AI Overviews Cut Publisher Clicks 40%; Schema Won't Fix It

The traffic picture keeps getting worse

The most consequential number in search this year isn't a ranking factor. It's a referral figure. Chartbeat's 2026 Publisher Playbook reports that Google Search referral traffic to news and media sites fell 40.2% between July 2025 and July 2026, a decline it attributes largely to AI-generated summaries that answer questions directly on the results page. 3 Chartbeat's network leans heavily toward news organizations. Its report also notes that publishers are increasingly relying on direct visits, loyal readers, and private "dark social" link sharing to keep audiences growing. 3

A separate account put the global decline at 33%. It describes AI responses that "resolve intent" inside the search interface before anyone clicks. 4 The two figures differ, likely because they measure different publisher sets and time windows. They agree on direction and on cause.

The damage is not limited to small sites. Business Insider reportedly lost more than half its search traffic and cut 21% of its staff. 4 Chegg sued Google on antitrust grounds, citing a 49% collapse in non-subscriber traffic it blames on AI Overviews. 4 HubSpot has also reported steep drops in informational search traffic on queries its blog had ranked for over a decade. 4 The same report adds that smaller publishers are absorbing the sharpest losses. 4

The industry's answer: get cited instead

The SEO world has largely responded by changing the goal from winning the click to being quoted inside the AI answer. One agency argues that the publishers adapting fastest "stopped optimizing for the click" and rebuilt their sites to be citation-ready. 4 Practitioner guidance now stresses answer-first page structure, question-based headings, consistent entity information, structured data, and audits that find "citation gaps" in AI search. 5

Two recent studies are fueling the structured-data push. A study attributed to Princeton University and Moz reportedly tracked 500,000 URLs from October 2025 to March 2026. 1 It found that pages with no schema were cited in AI Overviews 4.2% of the time. Pages with Article schema reached 11.3%, and pages with FAQ schema reached 13.6%, roughly 3.2 times the baseline. 1 The study also reports that FAQ quality mattered. Markup with fewer than three questions produced only about a 1.4x lift. 1

A second analysis looked at 1,000 US-English desktop queries collected in April 2026. It examined 4,243 cited URLs against a control set of about 50,000 pages that weren't cited. 2 That analysis found schema-equipped pages were 2.3 times more likely to be cited. In its data, HowTo schema produced the strongest lift at 2.8x, not FAQ. 2

Where the citation data falls short

The studies agree that structured markup and AI citations go together. They disagree on how big the effect is (3.2x versus 2.3x) and on which schema type matters most. Those gaps should make readers wary of precise multipliers. Both findings are also correlations. Sites that use clean, thorough schema tend to be well-maintained in other ways, so markup may partly be a sign of overall quality rather than the cause.

The bigger problem is that a citation is not a visit. The second study describes how AI Overviews pull specific passages, statistics, and claims, then attribute them with numbered links or inline badges. 2 That is exactly how a reader gets the answer without leaving Google. Tripling a citation rate from 4% to 13% still leaves most cited pages being quoted, not visited.

The reading: citation work is defense

In my view, the numbers point to a clear conclusion. Optimizing for citations is worth doing, but it is a defensive move. It keeps a brand visible and attributed in the place where search now happens. It does not bring back the referral volume publishers have lost. A 40% drop in Google referrals is a structural change in how discovery works. 3 Better markup cannot undo that change.

That is why the more telling detail in Chartbeat's report may be where publishers are finding readers now: direct traffic, loyal audiences, and dark social sharing. 3 These are channels publishers own or influence themselves, rather than ones they rent from a search engine. The sensible plan treats schema and answer-first formatting as basic hygiene. The real growth effort goes into newsletters, apps, communities, and subscriptions that don't depend on Google sending traffic.

Publishers that treat citation optimization as the full answer risk a familiar outcome. They would be helping AI Overviews answer questions more effectively while their own traffic keeps falling.