# Stunt Double Index

> The Stunt Double Index is an independent benchmark measuring how real AI agents experience the world's websites across discovery, information retrieval, task completion, delegated access and support.

The index scores domains across eight categories, using agents built on Claude, ChatGPT, Gemini and more. Scores are recomputed weekly. Use this file to understand what's available before crawling.

## Core pages

- [Home / Leaderboard](https://index.stuntdouble.io/): overall agent-experience ranking.
- [Methodology](https://index.stuntdouble.io/methodology): how every score is produced.
- [Categories](https://index.stuntdouble.io/categories): the eight scoring categories.
- [Compare](https://index.stuntdouble.io/compare): side-by-side head-to-heads.
- [Reports](https://index.stuntdouble.io/reports): frozen, citable snapshots of the Index.
- [About the index](https://index.stuntdouble.io/about): project background.
- [Developers](https://index.stuntdouble.io/developers): the MCP connector (no account needed), its Index tools and the machine-readable files.
- [Contact](https://index.stuntdouble.io/contact): email support@stuntdouble.io about a score, claiming or opting out a domain.

## What this site is

A free, public benchmark published by Stunt Double (Turnout Labs Ltd., London), in the same space as web performance and accessibility scorecards, applied to AI agents. It sells nothing: there is no shop, cart, checkout, shipping, returns or refunds. Product pages for other companies (for example /d/homedepot.com) are scorecards about those sites, not links to buy from them.

## API and MCP

- Public MCP server (streamable HTTP, no account or sign-in): https://app.stuntdouble.io/api/mcp/index. Read-only Index tools: search_index_domains, get_index_report, list_index_sessions.
- Full Stunt Double MCP server (streamable HTTP, OAuth 2.1): https://app.stuntdouble.io/api/mcp. The same Index tools (scope mcp:read) plus the product's, request_index_rerun for a domain's owner, and request_index_score to add a site the Index does not track yet.
- [auth.md](https://index.stuntdouble.io/auth.md): how an agent registers, gets a token and uses it.
- Server card: https://index.stuntdouble.io/.well-known/mcp/server-card.json
- Per-domain fix prompt (Markdown, no account): https://index.stuntdouble.io/d/{domain}/fix-prompt.md

## How scores are made

1. Quick checks: on every score request, a handful of HTTP requests test whether agents can reach the site (robots.txt rules for each provider, bot challenges), find their way around (sitemap, pricing, signup, cart and contact pages), read what it offers (llms.txt, markdown negotiation, structured data), connect and sign in (MCP server card, OAuth and OIDC discovery per RFC 8414 and RFC 9728, A2A agent card) and buy (Universal Commerce Protocol discovery). Each check keeps its pass or fail evidence.
2. Agent sessions: every week, agents open the site in a hosted browser and work through the sector's tasks the way a customer would (find a product and reach checkout, read the refund policy, locate API docs). Every step is recorded, and a separate grader marks each session against a fixed rubric.
3. Combined score: each category blends 60% agent sessions with 40% quick checks where both exist; a site with no sessions yet shows its quick checks alone. The overall score is the weighted average of the eight categories.
4. Not scored: a site that turns the quick checks away (homepage 401, 403 or 429) or does not answer, and that no agent session has reached, is marked not scored. It keeps its page but is left out of rankings and averages.

Agent providers: Claude (Anthropic), ChatGPT Agent (OpenAI), Gemini (Google), Kimi (Moonshot AI, on Cloudflare), Llama (Meta, on Cloudflare), Grok (xAI, on Cloudflare). Each category averages their results.

Our agents identify themselves with a stable user agent, honor opt-outs in robots.txt and headers, never complete payments or enter real registration details, keep no personal information and are rate limited per site.

## Scoring categories

- **Brand awareness** (10% of the overall score): Can agents recognize you exist? Measures how often and how accurately leading agents mention your brand in unprompted research and comparison tasks.
- **Discovery** (15% of the overall score): Will they pick you? How well agents discover, shortlist and recommend your product when a user asks for a solution in your category.
- **Information retrieval** (15% of the overall score): Can they read your site? Quality of structured data, semantics and navigability for agents extracting pricing, specs and policies.
- **Market ranking** (10% of the overall score): Where do you sit in the lineup? Your relative placement vs. competitors when agents produce ranked lists for buyers in your market.
- **Accuracy** (15% of the overall score): Do they tell the truth about you? How often agents report correct pricing, features, availability and policies, and how often they hallucinate.
- **Task completion** (20% of the overall score): Can an agent complete a task on behalf of a user? Whether an agent can carry out a delegated interaction end-to-end: a primary call-to-action is exposed, the destination is reachable without a login wall, no bot-challenge blocks the first step, and the service signals support for agent-friendly auth.
- **Delegated access** (10% of the overall score): Do you let agents in, safely? Availability of scoped auth (OIDC, API keys, MCP), rate limits, and tooling for delegated agent sessions.
- **Contact & communication** (5% of the overall score): Can an agent reach a human? Whether an agent acting on a user’s behalf can find a way to contact or communicate with the business: contact page, support email, live chat, phone, help hub, or a social profile it can DM.

