Lighthouse Agentic Browsing Audit: Why Your llms.txt Fails Without Markdown Links

Lighthouse Agentic Browsing: A New Audit Category with a Hidden Trap

Chrome's Lighthouse 13.3.0 introduced the Agentic Browsing category, a set of audits designed to evaluate how well a website serves AI agents. Among six default audits, the llms-txt check is the most controversial. It failed on nohacks.co with the error: 'File does not appear to contain any links.' The file actually contained dozens of links—but not in markdown syntax. The audit treats .txt files as markdown documents, demanding [text](url) format. This is a structural mismatch: the file extension and MIME type say plain text, but the parser expects markdown.

Why This Matters for Your AI Readiness

The llms.txt specification explicitly defines the format as a markdown document. Lighthouse enforces this strictly. If your file uses plain-text descriptions—even if every link works—you fail. The fix is mechanical: wrap each link in markdown bracket-paren syntax. But this creates a perverse incentive: auto-generated files from plugins like AIOSEO (used by 3M+ sites) pass by default, while hand-curated, accurate files fail. The audit measures parseability, not quality.

Strategic Consequences: Who Gains, Who Loses

Winners: AIOSEO and similar plugins that generate markdown links by default. Their users get a compliance advantage without effort. Also, sites that adopt proper markdown llms.txt gain better AI agent discoverability, potentially boosting traffic from agent-driven browsers.

Losers: Sites with carefully written plain-text llms.txt files. They fail the audit despite having superior content. Developers unfamiliar with markdown syntax face an unnecessary barrier. Alternative browser vendors (Firefox, Safari) may be pressured to adopt similar audits, creating a Chrome-centric standard.

The Bigger Picture: Machine-First Architecture

This audit is a narrow check within a broader shift toward machine-first architecture. Data models before page layouts, rendering independence, content that doesn't depend on client-side JavaScript. The llms-txt audit checks whether your file is mechanically parseable. It does not check whether it describes your website honestly. That responsibility remains yours.

What You Should Do Now

Run the Lighthouse Agentic Browsing audit on your site. If the llms-txt check fails, convert your links to markdown syntax. But don't stop there. Audit the quality of your llms.txt: does it accurately represent your site's content? Is it comprehensive? The audit passes, but your AI readiness depends on substance, not just syntax.

Source: Search Engine Journal