Google AI Overviews is an AI answer that appears at the top of the results page, assembling a response from several sources and linking to them. Getting a site into those answers is not a separate trick but an extension of good search engine optimisation: Google draws the answers from its index, and what gets drawn on most is content that answers the question directly and in a structured way. Here is how AI Overviews works and what needs to change on your site.
What Google AI Overviews is and when it appears
AI Overviews is Google’s generative answer, shown in some searches above the traditional results. It condenses the answer into a few paragraphs and displays the source pages the information was gathered from. The answer appears most often in question-shaped, information-seeking searches — ”how”, ”what does it cost”, ”what is the difference” — and less often in clear transactional searches where the user is already heading to a particular site.
For a company this has two consequences. Some clicks do not arrive, because the user gets the answer straight from the results page. At the same time, becoming a source is a new place to be visible: the source links in an AI answer collect clicks specifically from people who want to go deeper, which means better quality visitors.

How to become a source
Google has not published an exact recipe, but the pages that end up as sources share the same features. These are observations from the field rather than Google’s official criteria, and they are in line with what Google has said about search quality generally.
The page answers the question immediately. An AI answer is built from paragraphs that work as standalone answers. If your pricing article only gives the price in the tenth paragraph, a competitor’s page that gives it in the first is better raw material.
The structure is machine-readable. A clear heading hierarchy, question-shaped subheadings, and lists and tables where they genuinely organise the content. FAQ sections and the associated FAQPage schema make the question and answer pairs visible to machines too.
The site is trustworthy in Google’s eyes. The same E-E-A-T assessment that drives organic rankings drives source selection: expert content, an identifiable publisher, references from elsewhere.
The basics are in order. The page is indexed, fast and works on mobile. AI Overviews does not pick a source Google does not otherwise value; ranking in the top ten for a question search is in practice a precondition.
Find out in half a minute whether the AI knows your company
The AI visibility test asks a language model what it knows about you and shows the answer exactly as it comes.
What this means for content production
The most concrete change is moving the answers to the front of the content. Every important question gets its own section whose first one to three sentences form a complete answer: the definition, the price, the timescale, the conclusion of the comparison. The deeper text follows after that.
The second change is mapping the questions. AI answers trigger on question searches, so a content plan is best built on a list of the questions customers actually ask, including the long, specific ones nobody used to search for but which people now put to an AI in full sentences.
The third is restraint with the metrics. When some answers are given on the results page, the ratio of impressions to clicks changes. In Search Console it is worth watching particularly which questions your site already ranks at the top for, because those are the most likely AI Overviews slots.
Is AI Overviews visibility worth pursuing separately
Not as a separate project. Everything described above is the same work that improves organic rankings and visibility in other AI searches too, because Perplexity and ChatGPT reward exactly the same kind of directly answering content. A sensible way to think about it: build your content into an answering machine, and AI Overviews is one channel that picks the fruit.

What triggers an answer box
Answer boxes do not appear for every search, and the pattern is consistent enough to predict.
They appear most for questions with a stable factual answer: what something means, how something works, what the difference between two things is. They appear less for searches where a person clearly wants to choose between suppliers, and least for navigational searches where someone is heading to a known site.
They also appear more for longer, question-shaped queries than for two-word ones. This matters for planning: the pages most exposed are the explanatory ones, and the pages least exposed are the commercial ones. For most companies that means the traffic at risk is the traffic that converted least anyway.
Where the sources come from
An answer box is assembled from several pages rather than one, and the sources are drawn from search results for the query and related queries.
Two observations follow from watching this closely. Sources are not always the top-ranked pages: a result in fifth place that answers the specific sub-question directly is often preferred to the first result covering the topic broadly. And the box frequently cites several sources for a single answer, each supplying one part of it, which means being the source for one paragraph is a realistic target even against much larger competitors.
The practical reading: you are not competing for a position. You are competing to be the clearest available statement of one particular fact.
Reading the effect in Search Console
Clicks from an answer box are counted as ordinary organic clicks, so the effect has to be inferred. Three patterns indicate it.
Impressions holding steady or rising while clicks fall. The page is shown as often as before, but the answer is being read in the results.
Average position improving while click-through rate declines. This combination rarely has another explanation.
