> ## Documentation Index > Fetch the complete documentation index at: https://docs.37audits.com/llms.txt > Use this file to discover all available pages before exploring further. # Markdown Content Auditor > Validates that the site serves HTML to browsers and markdown to AI clients via Accept content negotiation **AI Readiness Auditor** — validates that the site serves HTML to browsers and markdown to AI clients via `Accept` content negotiation. AI agents that ask for markdown get clean markdown source instead of a full HTML page. Browsers keep receiving HTML by default; markdown is only served when explicitly requested. Large language models and AI agents consume markdown far more efficiently than rendered HTML: it is smaller, keeps the document structure, and drops navigation, scripts, and styling noise. A growing convention is to serve the same URL as markdown when the client sends `Accept: text/markdown`, while browsers — which send `Accept: text/html` — keep receiving the normal HTML page. This auditor requests the audited URL both ways and verifies each response's `Content-Type` and body. ## How it works The auditor issues two GET requests to the audited URL (following redirects, with HTTP Basic Auth when provided): 1. **Browser request** — sends `Accept: text/html` and expects a `Content-Type` of `text/html` or `application/xhtml+xml`. 2. **AI-client request** — sends `Accept: text/markdown` and expects a `Content-Type` of `text/markdown` or `text/x-markdown`. 3. **Body sanity check** — when the markdown response declares a markdown media type, the body must not start as a full HTML document (``37A-MarkdownContentAuditor-400` | `FAIL` | The request with `Accept: text/html` returns a missing or blank `Content-Type` header. | Send `Content-Type: text/html` for `{url}` when the client asks for HTML. | | `37A-MarkdownContentAuditor-401` | `FAIL` | The request with `Accept: text/html` is answered with a markdown media type (`text/markdown` or `text/x-markdown`). | Serve HTML to browsers: set Content-Type to `text/html` on `{url}` for `Accept: text/html` instead of `{contentType}`. | | `37A-MarkdownContentAuditor-402` | `FAIL` | The request with `Accept: text/html` is answered with a Content-Type that is present, not markdown, and not `text/html` or `application/xhtml+xml`. | Serve `{url}` as `text/html` when Accept is `text/html`; current Content-Type is `{contentType}`. | | `37A-MarkdownContentAuditor-403` | `FAIL` | The request with `Accept: text/markdown` returns a missing or blank `Content-Type` header. | Send `Content-Type: text/markdown` for `{url}` when the client sends `Accept: text/markdown`. | | `37A-MarkdownContentAuditor-404` | `FAIL` | The request with `Accept: text/markdown` is still answered with an HTML media type (`text/html` or `application/xhtml+xml`), meaning the server does not negotiate markdown. | Negotiate markdown: serve `{url}` as `text/markdown` (or `text/x-markdown`) when Accept is `text/markdown`; current Content-Type is `{contentType}`. | | `37A-MarkdownContentAuditor-405` | `FAIL` | The request with `Accept: text/markdown` is answered with a Content-Type that is present, not HTML, and not `text/markdown` or `text/x-markdown`. | Serve `{url}` as `text/markdown` when Accept is `text/markdown`; current Content-Type is `{contentType}`. | | `37A-MarkdownContentAuditor-406` | `FAIL` | The markdown response declares a markdown media type but the body starts as a full HTML document (``37A-MarkdownContentAuditor-200` | `SUCCESS` | Every rule above passed; one `SUCCESS` check is added with the message `"{url}" serves HTML ({htmlContentType}) by default and markdown ({markdownContentType}) when Accept is text/markdown.` | — | ## Output documentation | Status | Description | Test logic | | ----------- | ----------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **SUCCESS** | The URL serves HTML by default and markdown through content negotiation | `Accept: text/html` returns `text/html` or `application/xhtml+xml`; `Accept: text/markdown` returns `text/markdown` or `text/x-markdown` with a body that does not start as an HTML document. Both requests return HTTP 200. | | **FAIL** | Content negotiation is missing, inverted, or mislabeled | Missing `Content-Type` on either request (`400`, `403`); markdown served to browsers (`401`); non-HTML default (`402`); HTML still served when markdown was requested (`404`); a non-markdown type on the markdown request (`405`); a markdown-labeled response whose body is HTML (`406`); or a non-200 status on either request. No rule in this auditor emits `WARNING`. | | **ERROR** | The URL could not be fetched | The host is unreachable or the request fails before a response is received. | ## Risks and considerations ### AI readiness * Agents and LLM crawlers that request `text/markdown` and receive a full HTML page must parse markup, scripts, and navigation chrome, wasting context window and degrading answer quality. * Serving markdown through `Accept` negotiation keeps a single canonical URL per page — no separate `.md` URLs to publish, crawl, or keep in sync. ### Browser compatibility * Content negotiation must never leak markdown to browsers: a markdown response to `Accept: text/html` renders as plain text and breaks the page for every visitor. * Responses that vary by `Accept` should include `Vary: Accept` so caches and CDNs do not serve the markdown variant to browsers or vice versa. ### Correct labeling * A missing `Content-Type` forces clients to sniff the payload, which is unreliable and a known security concern. * Declaring `text/markdown` while returning an HTML body is worse than not negotiating at all: clients trust the label and feed HTML into markdown pipelines.