AI Search
How to appear in Google AI Overviews
A guide to Google AI Overview eligibility, the evidence a credible provider should show and the shortcuts no business should be sold.
If you’re wondering how to appear in Google AI Overviews, the honest answer is simpler than many AI Search pitches suggest. Google says the same search basics still apply. Your page needs to be open to Google, indexed and allowed to appear with a normal search snippet, while the content must be useful enough to support an AI answer.
Making a page eligible is the part an AI Search provider can work on and show you. Google still decides when to display an AI Overview and which pages to include. Nobody can submit your site for a source link, reserve a position or promise that a certain page will appear. Once you know where the provider’s work ends and Google’s choice begins, it’s much easier to judge what you’re buying.
What appearing in an AI Overview actually means
An AI Overview is a summary that can appear above or among Google’s usual search results. Its links let people explore parts of the answer in more detail. When your page appears, it isn’t ‘ranking number one in AI’. It’s one supporting source selected for one query, at one moment and for one searcher. Both the answer and its sources can change as Google uses different models, wording and context.
The useful goal, then, isn’t to win a fixed AI position. It’s to make the right pages eligible, helpful and clear enough for Google to understand. You can then measure whether that reach leads people towards the business. Google’s AI features guidance confirms that a page must be indexed and allowed to appear in Search with a snippet. It also warns that meeting those rules doesn’t guarantee a crawl, an index entry or a source link.
Those conditions create four connected gates. Each one answers a different question, and a provider should be able to explain the evidence at every stage rather than jumping straight to a visibility claim.
Gate 1: can Google access and index the page?
Before anyone talks about AI visibility, they should show whether Google can use the page in normal search. A provider can start with a small set of key service or product URLs. For each one, they should show whether the page loads without a login, Google can crawl it and Search Console reports the intended URL as indexed.
A page blocked by robots.txt, protected behind a login or carrying a noindex instruction can’t pass this first gate. The same is true when Google has crawled a page but chosen not to index it. Finding the cause may call for a wider review of content quality, duplicate pages or site structure. If Search Console reports Crawled - currently not indexed, the provider should work through the relevant indexing checks before they suggest AI-specific markup or copy changes.
JavaScript can also affect what Google receives. Key service details should appear in the page Google renders, rather than relying on an action a crawler may never take. Google’s AI optimisation guide points site owners back to its existing JavaScript SEO advice.
None of these checks can force selection, but they show whether a basic technical barrier makes the AI promise premature. Once Google can access and index the page, the conversation moves from technical eligibility to the reason anyone would choose the page as a source.
Gate 2: does the page answer a real customer question?
Indexation gets a page considered, but it doesn’t make the content useful. Google’s people-first content guidance asks whether a page gives visitors a good answer and adds first-hand detail, thought or experience. That’s a better test than writing in an imagined ‘AI style’.
Imagine an accountancy firm wants to support an Overview about changing accountants. A page that only says the firm is professional, friendly and experienced offers little that another business couldn’t claim. A stronger page could explain:
- which records a new accountant needs;
- who contacts the previous accountant and when;
- what can delay the handover;
- whether the client can move during a financial year;
- which parts vary by business type;
- what the firm has learnt from real handovers.
Those details help a future customer make a decision, which is also why they can make the page a useful source. A provider should draw out real processes, sums, service limits, photographs or named expertise from the business. They can then turn that material into an answer competitors couldn’t reproduce by changing a few words.
That doesn’t mean making dozens of near-identical pages for every version of the question. Google’s AI SEO guidance warns against scaled, low-value content. One well-supported answer is easier to maintain and less likely to compete with the rest of the site. With the substance in place, the next gate is whether Google can link that answer to the right subject and business.
Gate 3: is the meaning clear on the page and across the business?
Google needs to understand what the page covers, who stands behind it and which details it can trust. Clear writing helps, but the signals around the copy need to agree as well. The provider should review the title tag, main heading, section headings and internal links as one joined-up explanation, not as separate boxes to tick.
The answer should appear early, with conditions and exceptions introduced when the reader needs them. Internal links should describe the destination, while claims, calculations and expert advice should identify their source or method. When a named person helps the reader judge the advice, the page should explain their role. Vague credentials or unnamed praise aren’t a sound substitute.
Structured data can describe details people can already see, such as an organisation or local business. Google says the markup must match the visible content, so it isn’t a special AI Overview schema and can’t rescue a thin or blocked page. The same rule applies to Google Business Profile or Merchant Center where those services are relevant. Opening hours, locations, prices and stock shouldn’t conflict with the website.
