anthropic.comvsvercel.com
vercel.com beats anthropic.com by 4.0 composite points (74 vs 70, grades C vs C). The largest gaps are in structured data (67-point margin) and discoverability (9-point margin), which is where most of the score difference originates.
- · vercel.com leads on Structured data by 67 points.
- · anthropic.com leads on Performance by 14 points.
- · anthropic.com leads on Actionability by 11 points.
- · vercel.com leads on Discoverability by 9 points.
- · anthropic.com leads on Readability by 8 points.
Whether typed schema.org JSON-LD is present and valid for the page kind, such as Organization, WebSite, Product, Article, or FAQPage. Without typed data, agents infer everything from prose and hallucinate.
Whether the page renders fast enough that agents do not time out: TTFB under 1.5s, full render under 6s, payload under reasonable budgets. Most agents abort at the 6-second mark.
Whether agents can identify the primary offering and complete a task: a clear CTA, reachable pricing, contact route, and (where applicable) an MCP server card or agent-skills manifest.
Whether AI agents can find your canonical pages: robots.txt allowing GPTBot/ClaudeBot, /llms.txt at the apex, sitemap.xml, MCP server card. If agents cannot find you, every other score is academic.
Whether the page is parseable without executing JavaScript: server-rendered text, sane heading hierarchy, no JS-required content, no cookie-wall blocking the body. Agents read raw HTML before they render.
AgentSpeed compares two websites against the same public rubric, with checks across discoverability, readability, structured data, actionability, and performance. The comparison page is generated from real, completed public scans of both domains. There are no editorial picks, no curation, no model in the loop deciding which site looks “better”.
Each sub-score is the weighted average of the checks in that category that ran successfully on the scan. Composite is the weighted average of the five sub-scores. The grade is bound to a published band table on /rubric. Same rubric for both sides, same scan profile, same scoring code.
Categories are explained on the rubric page; the “Why these categories matter” cards above translate each one into the agent-readability problem it solves. The /learn hub has plain-English explainers if you want to dig deeper into any single signal.
Comparisons are domain-vs-domain. To compare a third domain or sweep a category, see /leaderboard. To improve your own score, run a free scan from the homepage and follow the per-finding fix pages at /fix.