Claude DAU is up 848%

Wall Street’s latest cut on the AI market is not a single scoreboard—it is diverging growth curves. In early April 2026, Bank of America Securities summarized third-party figures (Similarweb, Sensor Tower) for March. The read is sharp: Claude is breaking out, Gemini is compounding, ChatGPT is softer on some near-term measures, and no single assistant owns the field yet—rotation is the story.

Taken together, that snapshot sends a clear message: AI agents and AI coworkers are moving from experiments into real work and everyday life—how teams run pipelines, how people use phones and desktops, and how software stops being “a chat box” and starts behaving like a teammate.

The split leaderboard (March)

In one widely cited cut, Similarweb data puts Claude at roughly 20 million average daily web visits (up about 90% month over month), with mobile DAU up about 848% year over year to the single-digit millions—consistent with product-led adoption (e.g. coding and agent workflows) and model quality, not a one-off stunt.

ChatGPT is painted as mixed: global web traffic down about 6% month over month in that summary, with mobile DAU lower by on the order of ten million plus users—useful as a reminder that share shifts between assistants as people compare capability and price.

Gemini is described as compounding: on the order of eighty million average daily web visits (headline year-over-year growth in the hundreds of percent in some tables), mobile DAU up strongly year over year and still growing month over month. Analysts in the same coverage argue much of Claude’s uptake comes from share shifting among AI assistants and adjacent products—not a single static leaderboard.

Why we care

Zeus is on a mission to build the best AI employee infrastructure, with harness engineering—the controls and wiring that make agents reliable in production. We start with e-commerce because the workflows are concrete: inventory, tickets, campaigns, and money on the line. The future of the AI workforce is coming; we are building the layer teams need to deploy it without heroics.

Posts that break down full stacks—like our OpenClaw workflow notes—matter because they spell out the whole system, not just the flashy model name.

Figures summarized from public coverage of Bank of America Securities research (April 2026) and third-party analytics; see Longbridge: March AI landscape for cited ranges. Verify against primary vendor and issuer sources before trading or budgeting off headline stats.