Google AI Overview and AI Mode: eligibility, controls and measurement

What an AI Overview is, and how AI Mode differs

Google's Search Help describes an AI Overview as an AI-generated snapshot with key information and links to dig deeper. It appears inside the normal results page, usually near the top, and has three parts you'll see on most of them: a written summary, link cards or inline links to the supporting pages, and a Show more button that opens the full overview. From there a searcher can type a follow-up question and keep going.

It doesn't appear on every search. Google's Search Central page on AI features says AI Overviews are only shown when its systems decide they add something to classic Search, and as such often don't trigger. So two people searching slightly different wording can see different pages, and the same search can show an overview one week and not the next.

AI Mode is the separate, conversational tab in Google Search. Google says it's most helpful for questions that need further exploration, reasoning or complex comparisons, and Search Console counts each follow-up question in AI Mode as a new query. Google announced AI Mode for Australia, in English, in October 2025. Both features sit inside Google Search, so the eligibility rules, controls and reports in this guide apply to both.

The table separates the four kinds of answer an owner is most likely to confuse. The useful column for a business is the last one: each answer draws on different sources, so each needs a different check. For the bigger picture of how AI tools choose sources, see how AI search works.

How Google builds an answer and picks its links

Google doesn't publish its exact method, but its documentation describes the broad steps. Knowing them explains why eligibility comes first and why a page that ranks well can still be left out of the summary.

The gap between ranking and being cited matters for a local business. A page can sit high in the ordinary results directly under an AI Overview without being one of its links, and a page further down can be linked because one passage answered one of the fan-out searches well. Google says these features show a wider and more diverse set of helpful links than a classic results page. So treat the overview and the blue links as two separate places your page can appear, and check each one.

Your searches also differ in how often they're likely to carry an overview. Google doesn't publish trigger rules, so the table below is a way to sort your own keyword list for checking, not a forecast. Run the searches yourself and record what appears.

1. The searcher types a question or a phrase, or asks a follow-up in AI Mode.
2. Google decides whether an AI answer would add anything. For many searches it doesn't, and the normal results show on their own.
3. Both features may use what Google calls query fan-out: the system issues several related searches across subtopics and data sources to build a response.
4. Those searches return pages from Google's Search index. Only pages that are indexed and eligible to show a snippet can be used.
5. Google's generative AI writes the summary from that material, and Google attaches links to pages that support what it says.
6. The searcher reads the summary, opens a link, or asks a follow-up and moves into a longer conversation.

Eligibility: a page-by-page check

Google's guide to optimising for generative AI features in Search sets out the entry rules. A page must be indexed, it must be eligible to show in Google Search with a snippet, and it must meet the Search technical requirements: Googlebot isn't blocked, the page returns a working HTTP 200 response, and it has indexable content. The site must also be included in the Search generative AI setting in Search Console, which Google had rolled out to every site by 31 August 2026. Included is the default, so most sites qualify without touching it.

That's the full list. Google's documentation says there's no special schema.org markup to add, no AI text file such as llms.txt to create, and no need to break pages into tiny chunks or write in a special style for AI. Structured data can still make a page eligible for rich results in normal search. It just isn't an entry ticket to AI answers. Google also says that meeting the technical requirements doesn't guarantee a page will be indexed, let alone linked.

Run this check on the pages that bring in work, such as your main service pages, before you touch the content. It takes a few minutes per page in Search Console.

1. Paste the URL into the URL Inspection tool. "URL is on Google" means the page is indexed and can appear. If it isn't, read the "Crawl allowed?", "Page fetch" and "Indexing allowed?" lines to find what's stopping it.
2. Compare the user-declared canonical with the Google-selected canonical. Google treats your canonical as a hint, and results usually point to the page Google picked. If that's an old duplicate, sort the duplicate out first; the guide to noindex, canonicals and redirects explains which fix fits.
3. Look for snippet blockers in the robots meta tag or the X-Robots-Tag header: noindex, nosnippet or max-snippet:0. Plugins and staging settings sometimes leave these behind after a site launch.
4. Check robots.txt isn't blocking Googlebot. Google's robots documentation notes that a page blocked from crawling hides its noindex and snippet rules too, because Google never reads them.
5. Open Settings, then Search generative AI, and confirm the property is included. A property inherits this setting from its closest parent that changed it, so check the domain property as well as any URL-prefix property.
6. Make sure the key facts sit on the page as text. Google's AI features page recommends keeping important content in text form, supported by images and video, so an answer locked inside an image is hard to use.
7. Record the result for each page with the date, so a later change has a starting point to compare against.

Controls that limit what Google shows

Each control does one job. Pick the narrowest one that solves the real problem, because the broad ones also remove traffic you probably want.

The snippet controls carry a cost owners often miss. Google's robots meta tag documentation says nosnippet applies to all forms of search results, including ordinary web search, so the page's blue-link listing loses its text preview as well. A listing with no description gives a searcher less reason to click, which is why nosnippet suits a page you'd rather nobody quoted, and rarely suits a service page that has to win the click. A low max-snippet value has a softer version of the same trade-off.

Section-level exclusion works through a URL-prefix property: add a property for one folder of the site and set the Search generative AI control there. It isn't instant either. Google's help page says exclusion generally takes a few days, and some content can take longer because of caching.

Should a local business opt out at all? Rarely for the whole site. Google's June 2026 announcement of the control says sites that opt out won't receive traffic or impressions from its generative AI features. For a business that wants calls, that's a real cost. The narrower cases where holding content back can make sense:

None of these controls corrects a wrong fact. If an AI Overview repeats an old phone number or service area, fix the page and your Business Profile at the source, because hiding the snippet also hides the page that carries the right answer. The guide to AI answers and accurate company facts covers keeping those details consistent. Before changing any setting, note which pages currently earn AI impressions so you can see the effect, then test on one section before touching the domain property.

