# The Cracks in the Monolith: AI Assistant Market Data You Cannot Optimize Around > ChatGPT just fell below 50% of the AI assistant market for the first time. Six numbers explain why single-platform AI search strategies no longer work." - Canonical: https://www.aixindar.com/news/the-cracks-in-the-monolith-ai-assistant-market-data-you-cannot-optimize-around - Markdown: https://www.aixindar.com/news/the-cracks-in-the-monolith-ai-assistant-market-data-you-cannot-optimize-around.md - Author: Daoyu Guan — https://www.aixindar.com/experts/daoyu-guan - Published: 2026-09-03T08:28:05.434Z - Last updated: 2026-09-03T08:28:05.511Z - Evidence checked: Not separately recorded in CMS - Editorial status: Published - Corrections: No correction record supplied by CMS. For three years, "optimize for AI search" has implicitly meant "optimize for ChatGPT." That assumption quietly expired in May 2026, when ChatGPT's share of the AI assistant market **fell below 50% for the first time** — 46.4%, down from above 50% in January, according to Sensor Tower's State of AI 2026 report. This is not a story about ChatGPT declining. It is a story about the ground under every AI search strategy shifting from a monopoly to a portfolio problem. The numbers below are the ones that should be driving your platform decisions — and most of them are not the ones being shared in strategy decks. ## The Six Numbers ### 1. 46.4% — and falling download share ChatGPT remains enormous: roughly **1.1 billion monthly active users** (it crossed one billion in May 2026, the fastest climb in consumer software history) and annualized revenue around **$25 billion** as of February 2026. But the direction of travel is the story. In Q2 2026, ChatGPT accounted for **47% of AI app downloads, down from 67% a year earlier**, while Gemini took 22% and Claude took 14% — Claude's share of downloads was roughly 1% for all of 2025. A market where one player loses twenty points of download share in a year is a market where "the leader" is a moving target. ### 2. +627% — the growth asymmetry Claude's monthly active user base grew **627% year over year** to roughly 245 million (10.3% share); ChatGPT grew 67% in the same window. Gemini sits second at roughly 662 million MAUs (27.7%). Growth rates this lopsided mean the citation behavior you measure today will be running on a different audience mix within two quarters. ### 3. 360 billion hours — usage doubled in a year Time spent in AI apps roughly doubled from 172 billion hours in the first half of 2025 to about **360 billion hours in the first half of 2026**, with the top three assistants capturing 89% of it. Consumer spending on AI apps exceeded **$4.2 billion** in the same period, more than double the prior year. Whatever AI assistants are to search behavior, they are no longer a novelty layer on top of it. ### 4. $2.76 vs $1.74 — the monetization inversion Claude's average revenue per user (`2.76**) now substantially exceeds ChatGPT's (**`**1.74**), and Claude converts users to paid at **13%, the highest rate in the category**. In Q1 2026 mobile revenue growth, Claude led at +235%, Gemini +119%, Grok +103% — and ChatGPT +4%. The incumbent is winning volume; the challengers are winning value per user. Different incentives produce different product decisions — including different decisions about citations, sourcing, and answer formats. ### 5. \~90% — the old search is still the reference frame Context cuts both ways. By Statcounter's traditional measure, Google still handles roughly **90% of global search queries**. Reuters Institute's 2026 data put Google's referral volume at about **500× ChatGPT's** (1,300× including Discover). The honest framing: AI assistants own a fast-growing slice of user attention, while classic search still owns the overwhelming majority of referral volume. Rand Fishkin's caution applies — claims that AI "owns the start of the consumer journey" rest largely on survey perception, not observed behavior at scale. ### 6. 11% — the fragmentation tax Here is the number that makes the other five an optimization problem rather than trivia: per Profound's analysis of billions of citations, **the domains Perplexity cites and the domains ChatGPT cite overlap by only about 11%**. The engines do not merely weight sources differently — they largely read different internets. Add the platform citation profiles documented elsewhere in this series (Perplexity leans on Reddit and primary sources; ChatGPT leans on Wikipedia and established publishers; Google's AI Overviews lean on Reddit, YouTube, and Quora) and the conclusion is unavoidable: **there is no single "AI search" to rank in.** ## What the Portfolio Problem Actually Requires The strategic implications are mechanical, not aspirational: **1. Measure per-platform, never in aggregate.