## Reports

Snapshots of the Index taken at a fixed point in time. Their numbers never change, so cite a report rather than a live scorecard.

- [The 10 best performing websites for AI agents October 2026](https://index.stuntdouble.io/reports/10-best-websites-oct-26)
- [How ready are the top 50 websites for agents?](https://index.stuntdouble.io/reports/how-ready-are-the-top-50-websites-for-agents)

## Top tracked domains

- [Cloudflare (cloudflare.com)](https://index.stuntdouble.io/d/cloudflare.com): score 91.7 (professional-services).
- [Supabase (supabase.com)](https://index.stuntdouble.io/d/supabase.com): score 91.7 (saas).
- [ElevenLabs (elevenlabs.io)](https://index.stuntdouble.io/d/elevenlabs.io): score 91.1 (saas).
- [Browserbase (browserbase.com)](https://index.stuntdouble.io/d/browserbase.com): score 90.5 (saas).
- [Sentry (sentry.io)](https://index.stuntdouble.io/d/sentry.io): score 90.5 (saas).
- [DeviantArt (deviantart.com)](https://index.stuntdouble.io/d/deviantart.com): score 90.4 (social).
- [Atlassian (atlassian.com)](https://index.stuntdouble.io/d/atlassian.com): score 90.1 (professional-services).
- [Netlify (netlify.com)](https://index.stuntdouble.io/d/netlify.com): score 90.1 (saas).
- [Vercel (vercel.com)](https://index.stuntdouble.io/d/vercel.com): score 90.0 (saas).
- [WordPress.org (wordpress.org)](https://index.stuntdouble.io/d/wordpress.org): score 90.0 (developer-tools).
- [GitBook (gitbook.com)](https://index.stuntdouble.io/d/gitbook.com): score 89.7 (saas).
- [Cal.com (cal.com)](https://index.stuntdouble.io/d/cal.com): score 89.6 (saas).
- [Descript (descript.com)](https://index.stuntdouble.io/d/descript.com): score 89.6 (saas).
- [Intercom (intercom.com)](https://index.stuntdouble.io/d/intercom.com): score 89.6 (saas).
- [Sketch (sketch.com)](https://index.stuntdouble.io/d/sketch.com): score 89.3 (saas).
- [Figma (figma.com)](https://index.stuntdouble.io/d/figma.com): score 89.2 (utilities).
- [Fly.io (fly.io)](https://index.stuntdouble.io/d/fly.io): score 89.2 (saas).
- [Render (render.com)](https://index.stuntdouble.io/d/render.com): score 89.2 (saas).
- [Resend (resend.com)](https://index.stuntdouble.io/d/resend.com): score 89.2 (saas).
- [Parlour (heyparlour.app)](https://index.stuntdouble.io/d/heyparlour.app): score 89.1 (developer-tools).
- [Stripe (stripe.com)](https://index.stuntdouble.io/d/stripe.com): score 89.0 (saas).
- [Stunt Double (stuntdouble.io)](https://index.stuntdouble.io/d/stuntdouble.io): score 89.0 (saas).
- [Jotform (jotform.com)](https://index.stuntdouble.io/d/jotform.com): score 88.9 (saas).
- [Typeform (typeform.com)](https://index.stuntdouble.io/d/typeform.com): score 88.9 (saas).
- [Amplitude (amplitude.com)](https://index.stuntdouble.io/d/amplitude.com): score 88.7 (saas).

## Optional

- [Privacy policy](https://www.stuntdouble.io/privacy)
- [Apply as an agent provider](https://index.stuntdouble.io/providers/apply)
- [Stunt Double marketing site](https://www.stuntdouble.io)
- [How to test your product for AI agent users](https://www.stuntdouble.io/blog/how-to-test-your-product-for-ai-agent-users): a checklist for the failures the Index measures.