Google AI Overviews and AI Mode: website eligibility

A website needs to be eligible before Google can use it as a supporting source in AI Overviews or AI Mode. Eligibility is a starting condition, not a promise of selection. The useful question is which requirement you can verify and which result remains uncertain.

Google’s current AI optimisation guidance points to indexing, snippet eligibility and inclusion in its Search generative AI features. Ordinary search fundamentals still matter. A special “AI version” of every page is not required.

Follow this troubleshooting flow

1. Can the intended page be reached? Open its final public URL. Confirm that the actual service explanation loads, rather than a login screen, error or empty template. If it fails here, fix the website before testing AI answers.
2. Can Google crawl the page? Check robots rules, server responses and any verified crawler blocks in the firewall. A browser working for you does not prove that Google receives the same response. Ask for the specific finding and affected URL.
3. Is the right URL indexed? Use Search Console’s URL Inspection for the page you intend to promote. Compare the declared and selected canonical where reported. An indexed duplicate or an old URL may explain why the wrong version appears.
4. Is the relevant content eligible for a snippet? Review page-level and HTTP-header controls. Find out whether important information is deliberately excluded from snippets before proposing content changes.
5. Does the property permit Search generative AI inclusion? Check the setting and any inherited parent-property decision. A technically accessible page can still sit under an exclusion choice.
6. Is there a strong answer to the buyer’s question? Only after access checks, assess whether the page offers relevant facts, evidence and a useful next step. Technical eligibility cannot compensate for a vague offer.

Understand the separate controls

Google documents a Search generative AI control with include, exclude and inherit options. An exclusion applies to the covered generative features; Google says it does not itself change ordinary Search inclusion or ranking. Check the applicable property rather than assuming that a parent setting tells the whole story.

The Google crawler reference explains that Google-Extended controls specified training uses and grounding in Gemini Apps and Vertex AI. It is not solely a training toggle and does not control inclusion or ranking in Google Search. Keep those choices separate from the Search generative AI control. Do not change settings simply because a generic audit says “allow AI bots”.

For an owner, the practical deliverable is a short record of each current setting, why it exists and who approved a change. Confidential pages should have proper access protection; relying on search exclusion alone is inappropriate.

Ask for observable evidence

For example, a building surveyor might have a public page explaining what a survey includes but omit whether it covers flats or older properties. Answering those buyer questions is a defensible improvement. Calling that change “guaranteed AI inclusion” would not be.

Measure the right result

Google’s Generative AI performance report reports impressions for AI Overviews and AI Mode. Its documentation says the worldwide rollout completed on 31 August 2026. Access and data still need checking in your own property; a missing report does not automatically mean zero visibility.

Treat those impressions separately from visits and enquiries. The report’s activity is included within overall Web Search reporting, so do not add the two totals as if they were separate audiences. Preserve unavailable values when exporting; the documentation notes that some unavailable markers become zeros.

Use a consistent date window and note page changes, campaigns and seasonal effects. A rise after an edit is worth investigating, but it does not isolate the edit as the cause. Our tracking template helps preserve the observation conditions.

Prioritise useful fixes

Start with blocked or inaccurate commercial pages, then improve the explanations buyers need. Do not make a new page for every acronym. Our guide to schema and llms.txt separates useful implementation from optional extras, while the AI visibility audit guide covers the wider review.

If you need the findings turned into a bounded implementation plan, see our answer engine optimisation services. The deliverable is verified work and clear measurement, with selection by Google remaining outside any supplier’s control.

Official guidance checked on 27 October 2026.