The change concentrating in question-shaped queries. Filter your queries for ones containing how, why, what and whether, and compare their click-through rate against the rest. A widening gap is the clearest signal available.
None of this is precise measurement. It is enough to stop a team from concluding that its content quality collapsed in a month when the results page simply changed shape.
What to do about a page that lost clicks
The instinct is to make the page longer. That usually makes it worse. Three responses work better.
Move the page down the funnel. If the definition is now answered in the results, add the part that cannot be: your own figures, a worked example, the situations where the general answer fails. A page that offers something beyond the summary keeps its reason to be clicked.
Accept the visibility and take the credit. Being the cited source builds recognition even without a click, and it shows up later as brand searches. Measure that rather than mourning the click.
Consolidate. Several thin pages each losing their answer to a box are better combined into one page with enough depth to survive.
Whether you can opt out
The available controls are blunt and worth understanding before anyone suggests using them.
Snippet controls can limit how much of a page may be shown in results, and setting them aggressively does reduce what a summary can quote. It also reduces your ordinary search snippet, which is usually a worse trade than the problem it solves.
The nosnippet directive removes the page from summaries entirely and removes its search snippet too, leaving a bare title. Almost no commercial site benefits from this.
The honest position is that there is no setting which keeps full search visibility while excluding the page from answer summaries. Publishers with a paid content model sometimes accept the trade. For a company that wants to be found, the trade does not make sense.
Planning for the next two years
Two assumptions are safe enough to plan around, and one is not.
Safe: the share of searches showing a summary will grow, and informational traffic will keep declining. Also safe: pages tied to a decision will hold their traffic better than pages that explain a concept.
Not safe: any specific forecast of how much traffic will be lost. The numbers circulating vary by an order of magnitude, they come from different measurement methods, and most of them are produced by parties with something to sell.
Plan for the direction rather than the magnitude. That means shifting the content mix toward material that supports a decision, and starting to measure brand searches now so there is a baseline when the informational traffic goes.
Frequently asked questions
What is Google AI Overviews?
An AI-assembled answer shown above the search results, summarising information from several sources and linking to them. It appears particularly in question-shaped searches.
Can AI Overviews answers be influenced?
Yes, by the same means as organic rankings: directly answering content, a clear structure, sound technical foundations and a trustworthy site. There is no separate ”AI Overviews optimisation” apart from good SEO.
Does AI Overviews reduce my site’s traffic?
In some searches yes, because the user gets the answer on the results page and does not click. In return, a site that becomes a source gains visibility and better quality visitors. The net effect depends on whether you are a source in those answers or not.
Does AI Overviews appear in Finnish-language searches in Finland?
It does, and it is expanding. Coverage varies by search type and changes with Google’s updates, so it is not worth fixing a definitive list of triggering searches. Check the most important question searches in your own field regularly instead.
Can a page be excluded from AI Overviews while keeping normal search visibility?
Not with the controls that exist. Snippet limits and the nosnippet directive reduce what a summary can quote, but they cut your ordinary search snippet by the same amount, which usually costs more than it saves.
Why do answer boxes cite a page ranking fifth rather than the first result?
Sources are chosen for answering the specific sub-question clearly rather than for overall ranking. A focused page often supplies one paragraph of the answer while a broader top-ranked page supplies none, which is what makes the box winnable for smaller sites.
If you want to know whether AI searches already pick up your company, the AI visibility test shows it immediately. We cover building content into an answering machine on the GEO service page and the SEO service page.
Related articles
Bing visibility matters again — AI search draws on its index
Bing visibility became relevant again after years of quiet, because several AI services rely on Bing’s index for their web searches. A site that…
Local visibility in AI search — four actions that cover most of the benefit
Local visibility in AI search is decided largely by the same ingredients as Google Maps visibility: the Google Business Profile, reviews and consistent basic…
GEO for B2B — the AI writes the buyer shortlist, are you on it
For a B2B company GEO is an unusually strong lever, because AI search hits exactly the stage of the buying journey where B2B deals…
Shall we take the first step together?
Tell us briefly where you stand. We reply the same working day, usually within a couple of hours.
- A free 30-minute call, no strings attached
- You get a concrete view of where you stand
- We say plainly if we are not the right fit
- Phone+358 50 469 0039
- Emailmisa@purodigital.fi