AI Search Market Share 2026: Why the Leaderboard Won't Sit Still

AI search market share is not converging on a winner. It's still being fought over, live, and the fight is moving faster than almost any ranking shift a working SEO has ever had to track. In the six months between January and late July 2026, ChatGPT's share of AI-search activity fell from 78% to 56%, a 22-point drop inside two quarters. Over the same stretch, Google AI Mode crossed 1 billion monthly active users, and two analysts, using two different data pipelines, landed on nearly the same figure for how much of Google search now ends without a click: roughly 68%. Put those three facts next to each other and you don't get a stable leaderboard with Google and OpenAI trading a point or two. You get a market that is still being decided.

Most GEO advice published in the last two years assumes something close to a fixed hierarchy: optimize for ChatGPT first because it has the most usage, treat everything else as secondary. That assumption was defensible in 2025. It is not defensible now. The data from mid-2026 shows an AI search market share picture that moved more in two quarters than most Google algorithm eras moved in two years, and a program built around last quarter's leaderboard is already optimizing for the wrong weights. The practical cost of that mismatch isn't abstract. A team that spent Q1 budget shoring up ChatGPT-specific citation signals, while Gemini quietly picked up 30% of activity and AI Mode adoption pushed past a billion users, spent that budget defending a shrinking share of the problem instead of the growing one.

The AI search market share numbers nobody saw coming

Those figures come from Kevin Indig's "AI Halftime Report: H1 2026," published on Growth Memo on July 27, 2026. Indig has tracked AI-search activity share across engines for several cycles now, and this is the sharpest swing his reporting has captured. ChatGPT opened the year holding a commanding 78% share of the activity his methodology tracks. By the time the report published, that number had fallen to 56%, a drop of 22 percentage points in roughly six months. Over the same window, Gemini's share climbed to 30% and Claude's climbed to 10%. Those three figures don't sum to 100%, and that gap belongs to the smaller engines and tools Indig's report groups separately, our own arithmetic here, not a number the report states directly. Indig's methodology tracks activity, not raw query volume or ad revenue, which matters because activity share is the number that best approximates where buyer attention is actually going when someone reaches for an AI answer instead of a results page. It's the number a GEO budget should track first, ahead of citation counts or mention rate, because it sets the ceiling on how much any single engine's citations can possibly be worth this quarter.

The scale of the swing matters more than the direction. A 22-point move for the market leader in six months is not a gradual rebalancing. It's the kind of shift that would dominate the trade press for a year if it happened to a paid search platform's ad-spend share. In the anatomy of an AI citation, we argued that getting cited is a function of buildable signals rather than luck. That's still true at the page level. But the engine-share numbers underneath those citations are moving fast enough that a program tuned only to today's dominant engine is tuning itself out of relevance within a quarter or two. It also undercuts a common objection to acting on share data at all: that engine preferences are sticky, and switching costs keep users locked into whichever assistant they picked first. ChatGPT had the incumbency advantage, the largest user base, and the deepest habit formation of any AI assistant on the market, and it still lost more than a quarter of its own share inside two quarters. If incumbency wasn't sticky enough to hold 78%, no GEO program should assume today's runner-up stays in second place either.

Why ChatGPT's AI search market share cratered in six months

No single cause explains a 22-point drop, and Indig's report is a share-tracking study, not a causal one. What it does establish cleanly is the direction and the size of the move: ChatGPT lost share, Gemini and Claude both gained it, and the losses were not evenly distributed. Gemini's gain is the larger of the two, and it lines up with the simplest available explanation: Google has spent 2026 wiring Gemini into more default surfaces, Search, Workspace, Android, than any competitor can match through user choice alone. Default placement moves share in a way that product quality alone rarely does at this speed. None of that should read as a prediction that Gemini or Claude keeps gaining at this rate indefinitely. Distribution advantages plateau once they've captured the users reachable through that channel, and workflow-specific gains tend to concentrate in a ceiling set by the size of that workflow. The point isn't that Gemini becomes the permanent leader. It's that the mechanism behind this swing, default placement and workflow fit, can just as easily favor a different engine next year, which is exactly why quarterly re-measurement matters more than picking a new permanent leader to bet on.

Claude's rise to 10% is smaller in absolute terms but notable for where it's happening: enterprise and technical workflows, where Anthropic has focused product effort all year, rather than the consumer general-search use case ChatGPT still dominates. That's a structural detail worth sitting with. Share isn't just moving between engines, it's moving toward the specific tasks each engine has chosen to win, which means the AI engine market share numbers you should care about most are the ones inside your own buyer's workflow, not the aggregate headline figure. We wrote about this fragmentation problem directly in where citation readiness and search visibility diverge in 2026: a brand can be well cited on one engine and functionally invisible on another, and the aggregate share number hides exactly that split.

