The Top 100 Consumer AI Apps: Who’s Actually Paying?

The seventh Top 100 report added consumer card-spend data, excluding enterprise and SMB spending, and revealed a gap between attention and payment. Only 11 new products appeared across the web and mobile traffic lists, the fewest in the report series, while 29 of the 50 products ranked by spending appeared on neither list. Olivia Moore said traffic suggests the market is settling, but spending shows it remains a power-user market. About half of Americans report using AI and roughly 25% interact with it nearly every day, yet YIPPIT panel data puts the share paying for an AI subscription at about 4.5%; other estimates discussed range from 2% to 2.5% or somewhat higher.

The paying share has roughly doubled in a year, but revenue is highly concentrated: the top 10% of paying consumers account for more than half of AI revenue, the top 1% for 20%, and the bottom 50% for about 16%. Among consumers who pay for AI, the top 1% spend an average of $903 per month on personal credit cards, compared with a $25 median monthly spend. YIPPIT’s highest spenders skew toward developer, productivity, and creative tools, including N8N, Kretnola, Higgs field, and Manus. Moore said this reflects people using AI to code, make art and videos, and create things they could not make before. ChatGPT leads global consumer AI apps by usage and revenue; its web audience is about six times Claude’s and twice Gemini’s, though revenue may be especially informative when paying and usage are concentrated among power users.

In the US-spender panel, Claude has more paid subscribers than Gemini despite Gemini’s larger user and install base, while ChatGPT has about three times as many paid consumer subscribers as either. Anthropic’s no-ads approach was linked to a stronger subscription focus: about 7.5% of its subscribers are on plans costing at least $100 per month, compared with about 1% for ChatGPT and Gemini. Consumer assistants that act on users’ behalf are attracting early adopters but have not reached mainstream use. Instinct reported 100,000 users, 10% day-over-day growth, and 40% connecting a credit card in its first three weeks; average first-month spending exceeded $1,000. Muse recorded about 5 million downloads in the US and Canada over roughly 22 days, versus Threads’ 16 million, and its growth appeared less aggressively driven, partly because serving users is expensive.

Trust, privacy, security, and onboarding a user to a blank-box interface remain barriers to wider assistant adoption, and assistants have not yet developed person-to-person network effects. Some reportedly cost hundreds or thousands of dollars per user per month to serve, while products aimed at mainstream consumers were seeing costs in the tens of dollars. Amazon blocks Muse from browsing and purchasing on its site, although Muse has hundreds of other partnerships and can buy through Shopify. Assistant Benchmark’s community data showed coding and technical automation as the most common agent use cases, including among users of consumer assistant products.

About 85% of products on the consumer AI list monetize through subscriptions, 62% through credits or token-based extra usage, and 13% through ads or other models, with categories overlapping. Speakers argued that subscriptions alone are an unusual foundation for mass consumer technology revenue. Falling inference costs could make advertising and transaction fees more viable, but high serving costs currently push companies to charge early and make unmetered growth hard to control. One speaker said OpenAI had reached a $1 billion annualized advertising run rate by August, with the figure possibly higher the following month; the company’s last reported figure was 1.2 billion weekly active users. That scale and user knowledge could support targeted ads, but ads in AI conversations need clear labeling and relevance to preserve trust, and economics vary between inexpensive queries and costly tasks such as coding or image generation. OpenEvidence was reported to reach 50–60% of US physicians, suggesting focused vertical AI products may also support targeted advertising.

The strongest consumer AI successes so far cluster around getting things done: productivity, photo and video, design and editing, and search and answers. ElevenLabs and Suno rank near the top of both traffic and spending lists, with Suno leading in music; music-related intellectual-property issues were raised as one possible reason large labs may not prioritize competing models. OpenAI’s Images 2.0 was described as strong for image creation and editing, while Google’s Nano Banana and Veo have drawn traffic away from standalone image generators. Midjourney disappeared from traffic rankings but returned in revenue rankings, which Moore connected to power users valuing its aesthetic qualities and fine-tuning. Video remains a “wild frontier,” with the speaker citing Chinese companies’ data access and the challenge of generating and synchronizing visuals and sound.

The discussion’s broader bet is that value can accrue to the consumer experience layer, not only to the underlying model: tailored interfaces, workflows, personal context, and community can make AI capabilities accessible and harder to replace, while infrastructure value may accrue separately to model or inference providers. Work-adjacent apps such as WhisperFlow and Granola can spread through individual use and later add enterprise privacy, security, and team plans; Krea, Replit, ElevenLabs, and Gamma were cited as consumer products expanding into enterprise. Most pure consumers still use AI as a search replacement, with homework help and workplace email also common.