AI Overviews: How to Be the Cited Answer

The short answer: When Google shows an AI summary, people click an ordinary result about half as often, and almost nobody clicks the links inside the summary. You cannot switch it off for your queries without hiding your own snippets. So the goal changes from winning the click to being one of the sources the answer is built from, with your brand name in it. The way to do that is not a technical trick. It is writing pages that answer one real question completely, in the first few lines, with a fact nobody else has.

This guide shows the numbers, a paragraph rewritten answer-first, a table for mapping your customers' questions, and what to report when clicks stop telling the story.

What the research says

Source	What it measured	Finding
Pew Research Center, US, March 2025	68,879 real Google searches by 900 adults	People clicked a result on 8% of visits with an AI summary, against 15% without. Links inside the summary were clicked on 1% of visits
Same Pew study	How sessions ended	26% of visits with a summary ended the browsing session, against 16% without
Ahrefs, 300,000 keywords, December 2023 against December 2025	Click-through rate of the top result	58% lower when an AI Overview is present. Its earlier study had put the figure at 34.5%
Seer Interactive, 3,119 queries across 42 organisations, to September 2025	Brands cited in the AI Overview against brands not cited	Cited brands had an organic click-through rate of 0.70%, against 0.52% for uncited ones

Two honest notes. Pew's data is from the United States, and no equally careful study exists for India yet. And Seer says plainly that it cannot prove the citation causes the extra clicks. Stronger brands may simply be cited more often.

Pew also shows which searches trigger a summary. About 18% of all searches in its sample did. For searches phrased as a question the figure was 60%, and for searches of ten words or more it was 53%. For one or two words it was 8%. Long, specific questions are exactly where AI answers appear, and they are also the questions people ask just before they buy.

How Google picks what to quote

Google's documentation is direct about this, and it removes a lot of myths.

AI Overviews run on the normal search systems. A page has to be indexed and eligible for a snippet. There are no extra technical requirements.
No special markup or file is needed. Google says structured data is not required for its AI features, and that files such as llms.txt do nothing for Search.
The AI searches more than once. Google describes a "query fan-out": one question is broken into several related searches. A page that also answers the obvious follow-up questions has more chances to be picked up.
Unique beats common. Google's guide separates commodity content, which repeats what everyone knows, from content with a first-hand or expert view. It asks for the second.

One more change to know about. Google stopped showing FAQ rich results on 7 May 2026 and has since removed the documentation. Agencies still selling "FAQ schema for AI Overviews" as a deliverable are selling something Google says it does not need. Clear questions and answers on the page still help readers. The markup is no longer the point.

Before and after: an answer-first rewrite

Here is a paragraph of the kind most brand blogs publish. The brand is made up, and the figures in the rewrite are there to show the shape of a good answer, not to advise you on purifiers.

Before

"Choosing the right water purifier for your home is an important decision for every family. With so many options available in the market today, it can be confusing to know which technology is best. At Brand A, we believe every home deserves pure water. In this blog, we will explore the different types of purifiers and help you make an informed choice."

That is 60 words and not one fact. Nothing in it can be quoted.

After

"Which water purifier is right for my home? It depends on your water's TDS level, which a ₹200 meter will tell you. Below 200 ppm, as with most municipal supply, a UV or UF purifier is enough. Between 200 and 500 ppm, use RO with a mineral cartridge. Above 500 ppm, as with most borewell water, you need RO. The table below shows typical readings and what we would install for each."

What changed:

The heading is the customer's question, in the customer's words.
The answer comes first and stands on its own. Pew found the median AI summary is 67 words, so an answer of 40 to 70 words is the right size to be lifted whole.
It contains specifics: thresholds, a price, a clear recommendation for each case.
It promises the detail below, so the person who does click has a reason to.

This is not "chunking for AI". Google says you do not need to break content into pieces for its systems. It is simply what a good shop assistant does: answer first, explain after.

A quick test: read only the first two sentences under each heading. If a customer would be satisfied, the page is quotable. If they would still be waiting for the answer, rewrite it.

Map your customers' questions

Do this once, in a sheet. Start with twenty questions from sales calls, WhatsApp chats and search queries in Search Console. Search each one on Google and fill in the columns. Two rows filled in for the same made-up brand:

Customer question	AI Overview shown?	Who is quoted	Do we have a page?	Action
Which water purifier for borewell water?	Yes	Two rival brands, one news site	A product page, no guide	Write an answer-first guide with TDS table
How often should an RO filter be changed?	Yes	A rival brand, a YouTube video	Blog post, answer in fifth paragraph	Move the answer to the top, add a service-interval table

Then prioritise. Questions with an AI answer where a competitor is quoted and you already have a page are the quickest wins, because you only need to rewrite. Questions with no page come next.

Add original material wherever you can: your own service data, a test you ran, real photos, prices. That is the part no competitor and no AI can produce without you.

What to measure instead of clicks

Clicks alone will now under-report good work. Report these alongside them.

Measure	Where to find it	What it tells you
Citations in AI answers	The Generative AI performance report in Search Console, plus a manual check of your twenty questions each month	Whether you are a source
Impressions for your key questions	Search Console	Whether you are being shown, even when nobody clicks
Branded search volume	Search Console, filter queries containing your brand name	Whether being named is making people look for you
Conversion rate of organic visits	Your analytics	Whether the fewer visitors you get are better ones. Google says clicks from pages with AI Overviews are higher quality
Leads and sales from organic	Your CRM or store	The only number the business cares about

If impressions and branded searches rise while clicks fall, the work is succeeding and the reporting needs to catch up.

Four ways brands get this wrong

Burying the answer under an introduction about the brand.
Writing one page per keyword variation. Google's guide specifically warns against creating many versions of the same content to influence AI answers.
Blocking snippets to "protect" traffic. The controls that keep you out of AI Overviews, such as nosnippet, also remove your ordinary snippet.
Reporting clicks alone and concluding that search has stopped working.

Where this fits

Rewriting pages so they are quoted is what we mean by AEO. It sits on top of sound SEO, because Google's AI features draw on the same index and ranking systems, and it pairs with GEO, which covers ChatGPT, Gemini and Perplexity. For how those tools choose brands, see who ChatGPT quotes.