Between January and April 2026, 68.01% of US Google searches ended without a click to the open web. That comes from SparkToro’s analysis of Similarweb desktop and mobile panel data, reported by Danny Goodwin at Search Engine Land on 9 June 2026, and it is up from 60.45% in 2024.
In the same dataset, just 0.34% of searches transitioned into AI Mode.
Both of those numbers are real and they sit awkwardly next to the figure Google gave at I/O 2026, which is that AI Mode has passed a billion monthly active users with query volume more than doubling each quarter. Reconciling those three statistics is the most useful analytical exercise in search right now, and almost nobody is doing it, because each camp quotes the number that supports its position.
What follows is the fourteen Google AI Mode statistics worth carrying into planning, each with its source, followed by what the contradictions actually mean.
One note on sourcing before the numbers. Most pages collecting Google AI Mode statistics cite each other rather than the original studies, which is how a figure from a 2025 panel sample ends up presented as a 2026 measurement. Every number below is attributed to the party that produced it, with the period it covers, because the period is frequently the difference between two statistics that appear to contradict each other.
The Scale Numbers Google Reports
Four figures, all from Google, all directionally useful and all measuring adoption rather than habit.
AI Mode surpassed 1 billion monthly active users globally, per Google’s I/O 2026 disclosure. Query volume has more than doubled every quarter since launch, also Google’s figure. The average AI Mode query runs roughly three times the length of a traditional search query. And follow-up queries grew more than 40% month over month on average in the US.
Read those carefully and notice what they do not say. Monthly active users counts anyone who touched the surface once in a month, which for a feature wired into the default Google experience is a low bar. Doubling quarterly from a small base is still a small number. Neither figure tells you what share of search demand actually flows through AI Mode, which is the number you would want and the number Google has not published.
The Behaviour Numbers Are the Interesting Ones
These five change what you write, which makes them more actionable than the scale figures.
More than one in six AI Mode searches is multimodal rather than plain text, per Google. Image-input queries have grown more than 40% month over month since launch. Image creation queries have more than tripled since the start of the year. Queries beginning with “which” are growing 40% faster than AI Mode overall over the past six months. And planning queries are growing 80% faster than AI Mode overall in the US.
The “which” and planning figures are the two to act on. A query that starts with “which” is a comparison request, and a comparison answer names specific options. A planning query assembles a sequence of recommendations. Both are shaped like a decision rather than like a lookup, which means the content that gets drawn into them is content that makes a specific, attributable claim about specific named things.
The Click Numbers Are the Painful Ones
Five figures, from three independent sources, and they agree more than they disagree.
68.01% of US Google searches ended without a click in January through April 2026, up 7.56 points from 60.45% in 2024, per SparkToro and Similarweb. The share of searches producing at least one click fell 9.51 points over that period, a relative decline of 22.9%. The share of searches leading to another Google search rose 7.2 points, which is where a good deal of that lost click volume went.
AI Overviews appear on more than 20% of Google searches, and when they appear, click-through rates fall by close to 60%, per the same study. Ahrefs separately measured a 34.5% CTR reduction for the top-ranking organic page when an AI Overview is present, which is a narrower measurement of a narrower thing and a useful cross-check.
For AI Mode specifically, Semrush clickstream sampling put the share of sessions producing an external click at roughly 6% to 8%. Similarweb panel data analysed by iPullRank for May through August 2025 put it under 3%. The honest planning range is 3% to 8%, and it is low enough that the distinction barely matters operationally.
Why Do the Numbers Contradict Each Other?
Because they measure three different things and get quoted as though they measure one.
A billion monthly active users measures breadth of exposure. A 0.34% session transition rate measures depth of habitual use within one panel on one set of devices in one country. Query volume doubling quarterly measures growth rate from an unstated base. None of those is wrong and none of them is the number you want.
There is also a measurement gap that matters more than people acknowledge. SparkToro explicitly excluded searches inside Google’s mobile app, where AI surfaces are more prominent and zero-click behaviour is probably higher. Panel data also systematically undercounts logged-in mobile app usage, which is where the majority of consumer search now happens. So the 0.34% figure is a floor rather than an estimate.