Google AI Mode's 1 billion users change what "search" means

The AI Mode figure is the one enterprise teams are underweighting. A billion monthly active users is not a beta cohort or an early-adopter segment. It's a scale figure that puts AI Mode inside the same tier as Google's largest consumer products, and it means the surface most buyers now hit first, when they type a query into the box they've used for two decades, is already a synthesized-answer experience by default for a meaningful share of queries, not a toggle a minority of users opt into. That distinction changes how you should read every AI Mode adoption headline from here forward. A billion monthly active users inside a feature is a very different fact than a billion monthly active users inside a habit two decades deep. The second kind of growth doesn't require a marketing campaign, an app-store ranking, or a deliberate switching decision. It only requires Google to keep rolling the experience out to more of the query box's existing traffic, which is precisely what's been happening through 2026.

This is the part of the story that a pure engine-share chart misses. ChatGPT, Gemini, and Claude compete for share of a category, AI chat, that a user has to actively choose to open. AI Mode doesn't need that choice. It sits inside the query box that already gets the world's search volume, which means its growth curve isn't bounded by category adoption the way a standalone chatbot's is. A brand that has spent two years building citation readiness for ChatGPT and Perplexity, and treating generative engine optimization as a bolt-on to a Google-first program, now has to treat Google's own answer surface as a GEO target in its own right, not an SEO afterthought.

The clickless search number two analysts couldn't ignore

Engine share is one half of the story. What happens after an answer renders is the other half, and here two analysts working from different data independently landed on almost the same number. Indig's H1 2026 report puts the share of Google searches that end without a click at roughly 68%. Rand Fishkin's SparkToro post, "In 2026, Less than One Third of Google Searches Still Send a Click," published June 9, 2026 and built on Similarweb clickstream data, puts the figure at 68.01%, up from 60.45% in 2024. Both numbers describe the same underlying behavior from two directions. Indig is measuring how often an AI answer satisfies the query without a downstream click; Fishkin and SparkToro are measuring Google's raw clickstream directly, counting what actually happens after a search-box query, AI-generated answer or not. That the two lines converge on nearly the same number, using two structurally different measurement approaches, is the detail worth sitting with longer than either individual figure.

“Two analysts, two different data pipelines, landing within a hundredth of a percentage point of each other. That kind of agreement doesn't happen by accident in an industry this noisy. It happens when the underlying shift is real, large, and no longer close enough to the margin for methodology differences to hide it.”

That convergence is the strongest piece of evidence in this entire dataset, stronger in some ways than either number on its own. Indig and Fishkin are not using the same source data or the same tracking approach, and growth-memo.com and sparktoro.com are not affiliated. When two independent methodologies agree to within a hundredth of a point, on a metric that jumped nearly eight points in two years, the finding stops being a single analyst's estimate and starts being closer to a fact of the current market. Zero-click search isn't a talking point from one report anymore. It's a measured, corroborated baseline, and any content strategy still modeled on click-through as the primary success metric is modeling the wrong outcome for two out of every three Google queries. We built our own content strategy work around exactly this assumption: that citation and mention, not click-through, are now the primary currency, and the numbers above are the clearest confirmation yet that the assumption was right.

Build a GEO program you can re-weight quarterly

Here's the position this data actually supports, and it's a different position than most GEO writing takes. The instinct, seeing ChatGPT still holding the largest single share at 56%, is to keep building a program around ChatGPT first and treat everything else as secondary. That instinct is backward-looking. Six months ago the same instinct would have pointed you at a 78% share that has since shed more than a quarter of its value. There is no reason to assume the next six months look calmer than the last six, and every reason in this dataset to assume they won't. The engines that are gaining, Gemini through default placement and Claude through workflow-specific adoption, are gaining for structural reasons that aren't going to reverse on their own.

This is the same discipline we apply in engagements built around topical authority in AI search: breadth and re-testability beat a narrow bet on any single engine's current preferences, because the preferences keep moving and the breadth doesn't have to be rebuilt when they do. We've run this exact model with clients who came in over-indexed on one surface. Our work with Zenity started from a citation profile built almost entirely around one engine's quirks, and the fix wasn't a bigger bet on that engine, it was building measurement and content that held up as the underlying share numbers moved underneath it.

Do this next. Pull your own mention-rate data by engine, not blended, and check whether your program's effort allocation still matches Indig's H1 2026 shares: 56% ChatGPT, 30% Gemini, 10% Claude, plus AI Mode tracked as its own surface given its 1 billion monthly active users. If your content and measurement stack still treats ChatGPT as 70-plus percent of the opportunity, it's running against a market that closed that gap months ago. Rebuild the tracking cadence to quarterly, re-run the allocation math every time, and treat this quarter's numbers as a snapshot, not a forecast. The leaderboard is the volatile part. Build the program to survive that, not to bet on it staying still.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.