• Paid or members-only material, such as a course or a resource library, that you don't want summarised for free.
• Health or legal information that could mislead if one sentence is lifted out of context. A nosnippet rule on that page, or data-nosnippet on the block, is usually a better fit than a site-wide exclusion.
• Old sections kept for the record, such as a news archive, where noindex or a redirect is the cleaner fix.
• Material you've licensed from someone else under terms that bar reuse.

What to put on a page once it's eligible

Once the basics pass, Google's guidance is short: it says generative AI features are rooted in its core Search ranking and quality systems, and that content people find unique and useful will likely influence a site's presence in generative AI search. It asks for non-commodity content written from what you actually know, and a good page experience for the people who arrive. In practice that means making each section easy to lift and hard to misread.

Google still picks the links for each search. These are the same habits that help a page earn a featured snippet or a clear blue-link listing, applied with the fan-out step in mind: each passage may be judged on its own against one narrow question. The AI visibility checklist turns them into a repeatable review.

The local inputs sit outside the website. Google's guide says Business Profiles can help a business's products and services show in AI responses as well as other Google Search results, and its AI features page tells businesses to keep Business Profile information up to date. So the hours, phone number, categories, services and service areas on the profile should match the website word for word. The Business Profile audit checklist walks through each field.

• Answer the question in the first sentence or two of each section, then add the detail. A section that opens with background makes the answer harder to find.
• Make each passage stand on its own. Name the service and the place in the sentence ("hot water repairs in Caloundra"), so a quoted line still makes sense without the paragraph before it.
• Write the facts a customer checks: what the job involves, what affects the time it takes, what the customer needs to do before you arrive, and which suburbs you cover.
• Add what only your business knows, such as the questions customers ask on the phone, local conditions you deal with, or the checks you run on every job. That's the non-commodity content Google asks for.
• Keep tables and lists as real HTML text, and keep important details out of images and PDFs.
• Date and review pages that state rules, licences or service areas, so an answer drawn from them stays right.

How to measure AI Overview and AI Mode visibility

No single report answers everything. Use each one for the question it can answer, then add a manual check for the part no report shows.

Read clicks and impressions together. When an overview answers a question on the results page, a page's impressions can hold or rise while its clicks fall, because some searchers get what they needed without clicking. A falling click-through rate alongside steady positions and AI impressions points to a change in the results page, not a ranking loss. Google says clicks from results pages with AI Overviews are higher quality, meaning people are more likely to spend more time on the site, so judge the change by the enquiries the page produces as well as its clicks.

The reports can't show which searches triggered an overview, what it said, or which competitors it linked. A manual log fills that gap. Pick 10 to 20 searches that matter to the business, run them from Australia on the same day each month with a clean browser, and record a row like the one below for each. Results move, so judge the pattern over several months rather than any single check.

Be careful with any tool that claims exact AI Overview rankings. Google's guide says no third-party tool has access to its internal ranking or AI systems, so treat those figures as samples of searches the tool ran, not a count of what your customers saw. Comparing tools fairly is covered in how to compare AI visibility reports, and reading Google's own count in how to interpret Google AI impressions.

• Search Console Performance report: clicks and impressions from AI Overviews and AI Mode are counted in the Web search type, mixed in with ordinary results. You can't split them out here.
• Search Console Generative AI performance report: impressions only, grouped by page, country, device and date, with filters for text-based and multimodal searches. There are no clicks and no queries, and two links from your site in one AI answer count as one impression.
• Position: an AI Overview takes a single position on the results page and every link inside it gets that position. In AI Mode, a follow-up question counts as a new query.
• GA4: visits from AI Overviews and AI Mode land in the Organic Search channel. The AI Assistant channel covers tools such as ChatGPT, Gemini and Copilot, and Google's definition excludes its own AI features.
• Calls and forms: tie enquiries back to landing pages so a change in AI visibility can be judged by work won. Call tracking and conversion reconciliation covers the setup.

Worked example: a Caloundra plumber's hot water page

This example is hypothetical, and the business is invented to show the method in order.

Say a plumbing business in Caloundra finds its blocked drain guide earning AI impressions in Australia most weeks, while its hot water repairs page, the one that brings in paid jobs, has none.

Step one is eligibility. The owner runs the page-by-page check. URL Inspection says the hot water page is on Google, but the Google-selected canonical is an old duplicate URL left over from a site rebuild, and the duplicate is thin. The fix is a 301 redirect from the old URL and a clean canonical, then Request indexing on the live page. Google caps those requests each day and says a request doesn't guarantee indexing, so the owner notes the date and waits for the inspection result to show the right canonical.

Step two is the manual log. The owner runs eight searches, from "hot water repairs Caloundra" to "why is my hot water system leaking from the top". The local searches show the map pack and no overview. Three question searches do show an overview, each linking a manufacturer's page and a competitor's FAQ, and none links the owner's site.

Step three is the content. The owner compares the opening of the linked competitor answers with their own page, which opens with two paragraphs about the business's history. They rewrite each section to answer first: what a leak from the top usually means, when a valve can be replaced and when the unit needs replacing, and what to check before calling. Each answer names the service and the suburb, and the hours and service areas on the page are matched to the Business Profile.

Step four is measurement over the following months. The owner watches the page's clicks in the Performance report, its impressions in the Generative AI report, the monthly log, and the calls tagged to the page. If impressions appear, the log shows which questions the page is being used for. If they don't once the right page is indexed and rewritten, the owner checks the competitor pages again for what they cover that this page still doesn't, and leaves the settings alone, because eligibility was only ever the first filter.