AI coding agents: a growing market

These searches belong together because they are one decision: which AI tool a developer hands their code to. Some live in the editor and suggest and rewrite code (Cursor, Copilot), others work in the terminal as agents that read a whole codebase, edit files and run commands on their own (Claude Code, Codex). People search them by name, as "best AI coding assistant", and increasingly for open-source versions they can run without a subscription.

The size sits in the named tools. Claude Code draws 550,000 searches a month and Cursor AI 90,500. Read the growth by the last 12 months against the 12 before: Claude Code up 953% and Cursor AI up 112%. The generic searches are smaller but rising with them: AI coding assistant at 18,100 a month (up 386% on the same measure) and AI coding tools at 9,900 (up 347%). All four members with a year of history are rising, and three of four have kept rising two years running. Both generic searches had a two- or three-week burst in late August that came from somewhere else; growth here is read with those weeks left out, as direction.

This is not a search artefact. The library's other channels see the same thing beyond search, which is the bar for a young category. It is young, though: most of its members barely existed two years ago, and the leaders change every few months as new models ship.

The buyers are developers and engineering leads, and money follows them: advertisers pay a median $13.72 a click, the highest of any meta trend here, and median keyword difficulty is 36 of 100, so ranking is harder than in consumer niches. A founder should look past the agents themselves, which are a race between model companies, to what they create: code review and testing for AI-written code, cost and usage control for teams, setup guides, prompts and workflows, open-source alternatives, and teaching people who are not engineers to build with them.