google ai overviews · definition
Google AI Overviews: the answer that sits above the links
Google writes an answer above the results and cites a few sources under it. What the block is, why impressions can hold while clicks fall, and what makes a page one of the ones it quotes.
Google AI Overviews are AI-generated answers shown above the standard results for some queries, assembled from passages on several web pages and displayed with links to the sources they were drawn from.
reviewed 2026-09-02 · by the IT Master editorial team · how we check facts
What an AI Overview is in practice
An AI Overview is a block at the top of a Google results page: a few sentences or a short list, written by a model, with links to a handful of sources set beside or beneath it. The ordinary results still appear, further down the screen.
Three properties matter more than the layout.
- It is built from passages, not whole pages. Sentences are drawn from several sources, so a page can be cited for one paragraph while the rest of it is ignored.
- It does not fire on every query. Whether a block appears at all, and how long it runs, varies by query, by device and over time.
- The cited set is unstable. The same question asked twice can surface different sources, so a single screenshot proves very little.
The sources come from the same crawled, indexed web as the blue links. A page Google has not crawled and indexed cannot be quoted, and main content that only appears once client-side JavaScript has run tends to be crawled later and less reliably. The entry ticket is the one ordinary search has always asked for.
Google AI Overviews are AI-generated answers shown above the standard results for some queries, assembled from passages on several web pages and displayed with links to the sources they were drawn from.
Why it matters in 2026
The block sits between the searcher and your page. When a question has a short factual answer, that answer is now delivered in the results, and the click that used to follow it often does not happen. When someone needs to compare, price or buy, the click usually survives, because a paragraph cannot finish the job.
That split is what makes traffic reports confusing.
- Impressions can hold while clicks fall. The page still ranks; fewer people need to open it. Read click-through rate query by query, or you will diagnose a ranking problem you do not have.
- Being cited is not the same as ranking first. The sources named in a block are often not the top organic result, so being quotable and being first are two different jobs.
The practical response is to know which of your queries carry an Overview and which do not, then judge each page against the behaviour of its own query rather than against a site-wide sessions chart. Search Console is still the instrument; it just has to be read per query.
How we treat AI Overviews
There is no AI Overview switch, here or anywhere else. What the engine does is make a page a better candidate for being quoted: topics taken from real search demand, JSON-LD and FAQ schema on every article, internal links into the rest of the cluster, an IndexNow ping to the engines that accept one as soon as the article publishes, and a nightly loop that queues a refresh when a page's Search Console numbers decay. Facts are checked by models from a different vendor than the one that wrote them, because a confident wrong sentence is exactly what you do not want lifted into an answer.
Two honest limits. Search Console does not report Overviews as their own surface, so nobody, us included, can hand you a clean count of appearances inside the block. And our nightly measurement probes GPT and Perplexity, which are assistants rather than Google's results page. Treat that as a useful proxy for whether your pages are quotable, not as a readout of Overview citations. We report the two separately instead of blending them into one flattering number.
Common misunderstandings
That AI Overviews ended SEO. The sources in the block are crawled, indexed pages. The prerequisites did not change; the payoff for thin, restated pages did.
That some markup buys you in. Structured data helps a machine parse a page and is worth having, but no schema, meta tag or file grants entry to the block, and llms.txt does not either.
That blocking AI crawlers protects you. Overviews are served from Google's own search index, so Googlebot's rules are the ones that apply. Google-Extended, the control people usually reach for, governs Gemini's training and grounding rather than what Search itself shows. Check what you are actually blocking with the free AI crawler check.
That one appearance is a trend. Results vary between runs and between people. Repeat the check on a fixed query set over several weeks before concluding anything.
That publishing more raises your odds. Volume without a reason to be quoted adds pages, not citations. A page gets lifted when it answers one question cleanly and carries something the other results lack, which is usually first-party data.
questions people ask
How do you get cited in a Google AI Overview?
There is no submission form and no guarantee. What you can control is candidacy. Be crawlable and indexed, serve your main content in the HTML rather than leaving it to client-side rendering, and answer the question the page targets in one plain sentence near the top rather than after four paragraphs of preamble. Keep facts specific and checkable, since vague claims are hard to quote safely. Then give the page something the other results do not have, such as your own product data or support answers. Being quotable is the work; the appearance is the side effect.
Are AI Overviews the same thing as SGE?
Effectively yes. SGE stood for Search Generative Experience, the name Google used while the feature was an opt-in experiment. When it rolled out to ordinary results it was renamed AI Overviews, and that is the name Google uses now. People in the industry still say SGE out of habit, and older articles and tools use the term as well, so treat the two as the same feature at different stages rather than as two separate products competing for the same slot.
Do AI Overviews reduce clicks?
It depends on the query, so measure rather than assume. Questions with a short factual answer tend to lose clicks, because the answer is delivered in the block and there is little left to open. Commercial and comparison queries hold up better, since a paragraph cannot show stock, price or specification. The pattern to look for in Search Console is steady impressions with a falling click-through rate on the affected queries, not a fall in position. If position is stable and clicks are not, the block is a likely explanation.
Can I stop my pages appearing in AI Overviews?
Only bluntly. The nosnippet and data-nosnippet directives, and a max-snippet limit, restrict how much of your page Google may show, and that restriction covers Overviews as well. There is no Overview-only opt-out, so the same tags also shrink or remove your ordinary search snippet, which usually costs more traffic than it saves. Blocking Googlebot removes you from the block and from search together. For most sites the sensible move is to stay eligible and compete for the citation instead.
related
- Answer engine optimization: the practitioner's guide to AEO
- Generative engine optimization: what AI search actually rewards
- AI visibility: being named in the answer, not just the results
- Citation rate: the share of answers that credit you
- What FAQPage schema marks up, and what Google changed
- AI Visibility Checker
See what search engines and AI assistants find on your site
Free, no account. Type your address and we show you what is missing and what we would write first.