Google Keeps Expanding AI Overviews: Here's What It Means for Organic Traffic

Google's AI Overviews stopped being an edge case sometime in early 2026. By the time Google held its May I/O keynote, the company said AI Overviews had passed 2.5 billion monthly users and AI Mode, the fuller conversational version, had passed 1 billion. At that same event Google merged the two experiences: read a summary, ask a follow-up question, and you get carried straight into an AI Mode conversation without a page reload, on desktop and mobile, worldwide (Google blog). That single change tells you where this is headed. AI Overviews aren't a bolt-on anymore. They're the front door.

If you publish content for a living, the open question by October 2026 isn't whether this affects your traffic. It's which of your pages are exposed, how much they're losing, and what you can still do about it that isn't wishful thinking.

What Google actually shipped this year

Four moves matter more than the rest, and the dates are worth knowing because they show the direction of travel, not just a single launch.

In February, Google Search SVP Nick Fox announced AI Mode was expanding into 53 new languages in one go, the largest single language rollout since the feature launched. That took AI Mode to just under 100 languages, covering roughly another billion people who hadn't had access before.

Then came the May merger at I/O: AI Overviews and AI Mode became one continuous experience, with Gemini 3.5 Flash set as the new default model behind both. Google also said it was expanding Personal Intelligence inside AI Mode, the feature that lets responses draw on your own Gmail, Drive, and other connected Google data when relevant, to nearly 200 countries and 98 languages with no subscription required (Google I/O 2026 recap).

In August, Google extended generative UI, the interactive calculators and on-the-fly visualizations that had been exclusive to AI Mode since a November 2025 launch, into AI Overviews themselves. Ask about mortgage payments and you might get a working calculator embedded directly in the overview instead of a static paragraph pointing you somewhere else. Google had said at I/O this capability would roll out to everyone in Search for free over the summer; the AI Overviews rollout specifically began August 19, with full availability expected within weeks (Search Engine Journal).

A little over a week later, on August 28, Google confirmed AI Overviews can now auto-expand into a longer, AI Mode-style answer on their own, without a user tapping "Show more." For some query types, the overview simply decides it needs more room and takes it.

None of these four changes are reversals. They all point the same direction: fewer boundaries between a quick AI summary and a full AI-driven search session, and more surface area for Google's generative layer to occupy before a user ever reaches a blue link.

What the traffic data actually shows

This is the part most "AI Overviews are killing SEO" posts get sloppy on, so it's worth being specific about what's been measured versus what's just asserted.

Seer Interactive, an SEO agency, ran one of the more rigorous looks at this: 53 brands, 5.47 million distinct queries, and 2.43 billion organic impressions tracked from January 2025 through February 2026. Their numbers show organic CTR sitting around 2.36% on queries with an AI Overview present in February 2026, against 3.82% on queries without one. That gap had actually been worse in December 2025, when CTR on AIO queries bottomed out at 1.31% before partially recovering into the new year (Seer Interactive). The recovery matters. This isn't a one-way ratchet down. Google appears to be actively tuning how much real estate AI Overviews take on a given query, and publisher CTR moves with it, not against it.

Seer's data also isolated something more specific: pages actually cited inside an AI Overview earned roughly 120% more clicks per impression than uncited competitors on the same query, even though cited pages still trailed the CTR of a non-AIO search by about 38% overall. Translation: being invisible in the overview is worse than the overview shrinking the whole market. Being cited in it is the next best thing to the overview not existing at all for that query.

Pew Research ran a separate study with a different methodology: a panel of 900 US adults, nearly 69,000 real searches tracked through actual browsing behavior rather than self-report. Their headline finding was that people clicked a traditional search result on 8% of visits to a page carrying an AI summary, against 15% of visits where no summary appeared. Clicks on the citation links sitting inside the AI summary itself happened on just 1% of visits, which means the small footnote links Google added partly to answer publisher complaints are barely used in practice. Sessions also ended with zero clicks more often when a summary was present: 26% of the time, against 16% when it wasn't (reported by Search Engine Land).

Two different methodologies, two different exact numbers, same direction: an AI Overview roughly halves your odds of getting a click on that specific query, and being cited inside it is meaningfully better than not being cited, while still landing well short of what that query would have delivered with no overview at all.

