See how agents handle your site
How it works
From a first score to a fix you can prove, in four steps.
- 01
Check your domain
Enter it above. A domain we already track opens its report straight away. A new one gets its first report built live: quick checks of what agents can reach, then Claude, ChatGPT and Gemini trying real tasks in a real browser.
- 02
Read the report
One overall score and its rank, a score per category and per agent, and the friction that stopped them, each backed by the run it came from. The methodology explains every check.
- 03
Fix what blocks agents
Every report carries a fix prompt: the failing checks and frictions, written up for your coding agent. Paste it in, ship the fix, and the next run shows whether it worked.
- 04
Claim it
Sign in with a work email on the domain to claim it. Owners replay every agent session in full, edit how the domain is presented and request a re-run whenever they ship. Personal inboxes like Gmail can't prove ownership.
Go past the public tasks
The Index runs the same tasks for every site in a sector. Stunt Double puts agents and AI personas through your own flows, signed in, with the evidence to show what broke.
More for owners
Show your score
Embed a live badge of your score on your site or README. It follows your score: /badge/yourdomain.com.svg
Use it from your agent
Read reports and fix prompts from Claude, Cursor or any MCP client. See Developers.
Leave the Index
Disallow StuntDoubleIndex in robots.txt and we stop visiting on the next run. Questions go to Contact.