OpenAI scored 10 F
All three OpenAI domains — openai.com, chatgpt.com, chat.openai.com — scored 10/100. Identical. The world's largest AI company. The company most associated with "agents using the web." Lower than YouTube (also 10).
I scored 27 major technology companies — foundation AI labs, consumer AI products, AI developer platforms, dev-tool gold standards, marketplaces — on how well their websites are structured for AI agents to discover and call deterministically.
All three OpenAI domains — openai.com, chatgpt.com, chat.openai.com — scored 10/100. Identical. The world's largest AI company. The company most associated with "agents using the web." Lower than YouTube (also 10).
claude.ai (product): 60. anthropic.com (corporate): 20. The team that ships the product invested in agent-discoverability. The corporate marketing team didn't. The 40-point gap is the gap between product and marketing, made visible.
The most-scraped website in the world. The site every consumer-buying agent in 2026 will want to talk to. A 5/100 for structured discoverability. The agents are coming whether Amazon is ready or not.
Founded 2023, by ex-Vercel folks. Built with agent-discoverability assumptions from day one — llms.txt at root, OpenAPI advertised, JSON-LD, structured forms. The only company in the dataset to hit the ceiling. Proves the bar isn't aspirational.
HuggingFace: 45. Replicate: 25. Kaggle: 20. Runway: 25. The platforms where AI engineers literally build agents aren't structured for those agents to use. The supply side of the AI economy is invisible to the demand side it's creating.
The accelerator that's funded most of the AI-agent startups of the last three years. Their own site, where founders apply, where partners read pitches, where the next agent-native unicorn gets seed funding — invisible to the agents those startups are building.
Worst to best. Tap any row to score that domain yourself.
| Company | Category | Score | /100 |
|---|
No category cleared 50.
Each site was scored against a deterministic 7-signal rubric using the free agent-readiness scanner at solsticestudio.ai/readiness. Each signal is binary — full points or zero. Weights sum to 100, so the score is the percentage.
| Signal | Weight | What it checks |
|---|---|---|
| MCP server | 35 | /.well-known/mcp/server, /.well-known/mcp.json, or /mcp returns 200 |
| OpenAPI spec | 25 | /openapi.json, /swagger.json, or one of six common locations contains a valid spec |
| llms.txt | 15 | /llms.txt exists with content |
| Sitemap | 10 | /sitemap.xml or /sitemap_index.xml exists |
| robots.txt | 5 | /robots.txt exists |
| JSON-LD | 5 | Homepage has application/ld+json structured data |
| Semantic forms | 5 | Forms (if any) use labels and named inputs |
Grade scale: A (90+) · B (75-89) · C (60-74) · D (40-59) · F (under 40)
What it doesn't measure: content quality, API design quality, response time, security posture, or whether your endpoints actually work. It measures discoverability — can an agent find what it needs to call you without scraping HTML? A site can score 100 and ship a buggy API. We measure the front door, not the rooms inside.
Reproducibility: the rubric and code are deterministic. Run any URL through solsticestudio.ai/readiness and you'll get the same score in 5 seconds.
Dates: all scores in this dataset were collected on 2026-05-17.
Disclosure: Solstice AI Studio operates Arachne, a hosted MCP gateway and Shadow API compiler. Solstice's own site scores 100/100 (we publish llms.txt, OpenAPI, JSON-LD, and serve a discovery endpoint at /.well-known/mcp/server). Eating our own dog food was a precondition for publishing this.
Free, instant, anonymous. Five seconds. The scanner runs the same 7 signals against any URL.