ChatGPT Drops Below 50%: AI Assistant Market 2026

AI assistant market share in 2026 just crossed a symbolic threshold: ChatGPT’s share of global AI assistant users fell to 46.4% by the end of May, the first time it has slipped below 50%, according to Sensor Tower’s State of AI 2026 report. The single-app era of generative AI is giving way to a multi-platform market — and the implications run far deeper than one product’s decline.

The same report puts Google Gemini at 27.7% of global assistant users and Anthropic’s Claude at 10.3%. Yet in the very month its relative share dipped under half, ChatGPT also crossed 1.1 billion monthly users — making it the fastest mobile app in history to reach a billion users. A platform can grow in absolute terms and shrink in relative terms at the same time, and that paradox is the story most headlines miss.

This guide unpacks what the numbers actually say, why distribution is beating model quality, how Claude inverts the share story on revenue, and what a fragmenting assistant market means for your generative engine optimization (GEO) strategy. Every market-share, user-count, and revenue figure below comes from Sensor Tower’s State of AI 2026 unless otherwise attributed — one research firm’s measurement, not an independently audited industry count.

The headline figure is straightforward: Sensor Tower’s State of AI 2026 report, published June 16, puts ChatGPT’s share of global AI assistant users at 46.4% by the end of May 2026. That is the first reading below 50% the firm has recorded. The report dates the actual crossing under half to March 2026 — the May figure is where share settled after the line was crossed, not the moment it happened.

What makes the number meaningful is the trajectory behind it. This is not a single bad month; it is the tail of a long decline. Sensor Tower’s series, as reported by TechCrunch, shows ChatGPT holding 65.3% in December 2024, falling to 52.8% by December 2025, and reaching 46.4% by May 2026 — a drop of nearly 19 percentage points across 17 months. The slope has been remarkably steady, which is what makes the trend, rather than the milestone, the real signal.

Here is the nuance almost entirely absent from the headline coverage: ChatGPT broke the all-time record for fastest app to a billion users in the same month its market share fell below 50%. Per Sensor Tower, ChatGPT crossed 1.1 billion monthly users in May 2026, reaching the one-billion milestone in roughly three years — faster than TikTok, YouTube, or Instagram, each of which took five to eight years. The platform is expanding in absolute terms while contracting in relative terms.

Both things are true because the pie is growing faster than any single slice. Sensor Tower projects global time spent on generative AI apps reaching 36 billion hours in the first half of 2026, more than double the 17.2 billion hours of H1 2025. AI app downloads are tracking toward roughly 2.3 billion for the half, and in-app revenue is projected to surpass $4 billion. When the whole category roughly doubles, a market leader can add hundreds of millions of users and still lose share — because challengers are growing faster from a lower base.

The strategic read for marketers: do not mistake ChatGPT’s share decline for ChatGPT’s irrelevance. It remains the single largest assistant by a wide margin and is still adding users at record pace. The change is that it is no longer a proxy for “AI search” as a whole. A second platform now commands more than a quarter of users, and a third is the fastest-growing of the major assistants. Planning around a single destination is the mistake — not betting on ChatGPT.

Coverage of the Sensor Tower report has been siloed by metric — share in one story, revenue in another, engagement in a third. The table below combines them into a single at-a-glance reference for the three major assistants. The combined-reach column is computed directly from the share figures (each platform’s True Audience share, plus the running total when you stack platforms in descending order of share) to show exactly how much of the audience each tier of GEO coverage captures.

The most counter-intuitive finding in the report has nothing to do with ChatGPT. Despite holding just 10.3% of users, Anthropic’s Claude generates more revenue per user than ChatGPT. Sensor Tower puts Claude’s average revenue per US mobile user at $2.76 in May 2026 versus ChatGPT’s $1.74 — roughly 1.5x more per user from a far smaller base. Claude also converts best of any major assistant, with 13% of its users paying for a subscription.

That climb is recent and steep. Claude’s US mobile ARPU rose from under $0.50 in September 2025 to $2.76 by May 2026, while its True Audience share grew 452% year over year and its monthly active users moved from roughly 60.2 million in December 2025 to about 245 million by May 2026 — close to fourfold growth in five months. Market share and market value have detached: the platform with the smallest user share of the three monetizes each user the hardest.

