Guide

How to rank in AI Overviews

You do not rank in one. You get cited by one, or named inside one, and that difference decides which of the usual advice is worth your time.

What an Overview actually is

Google runs the query, retrieves a set of pages, and writes a summary from them with citations attached. There is no ranked list inside it and no position to hold. There is a set of pages it drew on, and either you are among them or you are mentioned by one of them, or you are absent.

Which means conventional ranking is the entry condition rather than the goal. A page that does not surface for the query cannot be retrieved, and a page that is never retrieved cannot be cited, however well marked up it is.

What we can say from our own measurement, and what we cannot

Being straight about the limits: on 26 July 2026 we measured 3 software categories against ChatGPT, Gemini and Perplexity, reading all 131 sources behind 18 answers. We did not measure Google AI Overviews specifically, so anything below that comes from that data is about assistant answers generally, and the read-across is reasonable rather than proven.

With that said, two findings travel well because they are about how retrieval-based answers behave rather than about one engine. Every category had a small number of domains deciding it, usually fewer than five. And those domains were different in every category, so a universal checklist was worth less than ten minutes of looking at your own.

What actually moves it

1. 1

   Rank for the question, not just the keyword

   Overviews fire on questions. A page that answers the question directly, near the top, in plain language, is a far better retrieval candidate than one that circles it for four hundred words before getting there.

2. 2

   Be present in the pages that already get cited

   You do not have to be the cited page. Being named inside it works too, and it is often much more achievable than outranking it. Read the citations on the Overviews in your category and you will usually find the same few sites.

3. 3

   Answer the follow-ups on the same page

   The related questions underneath an Overview are the model's own map of what else is being asked. Covering them properly on one page gives you more surfaces to be retrieved for.

4. 4

   Keep it verifiably true

   Retrieval-based answers lean on pages that agree with other pages. A page making a claim nothing else supports is a page a summariser has reason to skip, which is an underrated argument for publishing your actual numbers.

What does less than people hope

Schema markup helps a system understand a page it has already fetched. It does not decide whether the page gets fetched, and no amount of it puts you inside somebody else’s article. Worth doing, badly oversold.

Blocking Overviews is the other one. You can use nosnippet or max-snippet to keep Google from using your content, and it also strips your ordinary snippet, so you lose the citation and gain nothing. The traffic does not come back to you, it goes to whoever did not block.

How to tell whether any of it worked

Pick the questions your buyers actually ask, freeze the list, and check them on a schedule. Overviews are volatile for the same query, so one look tells you almost nothing and two looks a month apart tell you a direction. That is the entire discipline, and it is the part almost nobody does, which is why so much advice in this area is vibes.

Questions people ask next

How do you rank in Google AI Overviews?

You do not rank in one, you get cited by one. An AI Overview is assembled from pages Google retrieved for that query, so the job is being one of those pages, or being named inside one of them. Conventional ranking is the entry condition, because a page that does not surface for the query is not a candidate to be cited.

Can I opt out of or block AI Overviews?

You can stop Google using your content in them with the nosnippet or max-snippet directives, and doing so also removes your snippet from ordinary results. Almost nobody should. Blocking removes you from the answer without sending the traffic anywhere else, so you lose the citation and keep the loss.

Does schema markup get me into AI Overviews?

It helps a system understand a page it has already retrieved and it does not decide whether the page gets retrieved. Structured data is worth doing and it is not the lever people hope it is. The lever is being genuinely relevant to the question and being on pages the retrieval already trusts.

How long does it take to appear?

Nobody can tell you, and a vendor quoting a number invented it. Overviews change constantly for the same query, so a single check is a snapshot rather than a position. What you can do is measure the same set of questions on a schedule and watch whether your presence in the answers changes.

Is optimising for AI Overviews different from optimising for ChatGPT?

Less than the separate vocabulary suggests. Both assemble an answer from retrieved pages, so both reward being present in the pages that get retrieved. What differs is which pages each one reaches for. In the 3 categories we measured across ChatGPT, Gemini and Perplexity, the deciding sources were different in every category, so the safe assumption is that they differ per engine too and you should check rather than assume.

Our category measurements were taken 26 July 2026 against ChatGPT, Gemini and Perplexity, not against Google AI Overviews. Where this page reads across from one to the other it says so.