AI Overviews now appear in 43% of Google searches, up from 15% a year earlier, according to Similarweb’s new 2026 Generative AI Landscape report as covered by TechCrunch on July 27. That is a 28-percentage-point jump in twelve months — and notably, neither TechCrunch nor Similarweb’s own report page states the geography or device mix behind the figure.
The same report puts Google’s AI Mode at 279 million visits in May 2026, up from 126 million in June 2025, and finds that only 6.8% of US desktop ChatGPT queries carried a citation as of May 2026. Together the numbers sketch a search channel that is answering more and referring less — the structural shift we covered in depth in our zero-click data update. This post does not re-argue that thesis; it carries the new Similarweb data, and the denominator questions the coverage skipped.
Below: the headline figures with their scope caveats attached, a side-by-side scorecard of the competing AI-Overviews-presence estimates and the methods behind them, the two ChatGPT metrics everyone is about to conflate, and the measurement changes that follow for marketing teams.
- 0143% of Google searches now trigger an AI Overview.Per Similarweb’s 2026 Generative AI Landscape report, via TechCrunch — up from 15% a year earlier. Neither source states the geography or device scope, so treat it as Similarweb’s aggregate panel figure.
- 02Similarweb’s own page headlines a narrower claim.The report landing page leads with “>40% of US searches” — US-scoped messaging that is close to, but not confirmed identical with, TechCrunch’s scope-unspecified 43%. The two should not be silently merged.
- 03AI Mode visits grew roughly 2.2× in eleven months.From 126 million visits in June 2025 to 279 million by May 2026 — a 153-million-visit climb, per the same report, again with no geography or device breakdown disclosed.
- 04Semrush’s Sensor tool separately reads ~48%.A different method entirely — direct SERP observation across Sensor’s own keyword panel rather than panel-modeled traffic. Different denominator, same direction: AI Overviews presence is climbing fast on every tracker.
- 05Only 6.8% of US desktop ChatGPT queries carried a citation.As of May 2026, per TechCrunch. That query-level citation rate is a different metric from the share of ChatGPT visits that land on a webpage (25% in March 2026 rising to 60% by May 30) — this post keeps the two separated.
01 — The ReportFrom 15% to 43% in one year.
The core finding is stark. Similarweb — the Nasdaq-listed market intelligence firm whose panel data is routinely cited by SEO and marketing trade press — finds that Google’s AI Overviews have gone from appearing in roughly 15% of Google searches to 43% over the past year, per TechCrunch’s July 27 coverage of the firm’s 2026 Generative AI Landscape report. That is a 28-percentage-point expansion of AI-generated answers across Google’s result pages in twelve months.
One caveat belongs glued to that number, and most coverage has skipped it: the denominator is unstated. Neither TechCrunch’s article nor Similarweb’s public report page says whether the 43%/15% pair is US-only or global, desktop or mobile or blended. Similarweb’s own landing page instead headlines a separate, explicitly US-scoped teaser: “>40% of US searches now trigger an AI Overview.” The two figures are close, and they may even be the same measurement — but nothing published confirms that, so we present them as what they are: two adjacent, vendor-reported estimates.
AI Overviews presence · competing 2026 estimates vs the year-ago baseline
Sources: Similarweb 2026 Generative AI Landscape via TechCrunch; Semrush Sensor live dashboard“Google’s AI-generated answers in search, known as AI Overviews, have gone from appearing in 15% of searches to 43%, according to a new report.”— TechCrunch, July 27, 2026
Directionally, the finding is corroborated from outside Similarweb’s panel. Semrush’s Sensor tool — a live SERP-tracking dashboard that observes result pages directly across its own keyword panel rather than modeling traffic — currently reads AI Overviews presence at roughly 48% of the SERPs it tracks. Search Engine Roundtable’s own earlier coverage of this metric had it at about 13% to 20% a year ago, depending on the tracker. Whatever the exact denominator, every independent measurement agrees on the shape of the curve: AI Overviews went from a minority feature to something approaching half of tracked searches in about a year.
02 — Denominator ForensicsWhose 43%? A scorecard of the competing estimates.
