Sensor Tower data published this week documents a sharp uninstall spike for ChatGPT's mobile app: 132 percent year-over-year in April 2026, and a much larger 413 percent jump in March that the firm attributes to backlash over OpenAI's Pentagon contract โ€” which Sam Altman himself later described as "opportunistic and sloppy." Over the same period, Claude installs grew 1,000 percent year-over-year against ChatGPT's 14 percent. This data lands the same week the Wall Street Journal reported OpenAI missed its internal target of one billion weekly active ChatGPT users by end-of-year 2025, and missed revenue targets. CFO Sarah Friar has warned the company may struggle to honor future compute contracts if monetization doesn't accelerate.

The numbers are uninstalls, not gross subscribers โ€” important to disentangle. ChatGPT remains the largest AI consumer product by absolute scale; uninstalls happening at 132 percent of last year's rate doesn't mean total users are shrinking, but it does mean the leaky-bucket dynamics are getting worse precisely as growth flattens. The Pentagon backlash is the more directly causal story: a 413 percent month-over-month uninstall surge in March is the kind of acute brand event you can map to a single news cycle. The 1,000 percent Claude install growth against 14 percent for ChatGPT cuts the harder way โ€” Anthropic crossed a $1 trillion secondary-market valuation last week, and the demand-side data is now corroborating the valuation gap rather than diverging from it. Builders watching this closely have noticed: the consumer chatbot war has shifted from "OpenAI dominant, others catching up" to "OpenAI defensible but ceding the growth curve."

OpenAI's response is structural. The Information reported the company is exploring ad-supported revenue, an option Altman has long resisted publicly. The more concrete play is ChatGPT Go โ€” an $8/month tier targeting 112 million subscribers by year-end. Internal projections reportedly assume Plus ($20/month) drops roughly 80 percent to 9 million subscribers, with the Go tier picking up the volume โ€” netting 122 million paid users total, roughly double current. That's a steep target on top of decelerating consumer demand and a competitor whose installs are growing 70x faster. Plausible if and only if Go monetizes well, ChatGPT-as-OS launches with retention hooks, and the Pentagon contract drag dissipates. None of those are bankable yet.

For builders, the takeaway isn't that OpenAI is dying โ€” it's that the market structure for AI assistants is bifurcating faster than the IPO timeline anticipated. Three things to watch. First, whether ChatGPT Go's $8 price compresses the entire consumer AI assistant market โ€” Claude Pro at $20, Gemini Advanced at $20, and Perplexity Pro at $20 will face downward pressure if Go gets traction. Second, whether ad-supported chatbot tiers actually arrive at OpenAI; the moment they do, Anthropic and Google have a free differentiation point ("we don't ad-target your prompts") that resonates with enterprise. Third, the Pentagon-contract uninstall spike is the most legible signal yet that consumer trust is now a load-bearing variable for AI platform economics โ€” not just a PR concern. If you're building consumer AI products, brand-trust events are starting to move retention metrics by triple digits month-over-month. That changes how you think about features, partnerships, and which contracts you sign.