Google AI Overviews and Publisher Traffic: What the Numbers Show and How to Get Cited

You still rank on the first page, and your clicks are still falling. That is the confusing part, because the two used to move together, and now they do not. Around August 27, 2026, first spotted by SEO analyst Chris Long and reported by Search Engine Roundtable, publishers began seeing that Google had removed the "Show more" button from AI Overviews for a set of queries and let the answer expand on its own instead. A Google spokesperson said the overviews now "dynamically expand for topics where our systems determine it's most useful for people." In plain terms, the AI answer opens to full size by default, and your ranked link sits further down the page than it did the week before. This piece is about what the traffic data actually says, why a first-page ranking no longer guarantees the click, and what a publisher does about it that works.

The change sounds small. A button moved. What it signals is not small, because default expansion means the AI answer is now the first thing most people read, and the ten blue links have become the thing they scroll past to reach. Google did add one guardrail, so that if a reader has already started scrolling, the automatic expansion stops to keep their place. That tells you Google knows the expansion is aggressive. The question for a publisher is no longer whether this reshapes discovery. It is how much, and what you do with your own data once you can see it.

What Google actually changed in August 2026

Before this change, an AI Overview appeared as a condensed snippet, and a reader had to click "Show more" to see the full answer. That extra click was friction, and friction kept some readers moving down to the organic results. Google removed that friction for the queries where its systems decide a longer answer helps, and the overview now expands to full height on load. The result is more screen taken by the AI answer and less room above the fold for the sources underneath it.

This matters because attention on a search page follows position, and position just moved. An article that holds the number one organic spot can now open below a full AI Overview on a phone, which is where most publisher traffic lives. The ranking did not change. The real estate above it did. That gap between where you rank and where you actually appear is the mechanism behind the numbers in the next section, and it is the reason so many publishers describe the same strange pattern of steady rankings and shrinking traffic.

The traffic data, and what it actually says

The clearest large-sample read comes from Brainlabs, which analyzed Google Analytics data from 54 clients over 14 months, from January 2025 to April 2026, as reported by Digiday. The sample covered 24 U.S. firms, 24 British brands, and 6 international clients. Across the full period, organic traffic fell 10.5 percent, from 140.1 million sessions to 125.4 million, and 46 of the 54 clients saw a decline. The turning point was September 2025, the month AI Overviews began appearing on at least 30 percent of U.S. search results. Average monthly organic traffic for the group dropped from roughly 20 million sessions to under 15 million after that line was crossed.

A 10.5 percent group average hides how uneven the hit is. According to Ahrefs, AI Overviews have cut web traffic to some sites by as much as 58 percent, which is the ceiling, not the norm. A separate survey by Digital Content Next of 19 member publishers, run over eight weeks in May and June 2025, found a median year-over-year referral decline from Google Search of 10 percent overall, 7 percent for news brands, and 14 percent for non-news brands, with individual losses ranging from 1 percent to 25 percent. Three studies, three methods, and the same direction of travel. The disagreement is about size, not about whether it is happening.

The sharpest evidence is at the query level, and it explains the "still ranking, still losing" pattern better than any average. The Professional Publishers Association shared member data showing a lifestyle publisher whose click-through rate for a specific query fell from 5.1 percent to 0.6 percent even though the page held its first-page ranking. An automotive publisher saw a 25 percent traffic drop as its click-through rate slid from 2.75 percent to 1.71 percent. A Pew Research survey of 900 U.S. consumers pointed at the cause from the reader's side, finding that AI summaries make people less likely to click through to links at all.

The page still ranked first. The click-through fell from 5.1 percent to 0.6 percent. Ranking and traffic have quietly become two different things.

There is a second half to the Brainlabs data that most coverage skips, and it changes what you do next. Over the same period, AI referral traffic to those clients rose 163 percent, conversions from AI sources rose 335 percent, and traffic that arrived from an AI answer converted at 1.5 times the rate of traffic from classic organic search. So the picture is not only subtraction. A smaller number of readers are reaching sites through AI answers, and the ones who do arrive with more intent, because the AI has already done the browsing for them and sent through the person who still wanted the source. That is the opening, and it is why the response below is about being the cited source rather than fighting to reclaim a blue link.

Why ranking first no longer means getting the click

Search worked on a simple bargain for two decades. You answered the question well, Google ranked you, and Google sent the reader to your page to read the answer. The expanded AI Overview breaks the bargain at the last step, because now the answer is assembled and shown on the results page, and the reader gets what they came for without a click. Your ranking still decides whether the AI considers your page as a source. It no longer decides whether the reader lands on it.