Each piece of coverage we reviewed publishes a single number, and none of it puts the competing estimates side by side with their methods named — which is the only way to read this data honestly, because “43% of searches” means something different depending on who counted, what they counted, and how. The scorecard below is our reconciliation: four current estimates and two year-ago baselines, each with its window, stated scope, and measurement method.
| Tracker | Reported figure | Window | Stated scope | Method |
|---|---|---|---|---|
| Where the estimates stand now | ||||
| Similarweb, via TechCrunch | 43% of Google searches | Reported July 27, 2026 | Not stated — no geography or device disclosed | Blended traffic panel / clickstream modeling |
| Similarweb, own report page | “>40% of US searches” | Report teaser, undated | US searches (vendor’s own framing) | Same Similarweb panel, US-scoped messaging |
| Semrush Sensor | ~48% of tracked SERPs | Live tracker reading | Sensor’s own keyword panel — size and geography undisclosed | Direct SERP observation |
| BrightEdge + Ahrefs, per our May 17 data update | ~48% of tracked queries | March 2026 | Coverage tracking across verticals — our own post frames it as a 48-58% range depending on category | Direct SERP observation — different trackers and month; coincidentally similar to Semrush’s number, not the same data point |
| The year-ago baselines | ||||
| Similarweb, via TechCrunch | 15% of Google searches | ~July 2025 | Same unstated scope as the 43% figure | Blended traffic panel / clickstream modeling |
| Search Engine Roundtable, prior coverage | “about 13% to 20%” | ~Mid-2025 | Varies by the trackers it synthesized | Trade-press synthesis — trend corroboration only, not the same denominator |
Read together, the current estimates cluster between 43% and 48% across at least three independent trackers — Similarweb’s blended traffic panel, Semrush Sensor’s direct SERP observation, and the BrightEdge and Ahrefs coverage tracking reported in our May post. Two fundamentally different instruments sit behind those three readings: panel modeling on one side, direct SERP observation on the other. That convergence is the real story: no single tracker’s denominator is authoritative, but when panel modeling and direct SERP observation independently land in the same band, the underlying trend is hard to dismiss as measurement artifact.
One reconciliation trap worth naming: our own May 17 data update reported a ~48% figure for March 2026 — sourced there to BrightEdge and Ahrefs coverage tracking, and presented as part of a 48-58% range across verticals — that happens to numerically resemble Semrush Sensor’s current ~48% reading. They are different trackers measuring different keyword sets in different months — coincidentally similar, not corroborating each other. Similarly, Similarweb’s method is a multi-source blend — direct measurement, a contributor network of device-level data, algorithmically captured public data, and purchased partner data, per the company’s general methodology page — which is a fundamentally different instrument from a SERP scraper, so its 43% and Semrush’s ~48% should never be averaged or treated as bounds on one true number.
03 — AI Mode MomentumAI Mode visits, up 2.2× in eleven months.
AI Overviews are one of two Google AI surfaces in the report, and the second is growing faster. Google’s AI Mode — the separate, conversational, chat-style search surface — rose from 126 million visits in June 2025 to 279 million by May 2026, per the same Similarweb data via TechCrunch. That is 153 million additional monthly visits, roughly a 2.2× multiple in eleven months, with no geography or device breakdown disclosed for these figures either.
The momentum has an international dimension the report window only hints at: Google brought AI Mode to France on July 22, days before this report’s coverage landed — so the visit curve above was recorded largely before Google’s newest European expansion could contribute to it. Meanwhile the publisher side of this shift is escalating on its own track; some publishers are now openly weighing whether to block Google entirely.
Monthly visits · May 2026
Up from 126 million in June 2025, per Similarweb via TechCrunch. No geography or device breakdown was disclosed — treat as Similarweb’s aggregate panel measurement of the AI Mode surface.
Jun 2025 → May 2026
279 million visits against a 126-million base works out to roughly 2.2× in eleven months — while AI Mode was still rolling out to new markets, France included, only days before the report’s coverage.
15% → 43% in a year
The AI Overviews expansion and the AI Mode visit curve are two distinct metrics of the same underlying shift: answers rendered on Google’s own surfaces instead of clicks referred outward.
04 — The Citation CollapseTwo ChatGPT numbers everyone is about to conflate.
The report’s ChatGPT findings are where sloppy summaries will do the most damage, because TechCrunch’s article contains two different percentages with two different denominators. The first: just 6.8% of US desktop ChatGPT queries included a citation as of May 2026. Both qualifiers matter — the figure is explicitly US-only and desktop-only. TechCrunch notes citation rates grew more than fivefold over the prior year (the article does not state the baseline percentage, so neither will we) — yet even after that growth, more than nine in ten US desktop ChatGPT queries return an answer with no source attached at all.
The second number measures something else entirely: among ChatGPT visits, the share that actually lands on a webpage rose from 25% in March 2026 to 60% by May 30, 2026 — a jump TechCrunch ties to a May 7 ChatGPT search update. That is a visit-level metric about where clicks go when they happen, not a query-level metric about whether a source is shown at all.
Citation rate
Share of US desktop ChatGPT queries that included any citation, as of May 2026. Measures how often a source is surfaced at all — the metric that makes AI visibility hard to audit.
