peptidebenchmark.com

8 things to fix

Each one has a prompt written for a coding agent. Paste it into Claude Code, Cursor or whatever you build with. Or take all of them in one go.

• GPTBot is blocked

  Missing

  robots.txt disallows the root. OpenAI will not train on this site. ChatGPT search is unaffected, that is OAI-SearchBot.

  Remove the Disallow: / that applies to GPTBot, or keep it and accept the line above as the price.

  Copy prompt

• ClaudeBot is blocked

  Missing

  robots.txt disallows the root. Anthropic will not train on this site. Claude search is unaffected, that is Claude-SearchBot.

  Remove the Disallow: / that applies to ClaudeBot, or keep it and accept the line above as the price.

  Copy prompt

• Google-Extended is blocked

  Missing

  robots.txt disallows the root. Gemini will not train on or ground in this site. Google AI Overviews are unaffected, they use Googlebot.

  Remove the Disallow: / that applies to Google-Extended, or keep it and accept the line above as the price.

  Copy prompt

• No Link response headers

  Missing

  RFC 8288 Link headers let an agent find related resources without parsing the HTML.

  Send a Link header for the things an agent should find without parsing HTML: canonical, alternate, and your llms.txt.

  Copy prompt

• Accept: text/markdown returns HTML

  Missing

  Agents that ask for markdown get HTML they have to strip, which loses structure.

  Serve a markdown twin of each page when the request asks for text/markdown, or publish page.md beside page.

  Copy prompt

• API errors come back as HTML

  Missing

  /api/seobuilder-agent-readiness-probe-404 answered 200 as text/html. An agent calling your API gets a page where it expected an error object.

  Make every /api/ route answer with JSON on failure too: a real status, and a body with a message field. An HTML error page is a parse error to a caller.

  Copy prompt

• Missing pages answer 200

  Missing

  /seobuilder-agent-readiness-probe-404 answered 200. A page that does not exist must not answer 200, or every typo becomes a page an agent believes in.

  Return 404 or 410 for paths that do not exist, and put "404" or "Not found" in the title and the H1.

  Copy prompt

• Organization schema is missing contactPoint and address

  Partial

  An agent can see who you are but not how to reach you, which is the half that decides whether it recommends you.

  Add Organization JSON-LD with name, url, logo, contactPoint and address. The contact fields are the ones agents quote.

  Copy prompt

Already passing (17)

• robots.txt is published
• OAI-SearchBot can read the site
• Claude-SearchBot can read the site
• PerplexityBot can read the site
• The sitemap is valid XML
• robots.txt points at your sitemap
• /llms.txt is published
• llms.txt says what the site is for
• /.well-known/api-catalog is published
• An MCP server card is published
• The homepage has a title
• The homepage has a meta description
• Valid JSON-LD structured data
• One clear H1
• The page reads without JavaScript
• 11.94% of the homepage is readable text
• About, contact and privacy all have real content

Not your site?

The same check runs on anything. It fetches the page once and reads what an assistant would read, which is the server’s HTML, with no JavaScript and no scrolling.

Rank and bid from outbid.lol on 2026-08-22. Not affiliated with them. Want this page taken down? Ask us and it goes, no argument.