openai.com
Publicly-generated scan via agentspeed.com. Not affiliated with openai.com.
openai.com is not yet ready for AI agents (composite 40, grade F). Strongest sub-score: performance (100). Weakest: readability (22). The most actionable gaps detected: missing JSON-LD, no markdown content negotiation, and no published MCP server card.
Based on 30 checks · 20 deterministic · 9 heuristic · 1 model-assisted. see /rubric.
How openai.com declares its policy toward AI agents and crawlers. These rows are a presentation view over checks already counted under Discoverability and Readability, so they don’t add to the composite a second time.
What an AI agent fetching this site today is likely to understand. Each row is backed by a real signal that ran on this scan, or labels itself “not detected”. We never invent values.
robots.txt allows the major AI agents (GPTBot, ClaudeBot, PerplexityBot, …).
For a live UA-by-UA fetch matrix (status, redirects, content similarity, cloaking detection across GPTBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Google-Extended, Googlebot, Bingbot, CCBot, and a default browser), run the matrix tool against this domain.
No findings on the latest scan. Every check passed.
Running AI agents of your own? AgentSpeed also monitors them in production — every run, its latency, cost, and failures, with alerts when something breaks. Free for 10,000 events a month, no credit card.
first scan. History starts after the next scan.
No critical findings on the latest scan.
Sorted failures first. Every row was produced by the scan on 2026-07-25 05:42 UTC. Each check is defined in /rubric §6.
This scan is content-addressable. Download the JSON and confirm that the composite matches score(rubric, checks) for the referenced rubric. Re-running the command below executes against the current rubric; scores agree within ±1 point unless the target site changed.
curl -sS -X POST https://agentspeed.com/api/scans \
-H 'content-type: application/json' \
-d '{"url": "https://openai.com"}'What AgentSpeed scanned
Every scan is based on public website behaviour, fetched from a public client, no auth, no plugins, no private data. The nine signals below are what AgentSpeed reads about your site.
Show your AgentSpeed score in your README, footer, or sales deck. Four variants below; each is a static SVG served from /badge/<your-domain> with no JS, no tracking, no API calls. Documentation: /docs/badge.
Default. Brand mark + grade + score on a dark base. Pairs well with most footers and READMEs.
HTML snippet
<a href="https://agentspeed.com/score/openai.com" target="_blank" rel="noopener"> <img src="https://agentspeed.com/badge/openai.com?variant=dark" alt="AgentSpeed grade F for openai.com" width="220" height="32" /> </a>
Markdown snippet
[](https://agentspeed.com/score/openai.com)
Same dimensions as the dark variant; light background with dark text. For light-mode docs and pages.
HTML snippet
<a href="https://agentspeed.com/score/openai.com" target="_blank" rel="noopener"> <img src="https://agentspeed.com/badge/openai.com?variant=light" alt="AgentSpeed grade F for openai.com" width="220" height="32" /> </a>
Markdown snippet
[](https://agentspeed.com/score/openai.com)
Brand mark + grade only. Use when you have horizontal space constraints.
HTML snippet
<a href="https://agentspeed.com/score/openai.com" target="_blank" rel="noopener"> <img src="https://agentspeed.com/badge/openai.com?variant=compact" alt="AgentSpeed grade F for openai.com" width="130" height="32" /> </a>
Markdown snippet
[](https://agentspeed.com/score/openai.com)
Just the grade chip. Smallest variant, fits inline next to a heading or metric.
HTML snippet
<a href="https://agentspeed.com/score/openai.com" target="_blank" rel="noopener"> <img src="https://agentspeed.com/badge/openai.com?variant=score-only" alt="AgentSpeed grade F for openai.com" width="60" height="32" /> </a>
Markdown snippet
[](https://agentspeed.com/score/openai.com)
What does this AgentSpeed score mean?expand ›
The composite is a 0–100 weighted average of five sub-scores measuring how well a website performs for AI agents (ChatGPT, Claude, Perplexity, Operator, etc.). The full rubric (every check, weight, and grade band) is published at /rubric.
Is this an official score from the website owner?expand ›
No. AgentSpeed scores are publicly generated against openai.com's public surfaces (robots.txt, sitemap, the landing URL, /llms.txt, /.well-known/* discovery endpoints). AgentSpeed is not affiliated with or endorsed by openai.com. The score reflects what we observed at scan time, nothing more.
How often is this site scanned?expand ›
Public scans happen when someone submits the URL on the homepage. Once a domain is claimed in the (forthcoming) monitoring tier, the site is rescanned daily and a history is kept. For now, every public score reflects the most recent submission.
How can this score be improved?expand ›
Each failing check on the page expands to a remediation step with the exact change to make. The "Top fixes" section ranks them by score impact. Most domains move 5–10 points by adding an llms.txt, allowing well-known agent user-agents in robots.txt, and emitting JSON-LD that matches the page purpose.
What rubric version was used?expand ›
This scan was computed against rubric r2026.07.2. The rubric is versioned and immutable: historical scans always render against the rubric they were computed on, even after a new version ships. The current rubric and its full changelog are at /rubric.