Guide · Getting started
Understand Google AI search
Review Google AI Overviews with the right eligibility checks, source evidence and reporting limits.
Name the experience
Record whether you saw an AI Overview or ordinary search results. Google describes these as different experiences.
Keep the screenshot, exact prompt, country, language, date and sign in status. Write the experience's name beside every answer. One experience cannot establish another's result.
Google says the best practices for SEO remain relevant for AI Overviews. Good SEO is still the starting point.
Write down the device as well. A phone and a computer can lay out the same results differently, and a reader comparing screenshots needs to know which was which.
A different location or a follow up question makes a different comparison, so save it separately. Notes like these explain what you saw without suggesting everyone gets the same response.
Use a private browsing window, or note that you didn't. Your own search history and account settings are part of the context, and a reader can't see them.
Note it if you searched through a virtual private network or from abroad. Your apparent location then differs from your real one, and your customers' searches may differ from yours.
Expect AI Overviews to be missing on many searches
Google says AI Overviews are only shown when its systems judge them additive to classic Search. It adds that, as a result, they often don't trigger.
So a search with no AI Overview is a normal result, not a fault. Record it as no Overview shown, with the date and the exact wording you typed.
Google also says AI Overviews may use different models and techniques. The responses and links they give will vary as a result.
That is why each experience belongs in its own row. An AI Overview and ordinary search results for the same words are two observations, not one observation seen twice.
Ordinary results sit on the same page as an AI Overview. Save them separately if your review needs them, because they answer a different question.
Check eligibility first
Google says a supporting page must be indexed and eligible for a search snippet. Ask your website maintainer to inspect the exact page, and keep the date and the finding.
An eligible page can still be missing from an answer. Treat eligibility as a starting check, and keep Search Console findings with the time you inspected.
Eligibility belongs to a page, not to a whole website. Check the exact page you expect to support the answer, because a strong home page says nothing about a weak service page.
Google's newer guide adds that the site must be included in Search generative AI features in Search Console.
Inspect the page the way Google sees it
The URL Inspection tool in Search Console reports on the version of a page Google last indexed. It also shows when Google last crawled it.
Compare that date with the date the page last changed. If the crawl came first, Google's index holds an older copy than the one your customers see.
The tool also shows the canonical page Google selected. If Google chose a different address from the one you intended, note it, because that page may be the one linked.
Google says a status of URL is on Google means the page is eligible, not that it is certain to be shown. Record the status as a check, not as a result.
You can view the HTML Google received while crawling. Use it to confirm the important text is there, rather than only appearing after scripts run in a browser.
Run the basic checks Google lists
Google's AI features page describes the ordinary Search technical requirements below. Check the Search Console inclusion setting described earlier as well.
Crawling should be allowed in robots.txt and by any content delivery network or hosting service in front of the site. A firewall rule can block Googlebot without anyone noticing.
Important content should be available as text, and reachable through internal links. Images and video can support the text but shouldn't replace it.
Structured data should match the visible text on the page. Google also asks site owners to check that their Merchant Center and Business Profile information is current.
Google says you don't need new machine readable files, AI text files or special schema.org markup for these features. Other assistants publish their own guidance, so check each one separately.
Treat each item as a yes, no or unknown for the page you chose. Hand the no and unknown items to whoever maintains the site.
Ask the owner to review any proposed change and record what they publish and when. Keep the old answer so the next review has a clear starting point.
Read the page as a customer would, from the top. If the answer to the question sits low on the page or only inside an image, note that as a finding.
Write proposed changes as a short list of facts with their sources.
Follow the supporting link
Open the link attached to the detail you're checking and compare the visible page with the answer. Record the full address and the sentence that matters.
If the page changed after the answer, keep both dates. A link may back only part of an answer.
Check whether the linked page is yours, a directory, a review site or a news article. Each kind has a different owner, and only some of them are yours to change.
If the link opens a different page from the one you expected, record both addresses. A redirect or a moved page can make an old link point somewhere new.
Understand related searches
Google says its AI features may search related subtopics when building an answer. It calls this query fan-out. Your record may not show those searches.
Mark unseen queries as unknown. Plan your explanations around customer questions, not around a hidden list of searches you can't see.
Fan-out is a reason to cover a topic properly, not to chase every phrasing. A page that answers the main question and the obvious follow up questions in plain text can be read for each of them.
Look at the follow up questions customers actually ask you by phone or email. They are a better guide to related subtopics than guesses about hidden searches.
Know which controls change what Google can use
Google says robots.txt rules for Googlebot are the control for how a site is crawled for Search. It describes AI as built into Search, so the same Googlebot rules apply.