This hits informational and evergreen content hardest, which is the exact content most publishers built their libraries around. A "how many cups in a quart" page, a symptom checker, a definition, a step count, all of these are precisely what an AI Overview can answer in two sentences, so the click that used to follow the ranking now evaporates into the summary. Transactional and deeply reported pages hold up better, because a reader comparing products or chasing original reporting still needs to leave the results page. The optimization lens here is not a verdict that search is over. It is a sort. Your library now splits into pages the AI can fully answer and pages it can only point at, and the second group is where your surviving search traffic will concentrate.

Measure your own exposure before you react

The worst move right now is to act on a headline number instead of your own analytics, because the 10.5 percent group average and the 58 percent ceiling are both true and neither one is your figure. PIB's method on every channel is the same, which is continuous analysis and optimization of your own data rather than a one-time setup, and this change is a clean example of why. Before you change anything, read what your own pages are doing.

Start in Google Search Console and sort your top queries by impressions against clicks. The pages to worry about are the ones where impressions hold steady or rise while clicks fall, because that divergence is the AI Overview absorbing the answer above you. Then, in Google Analytics, segment your organic landing pages and separate the informational pages from the transactional ones, so you can see which part of your library is exposed and which part is insulated. Cross-check against the queries that now trigger an AI Overview, and you will have a ranked list of the pages losing clicks they used to earn.

The technical tease for the operators who want to instrument this properly is worth naming, because you can automate the read rather than checking by hand each week. You can wire a scheduled pull against the Search Console API in an n8n flow, run the same job as a Make scenario, or write a small cron against the API directly, and have it flag any query where the click-to-impression ratio drops week over week. You can also add referral segmentation in GA4 for the AI hosts, so visits from chatgpt.com, perplexity.ai, and gemini.google.com show up as their own channel instead of hiding inside direct traffic. That last step matters because the Brainlabs numbers only exist for publishers who were counting AI referrals in the first place, and most dashboards were not. If you are not measuring the new channel, you cannot tell the difference between a reader who vanished and a reader who switched roads.

Being the cited source, not just a ranked link

Once the answer is assembled on the results page, the win condition changes. You are no longer only competing to rank. You are competing to be one of the sources the AI pulls into its answer, and to be named when it does. This is the discipline PIB calls Generative Engine Optimization, and its core metric is AI Citation Presence, meaning how often and how prominently an AI answer cites you rather than a competitor. The GEO Readiness Score is the composite that tracks it across the pillars that decide it, from Content Authority to AI Citation Presence to Entity Positioning.

The practical work sits in three places. Technical Retrievability is whether an AI system can cleanly parse and lift your page, which rewards clear factual sentences, real structured data, and answers stated plainly near the top rather than buried under a preamble. Entity Positioning is whether the model understands who you are and what you are an authority on, which is built through consistent naming, an about page an engine can read, and third-party references that agree on your identity. Topical Authority is depth, meaning a cluster of useful, in-depth pages on one subject rather than a single thin post, because models cite the source that looks like it owns the topic. The tactical schema mechanics for news pages, from NewsArticle markup to freshness signals, are covered in our companion piece on getting cited in AI Overviews Top Stories, so the point to hold here is the strategic one. The page that gets cited is the page an engine can retrieve, trust, and attribute, and that is a property you build on purpose, not one you inherit from a ranking.

The real fix is diversification, not a better blue link

Every number in this piece points at one exposure, which is depending on a single funnel that a single company can narrow overnight with a button. The durable answer is not to win back the click that AI Overviews took. It is to stop being hostage to one source of discovery in the first place, a case we make in full in what the end of Google referral traffic means for publishers. PIB describes the response as the publisher operating system, and the identity claim underneath it is diversification for stability. A publishing business runs across Facebook, Google Discover, content syndication, AI search, and asset sales, so that any one channel narrowing is a dented quarter rather than a broken business.

On the Facebook side, that means treating the page as a distribution engine you control, where Curation decides what you publish and how you shape it to the audience, and Virality turns one strong post into a monetized reach event. On the Google side, it means keeping the transactional and original-reporting pages sharp while accepting that thin informational pages will keep feeding the summary, and it means working the discovery levers Google still hands publishers directly, such as becoming a Preferred Source in Google's feed and search profiles. On the AI side, it means building the citation presence above so that when the answer is assembled, your name is in it. The through-line across all of them is the same optimization loop, which is reading what is already earning and pushing more of it, rather than setting a channel up once and hoping it holds. The publishers who came through the last year in decent shape were not the ones who guessed which channel would win. They were the ones who were not betting the business on any single one.

