Google AI Overviews: What Actually Happens to Your Traffic
Plenty of site owners have noticed a dip in clicks without any obvious reason. Rankings look fine, impressions are holding, but the traffic just isn't arriving the way it used to. In many cases, Google's AI Overviews are sitting between the search result and the click. Understanding what that actually means, and which sites feel it most, is the clearest way to work out whether you need to change anything.
On this page
What an AI Overview Is
Google’s AI Overviews are generated summaries that appear at the very top of certain search results pages, above the traditional blue links, above paid ads, and above featured snippets.
Google builds them using its Gemini model, which reads and synthesises content from pages it has already crawled and indexed. It is not fetching live web pages at the moment you search. It draws on what it has already stored, weights sources it considers reliable for that query, and writes a condensed answer in its own words. The result sits inside a distinct grey box with a “Generated by AI” label and, usually, a small cluster of source attribution links tucked below the summary text. Those attribution links do show as clickable references, though they tend to appear smaller and less prominent than a standard organic result.
The overview itself does not quote pages verbatim. It paraphrases and compresses, meaning a site can contribute to an overview without a single sentence from its content appearing word for word.
They do not appear on every query. Google reserves them mainly for informational searches, questions that benefit from a synthesised answer rather than a direct visit to a specific page. Commercial searches, navigational queries, and anything where the user clearly needs to click through to complete a task tend to show standard results instead.
Traffic Patterns to Watch
Where clicks disappear
The split broadly falls along intent lines. Informational queries, the kind where someone wants a quick definition, a how-to summary, or a straightforward fact, are where Overviews hit hardest. Google surfaces a synthesised answer at the top of the page, the user reads it and leaves, and your page never gets the visit even if you ranked first. Think of searches like “what is crawl budget” or “how does noindex work”. The answer fits neatly into a few sentences, the Overview delivers it, and there is nothing left to pull the user further down the page. For sites built on high-volume, low-depth informational content, that can mean a noticeable drop in traffic without any change in rankings.
Where Overviews send traffic your way
The opposite happens with queries that carry commercial, comparative or nuanced intent. When someone searches for something like “best WordPress hosting for multiple sites”, an Overview cannot settle the question on its own. It has to cite sources. Being one of those cited sources drives clicks that would not have arrived through a standard ten-blue-links result, because the Overview itself points directly to your page as a reference. Understanding how AI search tools handle citations differently helps you judge which type of content to prioritise.
The practical question is which camp your content sits in. If it answers a closed question briefly, it is vulnerable. If it builds a case, compares options, or requires depth to be credible, it is a citation candidate.
Absorbed vs. Cited Content
The distinction matters more than most people realise. Definitions, factual summaries, and short explanatory passages tend to get absorbed straight into the Overview answer with no attribution at all. Google pulls the clearest version of a plain fact, paraphrases it, and the source disappears. If your page exists to define what a term means or summarise a well-known process in a few sentences, that content feeds the answer without sending a single visitor your way.
Step-by-step how-to content sits in a grey area. A tightly structured guide with numbered steps can get partially absorbed, but if the process is involved enough, Google will sometimes cite the source because the full sequence adds value beyond what fits in a two-line summary.
Product comparisons and opinion pieces behave differently. A comparison page with real specifics, genuine trade-offs, and a clear recommendation tends to earn a citation because Google cannot responsibly flatten that into a neutral overview. The same goes for first-person experience, strong editorial positions, and content that contradicts the obvious consensus.
The pattern that emerges is that how AI Overviews affect your traffic depends almost entirely on the shape of your content, not its quality alone. Generic, evergreen explainers are the most at risk of being consumed without credit. Content with a clear point of view, real comparative detail, or a structured process deep enough to warrant its own page is far more likely to hold its ground as a named source.
Reading Your Search Console Data Honestly
Impressions vs. clicks
The most useful thing you can do is pull your impressions alongside clicks over the same time window, then look at them separately. If impressions stay flat or grow while clicks fall, that pattern points squarely at something absorbing attention above your result, and AI Overviews are the obvious candidate for informational queries. If both impressions and clicks drop together, that is a ranking shift, not an Overview problem.
Conflating the two is the most common mistake people make when they panic about a traffic dip. A site that drops from position three to position eight will lose clicks for an entirely different reason, and no amount of rewriting your content for AI will fix a link profile or a page authority issue.
Breaking down by query type
The query-type breakdown matters a lot here. Filter your Search Console queries by the ones triggering position-zero appearances, particularly question-format queries beginning with “how”, “what” or “why”. Those are the searches most likely to get an Overview generated above all organic results. Compare their click-through rate against transactional queries on the same site. You will almost always find the informational ones have suffered more. Understanding what AI Overviews actually do to individual query types helps you prioritise which pages to address first.
Keep the two signals separate. Impressions versus clicks tells you about visibility absorption. Position movement tells you about ranking health. Mixing them produces the wrong diagnosis.
How Different Site Types Are Affected
Not every site takes the same hit. Where you land in that picture depends heavily on what your site does.
Informational content sites feel the sharpest squeeze. If your traffic has always come from people asking questions, such as “how does X work” or “what is the difference between Y and Z”, Google’s AI Overview often answers that question before anyone clicks. The user gets a synthesised reply pulled from multiple sources, reads it, and moves on. Your page may well be cited in the sources panel, but citations don’t pay the bills if click-through rates drop to near zero. Publishers built on ad revenue from high-volume how-to content are noticing this the most. It’s also worth understanding how Bing’s AI search is reshaping the same pattern, because both search engines are pulling in the same direction.
Service-based businesses sit in a noticeably different position. Someone searching “emergency plumber near me” or “web designer for small business” is not looking for a synthesised answer, they want a person. Overviews rarely satisfy that intent, so the click still happens.
Ecommerce sits somewhere between the two. Navigational queries like “buy blue running shoes size 9” tend to land on product listings rather than an Overview, but category-level research queries can get absorbed. If your pages exist to inform rather than to close, they are more exposed. If they exist to convert a buyer who already knows what they want, the risk is lower.
Adjusting Your Content Without Chasing the Algorithm
The sites that hold up well in an AI-heavy SERP tend to share one characteristic. They were already written for the reader, not for a ranking formula. Clear headings, direct answers near the top of each section, concrete examples that illustrate a point rather than pad a word count. None of that is new advice.
What changes is that thin, hedged, vaguely-worded content has even less room to hide. If a page dances around an answer for four paragraphs before getting to it, Google’s AI has no trouble extracting the eventual point and presenting it without ever sending the user your way. Structure has always mattered. It just has more immediate consequences now.
One practical adjustment is checking whether your key answers appear close to the relevant heading. A buried answer, even a good one, is harder for any system to surface cleanly. If you want to understand how Google is evaluating and scoring your pages beyond content alone, that broader picture matters too.
None of this requires a complete rethink. It requires honest editing.