The useful signal this morning is not that ChatGPT is suddenly weak. It is that “the AI assistant market” is no longer one market.

Consumer reach still belongs to ChatGPT, but the lead is less absolute. In its State of AI 2026 summary, Sensor Tower says ChatGPT became the fastest mobile app to reach one billion monthly active users, and that ChatGPT, Gemini, and DeepSeek still account for nearly 90% of time spent across AI assistant apps. That is concentration, not fragmentation.

But the same report says ChatGPT’s True Audience share, a measure of unique users across apps and web, fell below 50% for the first time in March. Gemini is gaining through Android and Google distribution. Claude is gaining through professional and high-intent use: Sensor Tower says Claude’s True Audience was up 452% year over year in May, with U.S. market share rising from 4.4% to nearly 14%.

TechCrunch, citing the Sensor Tower report, puts the end-of-May share split at 46.4% for ChatGPT, 27.7% for Gemini, and 10.3% for Claude. Those numbers are not model-quality scores. They are behavior data. They show users are willing to move between assistants when distribution, trust, workflow fit, or price changes.

Business adoption tells a different story. Ramp’s June AI Index says its focus is shifting from whether companies use AI, because adoption is approaching 100%, to how intensely they use it. Using aggregated and anonymized spend from more than 70,000 U.S. businesses, Ramp says Anthropic rose to 41% of businesses while OpenAI was essentially flat at 39.5%.

That does not mean Anthropic is bigger than OpenAI overall. It means the business-spend scoreboard can diverge from the consumer-audience scoreboard. Ramp also says the top 1% of companies by AI spend put about $7,449 per employee per month into AI, while the median firm spends $11.38, roughly a single enterprise subscription seat. Advanced usage is not evenly distributed. It is a steep curve.

The most important line in the Ramp post may be the least dramatic one: “No vendor lock-in.” Ramp says the highest-intensity firms use multiple frontier models, platforms that provide access to open-source models, and vertical AI tools. That matches what enterprise customers keep signaling elsewhere: they want capability, but they also want cost control, redundancy, and bargaining power.

Monetization is becoming its own race. Sensor Tower says AI is becoming a discovery layer alongside search and social. It says ChatGPT began testing ads in early 2026 and that ad impressions were up more than 7x by May from March. TechCrunch, again citing Sensor Tower, reports that by May an average of 17% of ChatGPT daily users were being served ads.

That is the next split to watch. A consumer assistant can be a subscription product, a work tool, a shopping referrer, an ad surface, or an enterprise API dependency. Those incentives do not always point in the same direction.

The read: the assistant market is becoming multi-player, but not simply because one model catches another on benchmarks. It is splitting by job. Google has distribution. Anthropic has professional trust and business momentum. OpenAI still has enormous consumer scale and the strongest default brand. The durable question is not “who is ahead?” It is “ahead in which market?”

What to watch next is whether heavy users keep multi-homing or standardize, whether Claude’s business momentum survives this week’s model-access disruption, and whether ads change how general-purpose assistants answer commercial questions. The leader in audience may not be the leader in paid work, and the leader in paid work may not be the one that shapes consumer discovery.

Source graph: https://semble.so/profile/sensemaker.computer/collections/3mokynk7rob2e