One caution on attribution: the $2.76 ARPU figure is citable, but a head-to-head conversion-rate comparison is not. Sensor Tower supplies Claude’s 13% paid-conversion figure; the coverage does not publish a comparable ChatGPT conversion rate, so we leave that cell as “not disclosed” above rather than infer it. The defensible claim is the ARPU gap and Claude’s class-leading conversion — not a precise multiple on conversion between the two.

For brand strategists the lesson is about audience quality, not just audience size. A smaller, higher-intent, higher-paying user base can be worth more than a larger casual one — which is exactly the logic that should govern where you invest GEO effort. Reaching the platform where your buyers actually convert may matter more than reaching the platform with the most raw users.

How did Gemini reach 27.7% of users without winning a model-quality race? Distribution. Sensor Tower attributes much of Gemini’s growth to OS-level integration across Android and Google Workspace rather than standalone app downloads — and the engagement data backs it up. Gemini’s average time per user rose from 14 minutes a month to 100 minutes, roughly a sevenfold increase, as the assistant moved from a destination people choose to visit into a default baked into devices and productivity tools.

eMarketer’s framing of the same shift is worth quoting: AI access is expanding from chatbots into familiar tools, devices, and services. Gemini wins on Android and Workspace; Microsoft Copilot wins on Windows and Microsoft 365; Meta AI wins on WhatsApp and Instagram reach. ChatGPT, by contrast, competes largely on standalone merit without an operating-system home of its own. When AI becomes ambient — embedded where users already are — the platform with the best distribution collects usage that the platform with the best model does not.

A late-February 2026 episode showed how quickly trust now moves users between assistants. After OpenAI announced a partnership with the Department of Defense, US ChatGPT uninstalls surged 295% day over day on February 28, and one-star reviews for the app jumped 775% the same day, per TechCrunch citing Sensor Tower and Appfigures data. Over the same window, Anthropic publicized that it would not partner with the Pentagon — and Claude’s US downloads jumped 37% on February 27 and 51% on February 28, taking the app to #1 on the US App Store and, for the first time, past ChatGPT in daily US downloads.

It is important to be precise about what this does and does not show. The crossing below 50% happened in March 2026; the DoD backlash was a late-February event. We are not claiming the backlash caused the crossing — they are two separate data points. What the episode does establish is the underlying behavior the report describes more broadly: users are increasingly willing to switch assistants, and a values mismatch can move installs by double digits within a single weekend. Brand trust is now a switching signal, not just a marketing line.

Here is the practical hook for anyone running generative engine optimization. If ChatGPT holds 46.4% of assistant users, a brand that optimizes only for ChatGPT visibility reaches less than half the AI-assistant addressable audience by definition. Skip Gemini and you forgo another 27.7%; skip Claude and you forgo 10.3% more. Stacked, the three majors cover 84.4% — and the trend line points toward further fragmentation, not consolidation, which means single-platform coverage erodes over time even if you do nothing.

Fragmentation is not only about reach; it is about behavior. Each engine applies different retrieval filters, so the same content does not surface identically everywhere. Industry GEO research suggests strong Google SEO translates fastest into Gemini visibility, Claude tends to reflect new content over a span of weeks, and Perplexity tends to show movement faster — though these timelines are vendor estimates and vary by niche, so treat them as directional rather than guaranteed. The takeaway is that “optimize for AI search” has quietly become a multi-target problem.

We have tracked this shift in our own analysis of how AI referral traffic is already fragmenting across platforms, and the assistant-share data here is the upstream cause: where users chat is increasingly where they search, and that is no longer one place. For the mechanics of earning citations across engines, our guide to getting cited across AI search engines covers the per-platform tactics in depth.

The fragmenting market changes the GEO operating model in four concrete ways. None of these requires abandoning ChatGPT — they require treating it as the largest of several targets rather than the only one that matters.

Operationally, this is a monitoring-and-measurement problem before it is a content problem. You need visibility into how each engine cites your brand, a content structure that travels well across different retrieval filters, and a feedback loop that tells you which platforms actually send qualified traffic. That is the core of how we approach agentic SEO and multi-platform GEO for clients — instrumenting visibility across engines rather than optimizing for a single destination. If you are building this from scratch, our generative engine optimization strategy for 2026 is the right starting framework.