Where to go from here

If you want to know exactly where you stand in AI search before you spend a quarter guessing, the GEO Authority System is built for this moment. It runs an LLM Visibility Evaluation across ChatGPT, Claude, Perplexity, and Gemini, scores your AI Citation Presence with the GEO Readiness Score, and hands you a 30, 60, and 90 day playbook to become the source an answer cites in your niche. Right now the ground in most niches is still open, because AI citation patterns are forming and have not hardened. You can get the GEO Authority System for $499 and start from a real diagnosis of your own visibility rather than a headline about someone else's traffic.

Frequently asked questions

What did Google change about AI Overviews in August 2026?

Google removed the "Show more" button for a set of queries and let AI Overviews expand to full size by default. A Google spokesperson said the overviews now dynamically expand where the systems decide a fuller answer is most useful. The effect is that the AI answer occupies more of the screen on load, which pushes organic results, including first-page rankings, further down the page.

How much traffic are publishers actually losing to AI search?

It depends heavily on the site, so treat any single figure with care. A Brainlabs analysis of 54 clients found organic traffic down 10.5 percent over 14 months, with 46 of 54 declining. Ahrefs put the ceiling at as much as 58 percent for some sites. A Digital Content Next survey of 19 publishers found a median decline of 10 percent, ranging from 1 percent to 25 percent. The direction is consistent even where the size is not.

Why is my traffic dropping if my rankings held?

Because ranking and traffic have separated. An expanded AI Overview can sit above your number one result, so the reader gets the answer without scrolling to your link. The Professional Publishers Association documented a page whose click-through fell from 5.1 percent to 0.6 percent while it kept its first-page ranking. The position stayed. The clicks did not.

Should I block Google's AI crawler to protect my content?

For most publishers this backfires, because Google uses overlapping systems for search and AI answers, so blocking the AI can cost you search visibility too. The more durable response is to measure your own exposure, keep your transactional and original pages strong, build citation presence so you are named in the answer, and diversify your discovery across more than one channel.

What is the difference between ranking in search and being cited in AI?

Ranking decides where your link appears in a list of results. Citation decides whether an AI answer pulls your page in as a source and names you. As answers move onto the results page itself, citation is becoming the more valuable position, because it puts your brand inside the answer the reader actually reads.

Key takeaways

• Around August 27, 2026, first spotted by Chris Long and reported by Search Engine Roundtable, Google let AI Overviews expand by default, which pushes organic results, including first-page rankings, lower on the page.
• The traffic data agrees on direction: 46 of 54 clients declined in the Brainlabs study, with a 10.5 percent group average and a 58 percent ceiling from Ahrefs.
• Ranking and traffic have separated, with documented cases of click-through collapsing while the ranking held, and informational and evergreen pages are the most exposed.
• The Brainlabs data also shows AI referral traffic up 163 percent and converting 1.5 times better than organic, so the readers who arrive through AI answers come with more intent.
• Read your own data first, using Search Console impression-to-click divergence and GA4 referral segmentation for the AI hosts, before acting on any headline number.
• The durable fix is being the cited source through AI Citation Presence and diversifying discovery across the full publisher operating system, so no single channel can narrow the business overnight.

Sources

• Search Engine Roundtable, Google removes "Show more" and expands AI Overviews by default, first spotted by Chris Long (2026-08-27): https://www.seroundtable.com/google-ai-overviews-push-ai-mode-responses-41974.html
• Digiday, In Graphic Detail: how AI search has impacted the web traffic of over 50 advertisers, Brainlabs analysis of 54 clients (2026): https://digiday.com/marketing/in-graphic-detail-how-ai-search-has-impacted-the-web-traffic-of-over-50-advertisers/
• Digiday, Google AI Overviews linked to a drop in publisher referral traffic, Digital Content Next member survey (2026): https://digiday.com/media/google-ai-overviews-linked-to-25-drop-in-publisher-referral-traffic-new-data-shows/
• Ahrefs research on AI Overview traffic impact, as cited by Digiday (2026)
• Professional Publishers Association member click-through-rate data, via Digiday (2026)
• Pew Research Center, survey of 900 U.S. consumers on AI summaries and click-through behavior (2026)