How to get cited in Google AI Overviews (2026 case study)

Between July and September 2026, Google's AI Overviews went from citing one of our clients on none of 13 tracked searches to citing it on 14 of 15. Nothing about that involved special markup, AI files or tricks. It came from ordinary work on the pages that answer the questions its buyers search for.

This guide covers what we tracked, what changed on the site, what the cited pages had in common, what happened during Google's October update, and how to track your own citations in a way you can trust.

How do you get cited in Google AI Overviews?

To get cited in a Google AI Overview, you usually need a page that ranks in the top 10 for that search and answers the question plainly. Google says there's no special optimisation beyond normal SEO. In our tracking, 21 of the 22 Overviews that cited our client pointed to a page already in the top 10.

Google's own documentation is fairly blunt about it. Its AI features page says "there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." A page needs to be indexed and eligible to show with a snippet, and that's it.

So most of the work is ordinary SEO, done properly, on the pages that answer the questions your customers search for.

What we tracked, and how

We tracked 16 searches a buyer might type when looking for a firm like this one, on fixed dates with the same settings: Google UK, desktop, top 10 results, with the AI Overview loaded. For each search we recorded whether an Overview appeared, and whether it linked to the firm.

They were the searches a buyer would make when looking for a firm like this in a specific sector. We ran the same list every time, so the dates can be compared fairly.

Keeping the two counts apart matters, because Google doesn't show an Overview on every search, and the number it shows changes from week to week. "14 of 15" and "7 of 11" read like a halving, but four fewer Overviews appeared, and a percentage on its own would hide that.

July to September: from 0 of 13 to 14 of 15

On 21 July 2026, Google showed an AI Overview on 13 of the 16 searches, and none of them cited the firm. By 4 September, Overviews showed on 15 searches and 14 of them cited it. That's the change we wanted to understand, so here's what happened on the site in between.

One caveat I'd want to know as a reader: the firm's new website went live on 16 July, five days before that first check, so Google had only just started recrawling it. Some of the July zero is a site Google hadn't caught up with yet, so I wouldn't credit the whole change to our work.

Here's what did change between the two checks:

• The sector pages that match these searches were rewritten or built out properly. Some went from a short stub of a couple of hundred words to a full page of well over a thousand.
• Each section on those pages opens with a short, direct answer to the question in its heading.
• FAQs were added where they fit, with FAQ markup that matches what's on the page.
• Internal links were tidied, so each sector page links to its parent and child pages.
• A batch of old blog articles that weren't doing anything was retired.

And here's what didn't happen in that window, which is worth knowing:

• No Google Business Profile work.
• No directory listings.
• No review campaign.
• No link building.

You can see more of our client work on the case studies page.

The pattern: cited pages were ranking pages

The clearest pattern in the September data is that citations followed rankings. Of the 22 Overviews that cited the firm across all 26 searches we checked that day, 21 cited a page that was also in the organic top 10 for the same search. Ranking wasn't enough on its own, but it came first.

That lines up with the larger studies. Ahrefs looked at 1.9 million citations from 1 million AI Overviews and found 76.1% of the cited pages ranked in the top 10. Our sample is tiny by comparison, but it points the same way.

On one search, though, the firm ranked first and the Overview still didn't cite it. So in our data, a top-10 page looks close to necessary but not sufficient.

What happened in October

On 1 October, during Google's September 2026 spam update, the check showed 7 of 11: fewer Overviews, and fewer of them citing the firm. After the update, on 18 October, an Overview cited the firm on 9 of the 10 searches where we got a usable result. So the October dip mostly didn't hold up, though fewer of our checks came back clean.

The update rolled out from 24 September to about 8 October. During an update, results reshuffle across the board, so my usual approach is to hold edits, avoid blaming any one page, and check again once the rollout ends. That's what we did here.

Two things came out of the re-check. First, about half of the 1 October results were clearly off: the top 10 we'd captured looked nothing like the usual results for those searches, while Search Console still showed the firm's pages near the top. On the 8 searches with a usable result, an Overview showed on 6 and all 6 cited the firm.

Second, on 18 October we ran each search three times, and 11 of the 16 gave at least one usable result. An Overview showed on 10 of those and cited the firm on 9, which is in line with September. We'll keep checking the same list each month.

AI Overviews change all the time anyway

AI Overviews change constantly, so a single check tells you very little. Ahrefs tracked 43,000 keywords and found that AI Overviews have a 70% chance of changing from one observation to the next, and that nearly half of the cited URLs are different each time. A swing on one or two searches is normal.

That's why we count across a fixed set of searches and repeat it, instead of reacting to one search on one day. A drop across most of the set, confirmed in Search Console, is worth acting on.

Where most AI Overview advice goes wrong

Most AI Overview advice I read makes one of two mistakes. The first is treating a single screenshot as proof, when the same search can give a different answer an hour later.

The second is chasing tricks, like special files or rewriting pages into tiny chunks, before the basics are in place: being easy to find in Google, and in Bing too, since some chat assistants lean on Bing's results.