** A blended "AI visibility score" across engines hides the 11% overlap problem. A brand can be dominant in Perplexity answers and absent from ChatGPT's — the average of those two facts is a number that means nothing. This mirrors the finding from Seer Interactive's measurement research that most "AI visibility" dashboards lack a unified KPI; the fix is platform-by-platform baselines, not better averaging. **2. Weight effort by audience, not by hype.** ChatGPT's user base is still the largest pool by an order of magnitude. A rational allocation for most brands still starts there — but the growth rates argue for a Claude and Gemini monitoring layer now, before the citation behavior of those audiences calcifies. **3. Diversify source formats by platform preference.** Because citation profiles differ structurally (listicles, comparison pages, Reddit discussions, primary data, Wikipedia adjacency), a single content format cannot serve all engines. The earning strategy is portfolio-shaped on the supply side too: structured comparison content for some engines, community presence and primary data for others. **4. Re-baseline quarterly.** With monthly citation volatility running at 40–59% across engines (Amsive's tracking) and market share moving this fast, any AI visibility measurement older than a quarter is historical document, not dashboard. Treat visibility as a dynamic share — closer to share-of-voice advertising metrics than to classic rankings. **5. Watch the monetization asymmetry.** Engines optimizing for revenue per user have different structural incentives around answer depth and citation than engines optimizing for raw engagement. The ARPU inversion above is early, but it is the kind of economic pressure that reshapes citation behavior over a product cycle. ## The Honest Caveats Three limits on this data, stated plainly: - **App-panel data skews mobile** and undercounts desktop, API, and third-party integrations — where a large share of AI usage actually happens. ChatGPT's official metrics (900 million weekly active users, 50 million paid subscribers, February 2026) use different definitions than Sensor Tower's monthly actives. Treat all numbers as directional, not gospel. - **Share is not sentiment.** Nothing here says users are dissatisfied with ChatGPT; usage hours grew across the board. Fragmentation and dissatisfaction are different phenomena, and only the former is clearly documented. - **The market is young enough to re-concentrate.** Two years of share data in a category this volatile is a weather report, not a climate model. ## The Bottom Line The comfortable era of "the AI engine" — singular — is over, and it ended faster than most strategists noticed. A market where the leader's download share drops twenty points in a year, where the fastest-growing challenger is growing at nine times the incumbent's rate, and where the engines cite overlapping sets of only 11% of domains, is a market where platform diversification is not optional sophistication. It is the baseline competence. The brands treating AI visibility as a portfolio — measured per engine, rebased quarterly, supplied by different content formats — will find the fragmentation navigable. The ones still optimizing for a single monolith are solving last year's problem. --- ### Sources - Sensor Tower, State of AI 2026 (May 2026 data): ChatGPT 46.4% share, 1.1B MAU; Gemini 662M / 27.7%; Claude 245M / 10.3%, MAU +627% YoY; Q2 2026 download shares; ARPU and conversion figures; 360B hours H1 2026; $4.2B consumer spend - OpenAI official metrics (February 2026): \~900M weekly active users, \~50M paid subscribers, \~$25B annualized revenue - Statcounter GlobalStats, search engine market share (\~90% Google, global, 2026) - Reuters Institute, Journalism, Media and Technology Trends and Predictions 2026 (280 media leaders, 51 countries): Google referrals \~500× ChatGPT (\~1,300× with Discover) - Profound cross-engine citation analysis (2024–2025): \~11% domain overlap between Perplexity and ChatGPT citations - Amsive monthly citation volatility tracking (2025–2026): Google AIO 59.3%, ChatGPT 54.1%, Copilot 53.4%, Perplexity 40.5% - Seer Interactive, "AI Visibility Is Lying to You" (2026): majority of AI search categories lack a unified KPI - Cross-context: SparkToro/Similarweb downstream impact study (2026) for the perception-vs-behavior caveat *All market figures are as of the report dates above; this category moves fast — verify before citing in evergreen contexts.* ## Editorial references - [Editorial policy](https://www.aixindar.com/editorial-policy) - [Research methodology](https://www.aixindar.com/research-methodology) - [Corrections policy](https://www.aixindar.com/corrections)