Which content is actually exposed

The pattern holds across both studies, and it matches what Google has said publicly about what AI Overviews are built to handle: queries with a clean, factual, extractable answer.

Pure reference and definition content, the "what is X," "how many Y," simple unit-conversion pages, takes the sharpest hit. If the entire value of a page is a single fact, an AI Overview that extracts that fact does the job completely, and there's no remaining reason to click through to confirm it.

Comparison and "best X" roundups land in the middle. AI Overviews will summarize the field, but someone actually choosing between options still tends to want current pricing, a real review, or a specific detail an AI summary is often deliberately vague about. These pages lose some clicks, but not at reference-content rates.

Long guides and multi-step tutorials hold up best of all. An overview can summarize five steps in a sentence each, but someone actually doing the thing, especially once step three doesn't match what they expected, still needs the full page with the screenshots, the edge cases, and whatever troubleshooting section sits underneath.

News and fast-moving topics see comparatively less AI Overview coverage in general. Google has consistently treated freshness and source diversity differently for time-sensitive queries, which is part of why timely content hasn't taken anywhere near the hit that evergreen reference pages have.

What Google actually tells site owners to do

Here's the part worth sitting with. Google's own Search Central documentation on AI features in Search states plainly that there are no special requirements to get picked up as a supporting link in AI Overviews or AI Mode. "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet," and beyond that, "there are no additional requirements... nor other special optimizations necessary." The same page is explicit that no new file format is needed either: "You don't need to create new machine readable files, AI text files, or markup to appear in these features" (Google Search Central).

That's a useful corrective against a growing pile of checklists selling specific markup or file formats as the fix for AI visibility. If your page already ranks and already shows a snippet in regular search, it's already eligible for AI Overviews and AI Mode. The lever that actually moves whether you get cited is the same one that's always mattered: how clearly and completely your page answers the exact question being asked, in a form Google's systems can extract cleanly. Our technical SEO checklist covers the indexing and crawlability side of that equation. Shaping the content itself into something extractable is a separate, more editorial skill, and we go deeper on it in how to get cited by ChatGPT and Google AI Overviews.

Site owners do have blunt controls if they'd rather opt a page out entirely. A nosnippet or data-nosnippet tag stops Google from pulling a snippet, which also stops it from citing that content in an AI summary. max-snippet caps how much text Google can pull. noindex removes the page from consideration altogether. None of these get you more traffic. They're defensive tools, not offensive ones, and worth using sparingly.

How to check whether this is actually happening to you

Before changing anything, look at your own Search Console data rather than assuming the industry averages apply to your site. Pull the Performance report, filter to your highest-impression queries, and sort by CTR. Queries sitting well below your site's average CTR for their position are worth a manual check: search that exact phrase and see whether an AI Overview appears above your listing.

Do this for a sample of 20 to 30 of your top queries rather than guessing from memory. Group the results by content type as you go: reference pages, comparisons, long guides, news. That grouping will usually tell you more about where you're exposed than any industry-wide percentage, because your content mix is not the average publisher's content mix. A site built mostly on long-form tutorials will see a smaller hit than a site built on quick-answer glossary pages, almost regardless of niche.

What publishers are actually doing differently

Watching how fast-moving publishers have adjusted their content mix through 2026, a few patterns show up repeatedly.

The clearest one: investment is shifting away from pure reference content and toward formats an AI summary structurally can't replace. Original research and surveys, tool comparisons that require someone to have actually used the products, and anything interactive, calculators, templates, community discussion, hold value an extracted paragraph doesn't.

Some publishers have gone the other direction on reference content specifically. Instead of fighting the extraction, they've restructured those pages to answer the core question in the first sentence, on the theory that the clicks they lose are mostly people who only wanted the one fact anyway. The clicks that remain skew toward readers who need actual depth, and those tend to convert better even at lower volume. That's a real tradeoff, not a hack, and it only works if you're honest about which of your pages are genuinely reference-only versus pages you've been padding to look comprehensive.