There is a third reason the figures diverge, and it is the one most likely to catch out a planning exercise. Google’s numbers are global. The clickstream studies are US-only. AI Mode adoption is uneven across markets, so a global monthly active user count and a US session share are not two views of one population. If your market is the UK, Australia, or anywhere outside the US, neither figure describes your traffic, and the honest position is that nobody has published a reliable number for you yet.
The resolution is unglamorous: AI Mode is a surface most people have touched and few people live in yet, growing fast, in a search environment where zero-click behaviour was already rising before AI Mode existed. The 2020 zero-click rate was 64.82% on SimilarWeb data, and 2024 was 58.5% in the US on Datos data. The trend line predates the feature.
What the Models Say When You Ask Them
So we ran the test rather than speculating about it.
On 9 October 2026 we asked Claude (Opus 5) directly: “What are the most reliable statistics on Google AI Mode usage in 2026, and how confident are you in them?” The response named the SparkToro and Similarweb zero-click study and Google’s I/O disclosures as the two primary sources, then flagged the reconciliation problem unprompted, noting that the billion-user figure and the sub-1% session share are not contradictory because they measure exposure and habit respectively, and that neither establishes what share of commercial search intent currently routes through AI Mode.
Two things in that answer are worth noting if you care about being cited. The model reached for named studies with named methodologies rather than for aggregated statistics pages, and it volunteered the limitation rather than presenting a clean figure. Content that states a number, names its source, dates it, and says what it does not prove maps onto how these systems construct answers. Content that presents a round number with no provenance does not get picked up, because there is nothing to attribute.
Ranking and Traffic Have Come Apart
This is the structural change the numbers describe, and it is bigger than any individual figure.
For twenty years, ranking first meant getting the click, so one metric tracked both visibility and traffic. That link is broken. You can rank first, be quoted in the answer, influence the decision, and receive no visit. A 34.5% CTR reduction on the top organic position is exactly what that looks like in a dashboard.
The practical consequence is that traffic is now a lagging and unreliable proxy for visibility. A brand can be gaining share of AI answers while its organic sessions decline, and a brand can be losing share of answers while its sessions hold steady on brand-name queries. Teams reporting only on sessions are flying on an instrument that has quietly stopped measuring the thing they care about.
What Should You Actually Measure?
Three things, and none of them is in a standard analytics report. The Google AI Mode statistics above tell you the environment has changed; measurement is how you find out whether it has changed for you specifically.
Citation share is the first: for the twenty or thirty questions that matter to your business, how often does an AI answer name you, and how often does it name a competitor instead. This requires running the queries on a schedule and recording the answers, which is manual and is the only way to get the number.
Branded search volume is the second, and it is the most honest proxy available for AI-driven awareness. A consumer who gets recommended in an AI answer and then searches your name converts into measurable demand. Rising branded search with flat non-branded traffic is the signature of AI visibility working.
Assisted conversions from direct traffic is the third. Visitors arriving with no referrer, on pages deep in your site, having clearly been told about something specific, are the measurable residue of an answer you were cited in.
What Actually Gets You Cited
The behaviour data points at the answer, and it is narrower than general content advice.
Make specific, attributable claims about specific named things, because comparison and planning queries are the fastest-growing categories and they need named options. Publish numbers with their methodology and date attached, because provenance is what makes a statistic quotable by a system that has to attribute. Answer the question in the first two sentences of a section, because extraction works on passages rather than on documents. And be described consistently across independent indexed sources, because entity confidence is what determines whether you are a candidate for citation at all.
That last point is where most of the real work sits. A brand with coherent, corroborated coverage across indexed publications is an entity these systems can resolve and cite. A brand with a website and nothing else is not, regardless of how well the website is written.
The Short Version
Fourteen numbers, three sources, one structural conclusion.
Zero-click is at 68.01% and rising, AI Overviews appear on over 20% of searches and cut CTR by close to 60%, and AI Mode sessions send a click to the open web somewhere between 3% and 8% of the time. Against that, AI Mode has a billion monthly users and under 1% of measured session share, which means the surface is early rather than dominant.
The actionable read is that the Google AI Mode statistics for 2026 describe a transition, not an arrival. Classic organic search still carries almost all the clicks that reach the open web, and the same indexed corpus feeds both. What has changed is that ranking no longer delivers traffic reliably, comparison and planning queries are growing fastest, and the thing worth optimizing for is being the source an answer cites rather than the link a user clicks.