Few people pay for AI, but those who do spend big

Andreessen Horowitz has published the seventh edition of its Top 100 AI list, a ranking of the most-used consumer AI products, and for the first time the venture firm has added observed spending data rather than relying only on web traffic and mobile user counts. The spending figures come from US consumer card panels supplied by YipitData, which a16z notes do not represent total revenue. The headline finding is a market with wide adoption and shallow monetization: nearly half of US consumers use AI, only a quarter do so daily, and just 4.5 percent had an active paid personal subscription to ChatGPT, Gemini or Claude in August. That paying share is up from roughly half that a year earlier, with ChatGPT accounting for most of those subscribers.

Within that narrow paying base, revenue is unusually concentrated. The top one percent of spenders generate nearly a fifth of all observed AI spending, more than the entire bottom half of payers combined. That top percentile averages about $900 per month and is still climbing, having increased its spending 79 percent since early 2025. The typical paying user spends about $25 and has held steady at that level for some time. The product preferences of the big spenders point to who they are: they disproportionately use building and automation tools such as n8n, Manus and fal, alongside creative tools including Higgsfield, Figma and HeyGen. A16z's reading is that these are prosumers buying software to build, design and work, and that professional use is what drives the spending. Only seven products appear in the top tier across all three rankings, among them ChatGPT, Claude, Perplexity and Canva.

At the top of the rankings little has changed. ChatGPT leads the web with roughly double Gemini's traffic and six times Claude's, and the mobile gap is wider still. The biggest shift is Claude's emergence as a clear number three. Claude did not appear on the first web list in 2023 but has since passed Deepseek and Perplexity in traffic and pulled even with Gemini among US subscribers. Anthropic also monetizes better than its rivals in one respect: a much larger share of Claude users sit on the priciest tier, which starts at $100 per month, compared with equivalent plans from OpenAI and Google. Claude passed Gemini in paid US subscribers in March 2026, and since Google restructured its plans in June the two have run neck and neck. By midsummer the momentum shifted again, with daily Claude sessions falling in July and August while ChatGPT regained speed on new models and ChatGPT Work. A16z frames the question less as who wins than as how well each company capitalizes on its window.

A second front is the race for a personal assistant. Six months ago an agent like OpenClaw was not usable for most consumers, according to a16z. Now startup agents such as Instinct and Tomo report hundreds of thousands of users, and larger players are responding with Meta's Muse, OpenAI's Dots and xAI's Grok Bot. Agents become an economic force once they trigger real purchases: Instinct founder Noah Shinn says 40 percent of users link a credit card within three weeks and then spend four-figure amounts per month through the service, much of it on travel. The platforms are starting to pick sides. Amazon blocked Meta's Muse within two weeks, while Shopify, Instacart, OpenTable and others signed official integrations. Muse reached a quarter million daily users in its first week and passed five million downloads in under a month, though that is modest next to Meta's Threads, which pulled in more than 15 million downloads in its first three weeks.

The business model question is the harder one. Established players look best placed to make money from AI: Canva and Notion rank in the web top ten, and Google alone holds five web spots. Startups carve out space mainly through a differentiated model, a multi-model experience, a specific audience, or by bypassing entrenched interfaces. Yet the vast majority of leading AI-native products rely on subscriptions or usage fees, with only a small fraction using advertising or transaction fees. A16z argues this inverts the pre-AI internet, where users were themselves the product and advertising drove the bulk of revenue for Meta and Alphabet, because high model costs make it hard for AI companies to forgo early revenue while building a large user base. Alternatives are appearing: OpenAI reported an annualized revenue run rate of $1 billion from ChatGPT advertising in August, and personal agents could open a path through affiliate fees or transaction cuts. The prior edition of the ranking, six months earlier, focused more on geographic splits, including the rise of Yandex Browser with its Alice assistant in Russia and Deepseek's role bridging Western and Chinese user bases; even then a16z warned that traffic and downloads say less and less about which products people actually use as AI shifts from a standalone product to an embedded feature. The new spending data is a direct attempt to close that gap.

The engagement gap is the other thread running through the report. Paid accounts offer far more features but are more complex and demand a steeper learning curve, so free users typically scratch the surface of what AI can do. That creates a skills gap between people who use AI at a high level and those who barely use it or do not use it at all, which could deepen existing social inequality, or alternatively AI keeps getting easier and cheaper until the gap stops mattering. OpenAI CEO Sam Altman has invoked "unlimited brilliance" for everyone in this context.

Read together, the numbers describe an AI consumer market that is broad in reach but narrow in payment, and highly stratified where payment exists. The panel-based spending figures cover US consumer cards only, cannot be split from bundled subscriptions in Gemini's case, and are not total revenue, so they are best treated as a directional signal rather than an accounting of the market. A likely trade-off is that the current subscription-heavy model maximizes revenue per professional user while capping reach among casual users, and advertising or agent-driven transaction fees would trade some of that density for scale.

Why it matters: developers and product teams should read the top one percent spending about $900 a month on build, automation and creative tools as evidence that professional workflows, not casual chat, are where consumer AI money currently sits. If you are pricing or positioning an AI product, the paid tier appears to reward depth and capability over simplicity, while the free tier remains a large but weakly monetized funnel. Watch whether agent commerce and advertising matures enough to fund that funnel, and whether cheaper models narrow the skills and subscription gap over the next year.