In this client's case, the citations followed ordinary work on pages that answer a specific question plainly.

What Google says not to bother with

Google says optimising for AI Overviews is still SEO, and lists several popular tactics it considers myths. Its guide on optimising for its generative AI features, last updated in July 2026, says that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO". The myths it lists:

• llms.txt files and other "special" markup.
• "Chunking" content into small pieces for AI.
• Rewriting content just for AI systems.
• Chasing inauthentic mentions.
• Over-focusing on structured data.

It also warns that creating separate pages for every variation of a question, mainly to influence AI answers, can count as scaled content abuse. And it says "no third-party tool has access to our internal ranking or AI systems", which is worth remembering when a tool or agency promises AI citations.

None of that was in the work above. The pages were rewritten for people searching for a firm in their sector, and the citations followed the rankings.

What I'd do if you want to be cited

If you want to be cited, I'd start with the searches your buyers make and the page that should answer each one, then work on getting that page into the top 10. That's what this firm's citations followed. It's one firm in one sector, so treat these as steps to test, not rules:

1. List the searches your buyers actually make. Start with your Search Console queries and the questions you hear on sales calls, not a keyword tool's guesses.
2. Make sure one page clearly answers each one. If the honest answer to "which page should rank for this?" is "none, really", that's the first job.
3. Open each section with the answer. A sentence or two that answers the heading directly, then the detail.
4. Add FAQs only where they're real questions. Keep any FAQ markup matching what's visible on the page.
5. Link the pages together sensibly. Parent to child, related to related, so each page has a clear place.
6. Work on ranking first. In our data, citations came almost entirely from pages already in the top 10.

If you're weighing whether to do this yourself or pay someone, our guide to GEO agencies vs doing it yourself covers the costs and what you can handle in-house.

How to track your own citations

You can track your own AI Overview citations with a fixed list of searches, a spreadsheet and a few habits that stop one odd result from misleading you. You don't need an expensive tool. What matters is checking the same searches the same way each time, and counting Overviews shown and Overviews citing you separately.

1. Write down 15 to 30 searches and freeze the list, so every check covers the same searches.
2. Check them on the same settings each time: same country, same device, ideally logged out.
3. Record two numbers per check: how many searches showed an AI Overview, and how many of those cited you.
4. Repeat each search more than once if you're using an automated tool, and note when the results disagree.
5. Set aside results that look scrambled, where most of the usual top-10 sites have disappeared at once.
6. Cross-check with Search Console. Its Generative AI report shows impressions from AI features by page, though not clicks or queries yet.

Step 5 is there because automated tools sometimes collect a top 10 that looks nothing like the real one. Several SEO data companies reported this happening more often from mid-September 2026. A simple test is to note the sites that usually rank for each search, and keep a result only if at least half of them are still there.

Free tracker: we've turned this into a spreadsheet you can copy: download the AI Overview citation tracker. It works in Excel, Google Sheets and Numbers, marks each result "Keep" or "Set aside" for you, and adds up Overviews shown and cited for each check date.

If an agency reports AI citations to you, it's fair to ask how they handle the same problems. Our guide on checking whether your SEO agency is working covers what you should be able to see each month.

For this client, Search Console's Generative AI report has kept rising through all of this: about 590 AI-feature impressions a day from mid-July to mid-September, about 920 a day from mid-September to early October, and about 1,210 a day in the week after. Part of that is probably Google showing AI answers on more searches, so I wouldn't read too much into it.

Is being cited worth it?

Being cited seems to soften the click loss from an AI Overview rather than add traffic. Seer Interactive looked at 3,119 informational searches across 42 organisations and found an organic click-through rate of 0.70% when a brand was cited in the AI Overview, against 0.52% when an Overview showed and didn't cite it.

Where there was no Overview at all, the rate was 1.45%. Seer is clear that it can't prove the citation causes the difference.

For a business that sells something considered, like executive search, I think the bigger value is being in the answer when a buyer is building a shortlist, even if they don't click. If you'd like to see whether AI tools mention you at the moment, our free AI visibility checker asks ChatGPT, Perplexity and Google's AI the questions your customers ask.

What we haven't figured out yet

There's a lot this one example can't tell you, and I'd rather say so than overstate it. It's one firm in one sector, the checks are snapshots, and AI Overviews and chat assistants behave differently. Here's what we haven't worked out yet:

• This is one firm in one sector. It's a real, tracked example, not a study.
• AI Overviews and chat assistants are different games. On a wider set of 26 questions across ChatGPT, Gemini, Perplexity and Claude, the assistants named the firm in 60 of 104 answers in September and 55 of 104 in October. That's a small change, and we haven't checked those answers as carefully yet.
• Our own site isn't cited yet. When we checked in October, no AI Overview cited Nest Content for questions about AI search. We're working on it the same way.

If you'd like a second pair of eyes on your own AI search tracking, book a 20-minute call.

Want to see where you stand first? Run the free check.