The documentation lookup tax is collapsing.
Mintlify’s co-founder, Han Wang, says nearly half of docs traffic across Mintlify-powered sites is now AI.
Some of that will be scraping. Much of it will be agents doing work.
I noticed this over Christmas when I rebuilt an app that’d been on my to-do list for months: user authentication and management, payments and webhooks plus an AI integration. I pointed an agent at the Stripe, Clerk.com and OpenAI docs and let it ingest. I mostly stayed in ‘tell me what to do next’ mode, stepping in only for keys, logins and a couple of judgement calls.
I didn’t read the documentation. The agent did.
If that behaviour scales, documentation stops being a learning resource and becomes a machine-facing surface. Documentation with clean structure, examples as few-shot prompts and predictable edge cases gets integrated first. The rest becomes friction.
This is a quiet advantage. If an agent can do the reading and the first integration pass, you keep your attention for the few steps that still need judgement and risk calls.
The readers changed. The advantage goes to whoever noticed.
