How to rank in AI Overviews: how Google picks its sources

To rank in Google's AI Overviews, a page must first be indexed and eligible to show in Google Search with a snippet. That is Google's only stated requirement. From there, Google runs several related searches behind the scenes and cites pages that answer part of the question clearly and credibly. So the practical route is to rank well for the specific sub-questions your buyers ask, and to write each answer so it can be lifted in a sentence or two.

That is less exciting than most "12 hacks for AI Overviews" posts. It is also what Google's own documentation says, and what I see in practice.

What does Google say about getting into AI Overviews?

Google's page on AI features and your website (last updated December 2025) is short and worth reading in full. The key lines:

• "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."
• To be shown as a supporting link, a page must be "indexed and eligible to be shown in Google Search with a snippet".
• The usual SEO best practices still apply: make important content available as text, keep structured data consistent with what is visible on the page, and look after page experience.

It also says traffic from AI features is counted within the "Web" search type in Search Console's Performance report. There is no separate AI Overviews report in Search Console at the time of writing.

How do AI Overviews pick which sources to cite?

Google describes a technique it calls query fan-out. In its AI features documentation, Google says AI Overviews and AI Mode may "issue multiple related searches across subtopics and data sources" to build a response. Its AI Mode announcement for the UK describes the same thing: breaking a question into subtopics and running many searches at once.

Here is what that means in practice. Someone searches "can I claim R&D tax relief as a small software company?". Google may also search, in the background, for eligibility rules, what counts as qualifying spend, how the claim is filed and what changed recently. The overview is assembled from pages that answer those parts well.

Three consequences follow:

1. You can be cited without ranking first for the main query. A strong page on one sub-question can earn a citation even if a bigger site owns the headline term.
2. Specific pages beat broad ones. One page that covers ten services thinly is unlikely to be the best answer to any sub-question.
3. Your page is competing on a passage, not the whole page. The overview needs a few clear sentences it can rely on.

What kind of page gets cited?

From the pages we see cited for professional services queries, a pattern shows up. The research backs it. The GEO study presented at KDD 2024 found that adding cited sources, quotations and statistics were the most effective changes to a page's visibility in AI-generated answers, each improving their main measure by around 27%. Keyword stuffing slightly hurt.

Does structured data help?

Google recommends making sure structured data matches the visible text. It does not say structured data gets you into AI Overviews. I treat it as good housekeeping: Organization (schema.org's spelling), Person and FAQ markup where it is accurate, and nothing that claims something the page does not show.

Do legal and financial topics face a higher bar?

Google's quality rater guidelines overview describes "Your Money or Your Life" topics, those that could significantly affect people's health, financial stability or safety, as needing the highest standard of trust. Raters do not directly change rankings, but their feedback is used to evaluate Google's systems. For law firms and accountants, that means named, qualified authors and sourced claims are not optional.

Is it worth being cited if people don't click?

This is the fair challenge. Pew Research studied 68,879 Google searches by 900 US adults in March 2025. With an AI summary on the page, people clicked a traditional result in 8% of visits, compared with 15% without one. Links inside the summary were clicked in 1% of visits.

So yes, clicks fall. But for a firm where one client is worth thousands, the question is not how many people click. It is whether the right person sees your name at the moment they are forming a shortlist, and later searches for you directly or calls. We have written more about this in our piece on SEO KPIs and why traffic is the wrong measure.

From our work

In September 2026 we looked at a UK accountancy practice that had lost more than 40% of the searches it appeared for since February. Over a third of what traffic remained came from one news article about HMRC and personal expenses. That kind of general, informational question is exactly what AI Overviews now answer on the results page, so those visits are the first to disappear. The practice's local service page, the one that brings clients, sat on page two.

What we'd do in your first 30 days

Days 1 to 7: map the questions. Take your five most valuable services. For each, list the six to ten questions a buyer asks before they hire (look at enquiry emails and call notes). Search each in Google and note whether an AI Overview appears and which sites are cited.

Days 8 to 14: check eligibility. In Search Console, confirm every relevant page is indexed. Check for nosnippet or max-snippet directives that might limit you. Confirm the answer text is in the HTML source.

Days 15 to 25: rewrite for the passage. For each question where an overview appears and you are not cited, open the pages that are. Write a better answer: first two sentences answer the question, then the detail, with sources, numbers and a named author. Use the question as the heading.

Days 26 to 30: measure properly. Record which queries now cite you. Watch Search Console for impressions on those pages. Most important, track enquiries on those services and ask each new client how they found you.

If you want the wider context, our AI search and generative engine optimisation pages explain how this fits with ChatGPT and Perplexity.

Checklist

• Page indexed and eligible for a snippet
• No accidental nosnippet or restrictive max-snippet
• Answer in the first two or three sentences
• Question-style headings that match how buyers ask
• Specific numbers with linked sources
• Named, qualified author
• Structured data that matches the visible page
• One page per topic, not one page for everything

FAQ

Can I opt out of AI Overviews?

Google lists nosnippet, data-nosnippet, max-snippet and noindex as controls that limit how your content appears in AI features. Using them also limits your normal snippets, so think carefully before you do.

Do I need to rank on page one to be cited?

Not necessarily for the main query, because of query fan-out. But you usually need to rank well for at least one of the related sub-questions.

How do I see if I am cited in AI Overviews?

Search Console does not break it out separately. You need to check the queries yourself, or use a tracking tool, and compare against enquiries. Our post on measuring AI search visibility covers the options.

Does AI Mode work the same way?

Google groups AI Overviews and AI Mode together in its guidance, with the same eligibility rules. AI Mode tends to handle longer, multi-part questions and runs more background searches.

Find out where you're cited

Want more revenue from Google and ChatGPT? Enter your website at our free Search Check. Our team reviews every site and you hear back within one working day.