An AI overview is a written summary that appears above the ordinary search results for some queries, assembled from several web pages and usually naming a few of them. It changes what a result page looks like rather than how pages are read.
Short answer
An AI overview is a machine-written summary placed above the ordinary search results, built from several sources and linking to some of them. There is no way to opt into appearing in one. The pages that get drawn on tend to be clearly structured, specific and directly relevant to the question being asked.
The summary sits at the top of the result page, runs to a few sentences or a short list, and carries links to some of the pages it drew on. Below it, the ordinary results continue as before.
They do not appear for every query. Broad informational questions attract them. Navigational queries, where somebody is looking for a specific site, and highly transactional queries tend not to.
Coverage also changes. A query that produced a summary last month may not this month, and the reverse happens too. Treat the presence of one as a condition of the query rather than a permanent feature.
The pattern matches what our page on how an answer engine chooses sources describes. Pages that address the exact question, state things plainly, and come from a site that is otherwise trusted are the ones that get drawn on.
Being in the ordinary results first appears to matter, so the conventional work of being findable is not bypassed. A page nothing can find is not a candidate for anything, summary or otherwise.
Specificity does the rest. A page built around one question, opening with a direct answer, gives a summarizer something clean to use. A page covering six subjects gives it a choice it has no reason to make in your favor.
You cannot submit a page, request inclusion, or pay for a place. There is no console setting and no report telling you which summaries used your material. Anyone offering that access is describing something that does not exist.
You can prevent your content being used in these summaries by blocking indexing entirely, which also removes you from the ordinary results. For almost every business that trade is plainly not worth making.
And you cannot correct an answer directly. If a summary repeats something outdated about your business, the remedy is to fix the sources it drew on, starting with your own pages and any stale directory listings.
The work below is the same work that improves the page for readers, which is what makes it safe to do without knowing exactly how any summary is assembled. None of it is speculative and none of it is wasted if the format changes.
For questions that a summary fully answers, some visits will not happen. That effect is real and it is concentrated on short informational queries, which were rarely the ones producing customers anyway.
For anything that ends in choosing a business, the summary is a step rather than a destination. Somebody still has to pick, check and make contact, and a named source is well placed for all three.
Watch your own reporting rather than industry commentary. Our page on the performance report covers how to see where impressions and clicks are actually moving on your site.
Writing pages addressed to the summarizer. Content that instructs a machine, repeats a question fifty times, or stuffs an answer with keywords reads badly to people and is exactly what quality systems look for.
Publishing hundreds of near-identical question pages to cover every phrasing. That is the doorway pattern, covered on our doorway pages page, and it damages a site rather than expanding it.
And chasing a summary that appeared once. Coverage shifts constantly, so building a strategy around a single observed result is building on something that may be gone before the work ships.
Track impressions and clicks over months rather than days, and compare pages against themselves rather than against a benchmark. A page whose impressions hold while clicks soften is probably sitting under a summary.
Sample the questions directly as well. Ask them the way a customer would, note what comes back, and look at what the named sources had that your page lacked. That comparison is more useful than any aggregate number.
Then fix the page rather than the metric. Every gap that comparison reveals is a real gap in the page, and closing it helps whether or not a summary is involved on any given day.
Only by blocking indexing of the page, which also removes it from the ordinary results. For nearly every business that trade is not worth making. There is no setting that keeps you in search while excluding you from summaries.
They usually link to some of the pages drawn on, but not to every one and not always prominently. Being used and being named are separate outcomes, and you have no control over which one you get.
For short factual queries, some visits will not happen. For queries that end in hiring or buying, the effect is much smaller because a person still has to choose a business. Watch your own reporting rather than general commentary.
Nothing beyond good practice. A direct answer near the top, one subject per page, honest sourced claims and clean structure. Content written specifically to influence a summary tends to read badly and be treated accordingly.
15-day free trial. Card required. Cancel before day 15 and you pay nothing.