3 min read

The a16z Consumer AI Report Results (Who Uses What)

The a16z Consumer AI Report Results (Who Uses What)
The a16z Consumer AI Report Results (Who Uses What)
5:52

ChatGPT has 900 million weekly active users. That's more than 10% of the global population, every week, using one product. By any reasonable measure, OpenAI has won the consumer AI race.

Except the race just changed shape.

Andreessen Horowitz's latest consumer AI report — tracking the top 50 generative AI products by web traffic and mobile MAU as of January 2026 — is the most comprehensive snapshot of where people are actually spending time with AI. Three years into this list's existence, the authors made a significant editorial decision: they widened the definition. CapCut, Canva, Notion, Grammarly — products most people think of as software tools, not AI products — now qualify, because AI has become load-bearing infrastructure inside them.

That definitional shift is the most important thing in the report. Not the rankings.

The "Default AI" Race Has a Different Finish Line

ChatGPT is 2.7 times larger than Gemini on web traffic. It's 8 times larger than Claude on paid subscribers. Those numbers are real and they matter.

What's also real: Claude grew paid subscribers over 200% year over year as of January 2026. Gemini grew 258%. Roughly 20% of weekly ChatGPT web users also used Gemini in the same week. The category is not consolidating around a single winner — it's exhibiting multi-tenant behavior, with users running different models for different tasks.

The report frames the endgame correctly: this may not resolve like the search wars, in which one player captured 90% of the market. It may resolve like mobile operating systems — two platforms with fundamentally different philosophies, both building trillion-dollar ecosystems. OpenAI is chasing the consumer super-app: shopping, booking, health, and daily transactions. Its app directory already includes 85+ integrations across travel, food, lifestyle, and entertainment — Expedia, Instacart, Zillow, MyFitnessPal. The vision is for ChatGPT to be the starting point for everything.

Claude's integration catalog looks nothing like that. Financial data terminals. Developer infrastructure. Medical and scientific tools. A growing open-source MCP community with no ChatGPT equivalent. Anthropic is, deliberately, building for the user who pays more, expects more, and works differently.

Both strategies are coherent. Neither is obviously wrong.

New call-to-action

The Quiet Collapse of Standalone Creative Tools

Three years ago, Midjourney ranked in the top ten. It's now at number 46. That trajectory tells the image generation story cleanly: as GPT Image and Gemini's Nano Banana improved, the value proposition for standalone image tools compressed. Nano Banana generated 200 million images and brought 10 million new users to Gemini in its first week. When the platform model gets good enough, the specialist tool loses its reason to exist.

The exceptions are instructive. Suno retained its rank in music generation. ElevenLabs has appeared on every edition of this list since September 2023. The pattern: where OpenAI and Google focused creative investment, standalone traffic collapsed. Where they didn't — music, voice — there's still room.

For anyone building or investing in AI-powered creative tools, the strategic question is whether your capability sits in a category the platform giants have targeted yet. If it does, the clock is running.

Agents Are Here. They're Just Not Mainstream Yet.

The most consequential section of the report covers the shift from AI that answers to AI that acts.

OpenClaw — an open-source local AI agent built by Austrian developer Peter Steinberger — went from a solo side project to 68,000 GitHub stars in weeks, and by early March became the most-starred project on GitHub, surpassing both React and Linux. The report frames it plainly: if ChatGPT was the moment consumers discovered AI could talk, OpenClaw may be the moment they discovered AI could act.

OpenAI acquired it in February.

But OpenClaw still requires Terminal knowledge to set up. Manus and Genspark, both on the list, handle open-ended multi-step tasks — research, analysis, slide generation — for users who don't want to configure anything. Manus was acquired by Meta in December 2025 for an estimated $2 billion. Genspark raised $300 million and announced a $100 million revenue run rate.

The agentic layer is forming. It isn't consumer-grade yet. The next six months will determine which platform — ChatGPT, Claude, Gemini, or a standalone agent — becomes the default executor for everyday tasks.

The Metric Problem Nobody Wants to Talk About

The report ends with an honest admission: the rankings are increasingly undercounting the AI products people use most. Claude Code hit a $1 billion annualized revenue run rate in six months — and barely registers in web traffic data because developers use it in the terminal, not a browser. A knowledge worker who dictates every email via an AI assistant isn't counted in app MAU figures.

AI is moving from destination to feature. That's not a problem for the technology. It's a problem for anyone trying to measure it, price it, or build a content and growth strategy around where attention is actually going.

The US ranks 20th globally in per capita AI adoption, behind Singapore, the UAE, Hong Kong, and South Korea. The country that built most of these products is not leading in their use.

That's the kind of data point that should probably show up in more strategy documents than it currently does.


Winsome Marketing helps growth teams build AI strategies calibrated to where the market is actually moving. Talk to our experts at winsomemarketing.com.

The Karpathy Contrarian: AI's Most Respected Voice Is Pumping the Brakes on Reinforcement Learning

The Karpathy Contrarian: AI's Most Respected Voice Is Pumping the Brakes on Reinforcement Learning

When Andrej Karpathy speaks, the AI community listens—and his latest message is one the industry desperately needs to hear. The former Tesla and...

Read More
AI Can't Hide Its Own Thinking — And That's Actually the Good News

AI Can't Hide Its Own Thinking — And That's Actually the Good News

OpenAI published research this week with a finding that sounds alarming on the surface but lands as something closer to reassuring: frontier AI...

Read More
AI's Real Test Isn't Efficiency—It's Crisis Response

AI's Real Test Isn't Efficiency—It's Crisis Response

The productivity metrics look impressive. Teams generate content faster, analyze data more efficiently, and process information at unprecedented...

Read More