a16z Top 100 AI: Only 4.5% of US Users Pay for Chatbots

Consumer AI has plenty of users and comparatively few customers. The seventh edition of Andreessen Horowitz's Top 100 ranking of consumer AI products shows that only a small slice of Americans pay for AI. That slice pays a lot.

For the first time, the venture capital firm (often shortened to a16z) adds actual spending data to its usual web traffic and mobile user counts. The spending figures come from YipitData panels of US consumer card transactions. They do not represent total revenue. They do give a rare look at how people pay for AI products.

A list that has stopped changing

After three years, the top of the ranking looks increasingly fixed. Only eleven products made the list for the first time, the lowest number of newcomers in any edition so far. Just seven companies rank in the top tier across all three of a16z's rankings. They include ChatGPT, Claude, Perplexity and Canva.

Broad use, thin wallets

The headline numbers point to a wide but shallow market:

1. Adoption: Nearly half of US consumers use AI. Only about a quarter use it daily.
2. Subscriptions: In August, 4.5 percent had an active paid personal subscription to ChatGPT, Gemini or Claude. That is roughly double the share from a year earlier.
3. Concentration: The top one percent of spenders accounts for nearly a fifth of all observed AI spending. That is more than the entire bottom half of spenders combined.
4. Spending levels: Those top spenders average about $900 a month, and that figure keeps rising. The typical paying user spends around $25, a level that has held steady for some time.

The heavy spenders are mostly prosumers. They use building and automation tools such as n8n, Manus and fal at much higher rates than other users. They also lean on creative tools like Higgsfield, Figma and HeyGen. Professional work is what drives the money, and a16z calls this group consumer AI's first real paying market. The firm's view that consumer AI is still early fits that picture.

The gap between users and payers also shows up in how deeply people use AI. Paid plans offer far more features but take more effort to learn. Free users tend to stay on the surface. The Decoder notes this could widen a skills gap between heavy users and everyone else, and possibly deepen existing social inequality. The other possibility is that AI keeps getting cheaper and easier until the gap stops mattering. OpenAI CEO Sam Altman has framed this as "unlimited brilliance" for everyone.

Three chatbots, three trajectories

ChatGPT still leads clearly. It gets about twice Gemini's web traffic and six times Claude's, and the lead is even larger on mobile.

The biggest change is Claude's rise to a clear third place. It was not on the first web list in 2023. Since then it has passed Deepseek and Perplexity in traffic and drawn level with Gemini among US subscribers. Anthropic also monetizes better than its rivals. A much larger share of Claude users is on the most expensive tier, which starts at $100 a month, than on comparable OpenAI and Google plans.

Momentum shifted in midsummer, though. Daily Claude sessions dropped in July and August, while ChatGPT picked up again with new models and ChatGPT Work. According to a16z, the real question is less who wins and more how well each company uses its window.

Personal agents start spending money

According to a16z, an agent like OpenClaw was not usable for most consumers six months ago. Now startup agents such as Instinct and Tomo report hundreds of thousands of users. Big 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. They then spend four-figure sums per month through the service, much of it on travel.

Platforms are picking 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 more than five million downloads in under a month. That is still modest next to Meta's Threads, which passed 15 million downloads in its first three weeks.

Subscriptions, not ads

Established companies are best placed to profit. Canva and Notion sit in the web top ten, and Google alone holds five web spots. According to a16z, startups find room mainly through a differentiated model, a multi-model experience, a specific audience, or by getting around entrenched interfaces.

The business model is the harder problem. Most top AI-native products rely on subscriptions or usage fees. Only a small fraction use ads or transaction fees. That reverses the pre-AI internet, where advertising paid for most of Meta's and Alphabet's revenue. High model costs mean AI companies cannot give up early revenue to build a user base first.

Some alternatives are taking shape. OpenAI reported a $1 billion annualized revenue run rate from ChatGPT advertising in August. Personal agents could also earn through affiliate fees or a cut of transactions.

Our Take

The spending data changes how the market should be read. Usage rankings have long suggested a mass market. Card data suggests the money comes from a small group of professionals who treat AI as work software. For readers, this is a useful reminder that free tiers are not the whole product, and that the gap between casual and power users is partly a question of what people pay for.

This also adds context to earlier signals. Six months ago, a16z warned that traffic and downloads say less and less as AI becomes an embedded feature. Moves like ChatGPT's visual display ads suggest providers are already testing ways around subscription limits.

Three things are worth watching. First, whether Claude's monetization lead survives its summer slowdown. Second, whether agent commerce grows beyond early adopters, especially as rules for AI shopping agents take shape and platforms like Amazon decide whom to let in. Third, whether the $25 typical spend starts to move at all.