Early ASO Examples
On 2026-06-11 we ran the automated ASO scanner against the apex domains of ten companies everyone knows. Every site allows AI crawlers. Almost none of them tell agents what they are, what they sell, what it costs, or how to act. The agent-readable web is still mostly unbuilt -- even at the top.
Leaderboard
| # | Site | Score | Level | Discover /20 |
Identity /20 |
Trust /15 |
Commerce /15 |
Reputation /15 |
Memory /15 |
|---|---|---|---|---|---|---|---|---|---|
| 1 | cloudflare.com | 66 | ASO-3 Invocable | 20 | 20 | 8 | 2.5 | 0 | 15 |
| 2 | vercel.com | 46 | ASO-2 Understandable | 12 | 5 | 15 | 2.5 | 4 | 7.5 |
| 3 | stripe.com | 39 | ASO-2 Understandable | 12 | 5 | 9.5 | 2.5 | 2 | 7.5 |
| 3 | shopify.com | 39 | ASO-2 Understandable | 16 | 5 | 8 | 2.5 | 0 | 7.5 |
| 5 | github.com | 33 | ASO-2 Understandable | 8 | 2.5 | 11 | 2.5 | 4 | 5 |
| 6 | huggingface.co | 29 | ASO-1 Discoverable | 8 | 2.5 | 13.5 | 0 | 2 | 2.5 |
| 7 | anthropic.com | 25 | ASO-1 Discoverable | 8 | 2.5 | 9.5 | 0 | 2 | 2.5 |
| 8 | nike.com | 20 | ASO-1 Discoverable | 8 | 5 | 4 | 0 | 0 | 2.5 |
| 9 | openai.com | 19 | ASO-1 Discoverable | 8 | 0 | 8 | 0 | 0 | 2.5 |
| 10 | amazon.com | 15 | ASO-1 Discoverable | 4 | 2.5 | 8 | 0 | 0 | 0 |
Average score: 33.1. Median: 31. ASO-3 Invocable starts at 50. One site out of ten reached that level.
What the scans show
All ten sites allow AI crawlers in robots.txt. But access is not communication: only one site publishes an agent identity manifest, and none publish an A2A agent card or a payment manifest at the paths agents check.
cloudflare.com scored 66 (ASO-3 Invocable): perfect discoverability and identity, llms.txt plus llms-full.txt, an agent.json manifest ("Cloudflare Site Agent"), markdown negotiation via /index.md, and an OpenAPI spec. It is the only site in the batch that reached ASO-3 Invocable.
Cloudflare, Vercel, Stripe, Shopify, and GitHub publish llms.txt. Hugging Face, Anthropic, Nike, and Amazon return 404. OpenAI returns 403 -- the file path is actively blocked on the apex domain.
No site in the batch exposes machine-readable pricing, a payment manifest, or x402 support at the apex domain. The highest commerce pillar score was 2.5 out of 15. The companies that process the world's payments are not yet payable by agents.
openai.com scored 19 and anthropic.com scored 25 -- both ASO-1 Discoverable, behind Nike-adjacent retail basics on identity signals. Building frontier models and publishing agent-readable signals are evidently different muscles.
Method and limits
These are point-in-time automated scans of public signals at each apex domain, run on 2026-06-11 with the beta scanner. Subdomains, docs sites, and API hosts were not scanned, so a company with a strong developer subdomain can outscore its apex result. The reputation pillar is an emerging category that automation can only partially verify. A score is not an endorsement or a criticism -- it is a snapshot of what an autonomous agent can find, parse, and act on at that origin today. Any site here could change its score overnight by publishing the missing files.
Reproduce it
# Claude Code
claude mcp add aso -- npx -y @forgemeshlabs/aso-audit-mcp
# Then ask your agent
"Scan stripe.com for agent readiness"
"Scan my own site and give me a fix plan"Your turn