A smaller but growing group has started tracking AI Overview citations directly as a metric, the way they'd track rank position, and using the overview itself as a research tool: search your own target queries, read what gets cited and how the answer is structured, and treat that as a direct signal for what counts as a complete, extractable answer on that specific query. We go deeper on building that habit into a regular workflow in our guide to generative engine optimization for beginners, and in our roundup of free AI SEO tools if you'd rather not buy anything new to start.

Where AI Mode changes the attribution question

There's a second-order effect that's easy to miss if you're only watching click-through rate. Once a user follows up inside an AI Mode conversation instead of running a fresh search, your analytics stop seeing that person as a new query entirely. They're inside Google's interface for several turns before, if ever, a link takes them to your site. That's a different problem from losing a click on a single query. It's losing visibility into an entire research session that used to generate three or four separate search impressions you could track and optimize against individually.

This is part of why citation, not ranking position, is becoming the more useful thing to measure for AI-driven queries. A page can rank first for a query and still never get read if the AI Overview or AI Mode answer satisfies the question without surfacing a link at all. Conversely, a page ranking fourth or fifth that gets explicitly named and quoted inside the generated answer is doing real work, even if it never shows up as a click in Analytics. Treat citation tracking as a companion metric to rank tracking going forward, not a replacement for it. Rank still determines whether you're in the pool Google draws citations from in the first place.

The part that doesn't have a workaround

There's no tactic that reverses this for a genuinely single-fact query. If the complete, honest answer to a search fits in two sentences, an AI Overview is going to keep winning that exchange, and winning it more often as Google keeps expanding where overviews show up and how much room they're allowed to take. Chasing a workaround for that specific fight is a waste of a quarter.

The realistic response is an honest audit of your content mix: how many of your pages exist purely to answer one fact, how many require actual judgment, comparison, or multi-step execution to be useful, and whether the content you're commissioning right now keeps feeding the shrinking category instead of the growing one. That's not a one-time fix. It's closer to a standing editorial policy, and the 2026 numbers suggest it's the only lever that's actually working at scale. If you want this kind of analysis as it develops, our newsletter tracks these shifts as Google ships them, not months after the fact.

AI Overviews aren't the only place Google's models answer questions about your brand. The Gemini app draws on the same index, and our guide to getting your brand mentioned in Google Gemini explains how that grounding works.

Since Search Console blends AI Overview data into regular results, a dedicated tool is the practical way to watch your citations. See our roundup of Google AI Overviews trackers.

For the bigger picture behind these numbers, read what zero-click search is and how to win without the click. AI Mode's answers are also built with a technique Google calls query fan-out, which changes which pages get pulled in.

The latest expansion is on brand names: AI Overviews now show on about 80% of branded searches. For a weekly digest of changes like this, see our AI search news roundup.

FAQ

Did Google officially confirm AI Overviews cut organic traffic?

Google hasn't published its own click-through comparison publicly. The numbers in this piece come from independent research: Seer Interactive's tracking of 53 brands and 2.43 billion impressions, and Pew Research's panel study of actual search behavior across nearly 69,000 real queries. Both point the same direction, with different specific figures, which is why both are worth citing rather than relying on either alone.

Does getting cited inside an AI Overview make up for the lost clicks?

Partially. Seer's data found cited pages earn about 120% more clicks per impression than uncited competitors on the same query, but even cited pages still trail the click rate of a query with no AI Overview at all by roughly 38%. Being cited beats not being cited. Neither beats the overview not existing.

Do I need special schema markup or an AI text file to appear in AI Overviews?

No. Google's own documentation says there are no special requirements or optimizations beyond being indexed and eligible to show a normal snippet in Search. If a page already ranks and shows a snippet, it's already eligible for AI Overviews and AI Mode.

Which of my pages are safest from AI Overview traffic loss?

Long guides, multi-step tutorials, and anything requiring ongoing interaction or judgment, comparisons, troubleshooting, calculators, tools. Pure definitional or single-fact reference pages are the most exposed, based on both the Seer and Pew data.

Should I block AI Overviews from citing my content?

You can, with nosnippet or noindex, but both also remove the page from normal search snippets, which usually costs more traffic than it saves. It only makes sense for content you genuinely don't want surfaced in any summarized form, regardless of the source.