Referral landing rate
Share of ChatGPT visits that land on a webpage — 25% in March 2026 rising to 60% by May 30, 2026, which TechCrunch ties to a May 7 ChatGPT search update. Measures where engagement goes when it leaves.
05 — The FramingGateway to destination.
TechCrunch’s framing of the report is the cleanest one-line summary of what the numbers mean. The piece describes a shift from an era when Google provided a simple list of blue links to click through and read, to one in which Google itself is the destination — closing with the line that “Google transforms itself from a gateway to the wider web into a destination in its own right.” When AI-generated answers sit on approaching half of result pages and AI Mode adds 153 million monthly visits in under a year, the search engine’s center of gravity has moved from referring attention outward to holding it.
The report’s wider teaser stats — all vendor-stated on Similarweb’s landing page, with no public methodology attached — sketch the same fragmentation from other angles: Claude’s unique visitors reportedly grew about 180% in the second half of the period versus the first, ChatGPT’s usage share is described as declining as Gemini and Claude gain ground, Google search query length is up 5.4% since AI Mode arrived, and 26% of ChatGPT responses are said to now contain ads. Treat each of those as directional vendor color rather than audited findings — but the direction is consistent: the AI-search field is both growing and splintering at once.
Our read on the trend: this is the quarter the “gateway” model of search measurement formally broke. For twenty years the implicit contract was that presence in Google converted to sessions you could count in your own analytics. A 43%-and-climbing AI Overviews rate — whatever its exact denominator — means an expanding share of your search presence now produces impressions and brand exposure on Google’s surfaces that your analytics never sees. Projecting forward, if the past year’s trajectory holds even partially, every tracker’s estimate crosses 50% within quarters, not years — at which point “most searches” include an AI answer and session-based SEO reporting undercounts the channel by design.
06 — Operational ShiftWhat to measure now.
TechCrunch’s coverage ends at “Google is a destination now.” It does not tell a marketing team what changes operationally on Monday. Here is our translation of the gateway-to-destination shift into concrete measurement changes — the same reframing we run inside our agentic SEO engagements when a client’s organic sessions and their actual search visibility start telling different stories.
Track AI-surface presence directly
Organic sessions no longer proxy search visibility when AI answers sit on ~43-48% of tracked results. Add brand-mention and citation presence inside AI Overviews and AI Mode responses to your KPI set, monitored across your priority queries.
Treat citation rate as surface-specific
ChatGPT’s citation rate moved more than fivefold in a year, and its referral-landing share went from 25% of visits in March 2026 to 60% by May 30 — a move TechCrunch ties to a May 7 product update that landed late in that window. Benchmark per surface and per month — never bake a single citation rate into an annual plan.
Use Google’s own surface breakdowns
Third-party trackers disagree on the denominator. Google’s own reporting surfaces are the closest thing to ground truth it currently offers — start with the Search Console platform-properties breakdown covered in our July 10 post.
Name the denominator in every report
The lesson of the 43%-vs->40%-vs-~48% muddle: a percentage without its denominator is a vibe, not a metric. Every AI-search stat in client reporting should carry its tracker, window, and scope inline.
On the instrumentation point: Search Console’s platform properties give sites performance data broken out by surface — the nearest first-party signal for separating classic-search performance from AI-surface exposure as the mix shifts. Pair that with a measurement layer that reconciles first-party analytics against AI-visibility tracking — the stack design work we do in our analytics engagements — and the gateway-to-destination shift becomes something you can actually see in a dashboard rather than infer from declining sessions.
07 — ConclusionThe number is contested. The direction is not.
Every tracker disagrees on the denominator — and agrees on the direction.
Similarweb’s report gives the clearest single-vendor snapshot yet of the gateway-to-destination shift: AI Overviews on 43% of Google searches against 15% a year earlier, AI Mode visits up 2.2× to 279 million, and a ChatGPT ecosystem where only 6.8% of US desktop queries surface a citation at all. Every one of those numbers carries a scope caveat — and every independent tracker still lands in the same band, which is precisely what makes the trend credible.
The measurement mandate follows directly. Session-based SEO reporting undercounts a channel where approaching half of searches resolve on Google’s own surfaces. The teams that adapt first will be the ones tracking AI-surface presence alongside sessions, keeping query-level and visit-level metrics separated, and naming the tracker behind every percentage they put in front of a stakeholder.
And the honest posture on the data itself: hold the 43% loosely and the direction firmly. The specific figure is one vendor’s panel estimate with an unpublished denominator. The direction — corroborated by direct SERP observation from at least two independent trackers — is the thing to build strategy on.