By this point, the provider should have a clear record of any technical barrier, the useful material added and the signals brought into line. That record becomes the starting point for the final gate. Without it, neither side can tell whether the work changed anything that matters.
Gate 4: can progress be measured without relying on spot checks?
Searching a target phrase once isn’t a sound measure. AI Overviews don’t appear for every query, their sources can change and results may vary by searcher. A screenshot can show that a page appeared once, but it can’t prove lasting reach or business value.
In June 2026, Google announced a Generative AI performance report in Search Console. It can show when Google’s AI features displayed your pages, with views by country, device and date. The report is still rolling out, so some sites won’t have it yet. Where it’s available, a provider should use it with wider Search Console, analytics and enquiry data to ask:
- Which useful pages receive impressions?
- Do those impressions come from countries the business serves?
- Do visits continue to service, product or contact pages?
- Are resulting enquiries qualified and attributable with reasonable confidence?
- Did visibility change after a recorded page improvement?
Google also says traffic from its AI features appears in the overall Web performance report. Those totals can’t always isolate one AI Overview, and Search Console has limits and delayed data. An impression is proof that a page appeared, not that one edit caused the result. searchDecoded’s AI citation tracking guide explains why a mention, a source link and a useful business result are different forms of proof. The Search appearance simulator also shows why a page may support part of an answer rather than own the whole result.
The limits don’t make measurement pointless. They define what a responsible provider can report with confidence and what must remain an informed interpretation.
AI Overview shortcuts that Google’s guidance doesn’t support
Once the four gates are clear, many AI-branded shortcuts become easier to judge. A new label can make familiar search work sound like a new product. The useful question isn’t whether a tactic sounds advanced, but what will change on the site and what proof supports it.
‘You need an llms.txt file for Google’
Google’s current AI optimisation guide says you don’t need new files for machines, AI text files or special markup to appear in its AI features. An llms.txt file may have uses elsewhere, but Google doesn’t list it as a rule for AI Overview eligibility.
‘There’s a special AI Overview schema’
Google says no special schema is required. Supported structured data can help describe visible content when it suits the page. It can’t rescue content that is thin, blocked or missing from the index.
‘Every paragraph must be rewritten into AI-sized chunks’
Short sections can help people scan, but Google hasn’t published a chunk length that guarantees a source link. A provider should shape the answer around the reader’s choices and vary the pace where it helps. Making each paragraph the same size can strip out useful context to meet an invented rule.
‘We can guarantee an AI Overview placement’
Eligibility and selection aren’t the same. A specialist can remove barriers, improve the page and measure the available evidence, while Google decides whether an Overview appears and which sources support it. A permanent place in an organic AI answer isn’t something a provider can sell.
Use the checklist to brief a provider, not replace one
The four gates let you ask for a defined piece of work without taking on the technical work yourself. The separate checklist helps you record the pages and business goals that matter, then compare a provider’s findings and spot gaps in the proof they return.
Build your Google AI Overview provider brief Define the work, compare the evidence you receive and keep the project tied to a useful business outcome.The checklist doesn’t produce a readiness score or predict selection because neither would be honest. Instead, it asks the provider to show the index proof and explain what makes each page distinct. They must also record the changes they advise and define how progress will be reviewed. This keeps the business involved in key choices while the specialist remains responsible for the review, work and reporting.
What a useful AI Search engagement looks like
A useful engagement begins with a set number of pages and a clear business need, then follows the proof through all four gates. The work may involve fixing indexation problems, testing content built with JavaScript or joining duplicate pages. It could also cover structured data or first-hand detail that only the business can supply.
The provider should explain what they found, which changes matter first and who will carry out the work. They should also say what the data can show and where it falls short. Low search demand, gaps in reports or Google’s changing source choices may all limit the finding. That candour is worth more than a dashboard full of unexplained citation counts or a promise to ‘make the website rank in AI’.
The business still has an important role. Only you can confirm which services, locations and customer questions come first, and whether the final copy reflects the work. That input doesn’t turn the project into a do-it-yourself task. It gives the specialist the business context needed to make sound choices.
Make the page ready, then make it worth choosing
Appearing in Google AI Overviews begins with the normal rules for Search. Google must be able to reach and index the page, and the page must be allowed to appear with a snippet. Those rules only open the door.
The real chance comes from giving Google and future customers something worth choosing: a clear answer, first-hand detail, sound facts and proof they can trust. A credible provider should link that content to the technical checks and show the work completed. They should also report progress without claiming control over Google’s final choice.