To limit what Google shows from a page, its guidance points to nosnippet, data-nosnippet, max-snippet and noindex. These are rules a maintainer adds to the page or its headers.
Google's robots meta tag documentation says nosnippet also stops the content being used as a direct input for AI Overviews.
Check that none of these rules was added by accident. A nosnippet left on a template can affect every page built from it.
Google-Extended is a different control. Google says it governs whether crawled content may train future Gemini models or ground answers in Gemini Apps and on Vertex AI.
Google also says Google-Extended doesn't affect a site's inclusion in Google Search. Blocking it is a decision about other Google products, not about AI Overviews.
Record which controls are in place on the date of each review. A changed rule is one of the other changes your comparison has to mention.
Know the measurement limits
Google counts its AI features within overall Web search traffic in Search Console. That total doesn't isolate the answer you saved.
Keep answer observations apart from website visits, and ask which measurement source backs each number. A rise in overall traffic doesn't point to an AI feature as the cause.
Keep your saved answers as a separate record with their own dates. A traffic total can't tell you what any answer said.
Google's newer guide also documents a Generative AI performance report in Search Console. Check which report your property provides and its coverage.
Find AI Overview visits in Analytics
Google Analytics sorts visits into channels. Google defines an AI Assistant channel for visits from sources such as ChatGPT, Gemini and Copilot.
Google says that channel excludes AI Overviews. Visits from those two features are counted in the Organic Search channel instead.
So you can't read AI Overview visits from the AI Assistant channel. You also can't separate them from ordinary search visits using the default channels alone.
Write the channel name and its definition beside any visit figure you report. A reader who assumes AI Assistant covers every AI source will read the figure wrongly.
Visits are one more kind of evidence, apart from the answers you saved. A visit doesn't show which answer the person read before clicking.
Repeat the review
Run the same review again after the owner makes a change. Save the new answer even if it's unchanged.
Describe the two observations and the time between them. Other changes may have happened too, so a later citation can't show your page edit caused the assistant to pick it.
Keep the interval between reviews steady, and long enough for a recrawl to happen. A review the day after a change mostly measures the old index.
Report a wrong statement to Google
Google's help pages say AI responses may include mistakes. When an AI Overview states something wrong about your business, you can tell Google directly.
Google says each overview has thumbs up and thumbs down icons at the bottom. Choosing thumbs down lets you report a problem and pick the category that fits.
Google says the feedback you submit includes your most recent search and its results. Don't search for anything private just before you send it.
Keep a copy of what you reported, with the date and a screenshot of the overview. Feedback is a request, and there's no promise the wording will change.
Correct your own page as well, if it contributed to the mistake. Then repeat the review on the usual schedule and record what you see.
A worked example
Fern Workshop records an AI Overview for one search and no AI Overview for a separate search on the same service.
| Detail | Observation | Review or next step |
|---|---|---|
| AI Overview | The saved answer links to the service page. | Keep the response and its linked page together. |
| No AI Overview | No AI Overview appeared in the separate search. | Record that on its own. It doesn't cancel the saved answer. |
Keep each observation apart. A missing Overview in one search doesn't cancel the answer saved from another.
Common questions
Does eligible mean included?
No. Google says eligibility does not ensure crawling, indexing or inclusion.
Can an answer record isolate Google AI traffic?
No. An answer record isn't a traffic report. Keep each measurement tied to its own source.
Does eligible mean included?
No. Google says eligibility does not ensure crawling, indexing or inclusion.
Can an answer record isolate Google AI traffic?
No. An answer record isn't a traffic report. Keep each measurement tied to its own source.
Check your understanding
An eligible page is absent from one AI Overview. What follows?
Read the answer and explanation
- It can never be cited.
- Eligibility proves inclusion tomorrow.
- Eligibility and observed selection are different findings.
Answer. Eligibility and observed selection are different findings.
Google says eligibility does not ensure inclusion. Keep the page finding apart from the answer record.
Answer and explanation
Eligibility and observed selection are different findings.
Google says eligibility does not ensure inclusion. Keep the page finding apart from the answer record.
Your next steps
Sources and method
A Carqo review method. Provider statements link to the relevant documentation.
- Google guidance on AI features Link checked 4 November 2026
- Search Console Help, URL Inspection tool Link checked 4 November 2026
- Google Search Central, robots meta tag, data-nosnippet and X-Robots-Tag Link checked 29 October 2026
- Google crawling documentation, Google's common crawlers Link checked 4 November 2026
- Google Analytics channel definitions Link checked 4 November 2026
- Google Search Help, AI Overviews in Google Search Link checked 29 October 2026
- Google, optimizing your website for generative AI features on Google Search Link checked 4 November 2026