The AI Tool Landscape: Every Major Player, Mapped by Use Case

The AI tool landscape is usually drawn as a single leaderboard, with ChatGPT sitting on top of everything else. By one measure that is fair: ChatGPT still records more web visits than any rival, by a wide margin. But the word doing the quiet work here is “used.” A leaderboard treats usage as one number you can rank, when the figure underneath it is at least three different things.

Disaggregate them and the map changes shape. Web visits measure the consumer website, not the app, the API, or the tool embedded inside Word or an IDE. And those gaps fall unevenly across products. Separate the interface from the model it runs on, and a sprawl of dozens of “players” resolves into a handful of model-makers. The landscape is not a ranking. It is a structure, and the structure is what this map is built to show.

ChatGPT’s lead is growing and shrinking at the same time

The headline is 900 million weekly users, a record. Read as dominance, it misleads. ChatGPT’s share of chatbot web traffic fell from roughly 77% to about 55% in a year, while Gemini’s rose from around 6% to 27%. Absolute reach and relative share are moving in opposite directions. The mechanism is a market growing faster than any one player: the leader can add users while losing the category, so “biggest” and “winning” have quietly come apart.

The tools that look smallest are often the most embedded

On a web-traffic map, Microsoft Copilot looks marginal and Meta AI barely registers. Both readings are artefacts of the metric. Copilot sits on roughly 20 million paid Microsoft 365 seats; Meta AI has over a billion users, almost all inside WhatsApp and Instagram. Web visits measure destinations, and these tools are not destinations. They are features inside software people already had open. The dashed rings on this map mark where that gap is largest.

A dozen “competitors” run on the same few engines

Counted as logos, the landscape looks crowded and competitive. Counted as models, it consolidates fast. Copilot runs on GPT; Perplexity and Cursor route between Claude, GPT and Gemini; Codex and Claude Code are OpenAI and Anthropic in developer clothing. Colour the map by who owns the underlying model and most of the apparent variety collapses onto four or five labs. The competition you can see sits on an oligopoly you cannot.

Sora was the video “leader”; until it was switched off

A map by use case implies the categories are stable. They are not. As ChatGPT and Gemini absorbed image, voice and video, standalone specialists lost web share, and Midjourney slipped from the top tier it once held. In April 2026 OpenAI shut down Sora as a standalone app and folded it into ChatGPT. The mechanism is bundling: when a general assistant swallows a capability, the dedicated tool’s separate existence becomes optional; and sometimes ends.

Methodology

Tools are grouped into use-case lanes editorially; many span more than one and are placed in their primary category. Bubble size is a banded tier of estimated monthly web visits (Similarweb and aggregated reporting), not an exact area: tier 1 is roughly 1 billion or more monthly visits, tier 2 about 250 million to 1 billion, tier 3 about 50 to 250 million, and tier 4 below that. Bands are used rather than precise areas because the figures are estimates spanning more than two orders of magnitude. Colour denotes the company that owns the underlying model. A dashed ring marks tools whose real usage web visits materially misstate, in either direction, because most of it occurs in native apps, via API, or embedded inside other software; the tooltip notes which. Sora is shown as closed following the April 2026 shutdown of its standalone app. Figures reflect April–May 2026 and will drift; this is a snapshot, not a live tracker. Sources: Similarweb, a16z Top 100 Gen AI Consumer Apps, and company announcements.