descriptive alt text and surrounding text, since overviews can include media too.
• A named author or organisation, a review date and links to the sources you rely on.

Step 6: Know what to ignore

The same Google guide lists several popular tactics it says you can ignore for Google Search. Spending time on them is time not spent on the steps above.

• llms.txt and other special files. Google Search does not use them. It will neither help nor harm your visibility. See our explainer, llms.txt explained: what it is and whether you need one.
• Chunking content into tiny pieces for AI.
• Rewriting content just for AI systems. Synonyms and meaning are understood, so you do not need every long-tail variant.
• Seeking inauthentic mentions around the web.
• Overfocusing on structured data. No special schema is required, though standard markup that matches the page remains good practice. Our post on FAQ schema and AEO covers the details.

Diagnose by symptom

Match what you see in Step 1 to the likely cause.

No AI Overview appears for the query

Not every query triggers one. Do not force it. Move your effort to queries where overviews do appear, and keep the page strong for ordinary results.

An overview appears but links only other sites

Open the cited pages and compare them with yours using the originality test. Look for what they offer that you do not, such as clearer definitions, original data or a more complete answer. Improve your page on that dimension.

You were cited and then dropped out

Check for technical changes first: noindex, snippet controls, redirects or a rendering fault. Then look for a fresher or more complete competing page. Set an alert so you catch these changes quickly. Our guide to a SERP alerts tool shows how.

You rank well but are not cited

Your page may be relevant but not the clearest or most original source. Tighten the opening answer, add evidence and fill the related-question gaps.

Your clicks fell after an overview appeared

This is a real trade-off. Concentrate on queries where being the cited source builds recognition, write pages that give readers a reason to click through for detail, and measure conversions as well as traffic.

Step 7: Measure what happens

Google recommends the Generative AI performance report in Search Console for seeing how your content performs in AI features on Google Search. Check it each week, and combine it with your manual query log so you can see which pages and queries moved. Google also cautions that no third-party tool has access to its internal ranking or AI systems, so treat external trackers as samples of what searchers see. Our guides on how to track AEO and the AEO and GEO KPIs scorecard give a complete measurement routine, and our post on getting cited by ChatGPT covers the same discipline for another engine.

A 30-day audit and fix plan

• Week 1: build the query list, run the baseline checks and fix indexing and snippet blockers.
• Week 2: score pages on originality and coverage, and pick the five weakest pages that matter most to your pipeline.
• Week 3: rewrite those five pages with first-hand evidence, an answer-first opening, clear headings and visible dates.
• Week 4: request indexing for updated pages, set alerts and begin weekly measurement. Judge results over several weeks, not days.

How Insyrge can help you grow

Insyrge runs AI Overview audits as part of its answer engine optimisation and SEO work. We check eligibility, benchmark your pages against the sources currently cited, rewrite the pages that matter most and set up measurement and alerts tied to leads in your CRM. You get a prioritised plan and a clear view of progress. Begin with the free AI visibility check, explore our AEO agency service and SEO services, or contact us to book an audit.

Frequently asked questions

Can I submit my site to Google AI Overviews?

No. There is no submission process. Eligibility depends on a page being indexed and allowed to show a snippet, and selection depends on relevance and quality.

Do I need special schema markup to appear?

No. Google says no special markup is required. Keep using accurate structured data as part of normal SEO, but do not expect it to unlock overviews on its own.

Do I have to rank in the top ten to be cited?

Ranking well helps because the system retrieves from the Search index, and query fan-out can also surface pages that rank for related queries. Treat strong organic performance as the base and relevance and originality as the differentiators.

Should I write a separate page for every related question?

Generally no. Make your main page complete. Google warns against creating pages for every query variation mainly to influence rankings or AI responses, because it can break its spam policies.

Does a nosnippet tag stop my page appearing?

Restrictive snippet controls limit how your content can be shown, and Google requires snippet eligibility for generative AI features. Remove them from pages you want to be cited.

How do I know if it is working?

Use the Generative AI performance report in Search Console and a weekly manual log of your priority queries. Look for gains in citations and impressions over several weeks.