BridgeToAgent
Get the kit — $49
All posts
Announcement7 min read

Chrome just made Agentic Browsing default in Lighthouse — here's how to pass before your competitors notice

Lighthouse 13.3.0 shipped May 7 with the Agentic Browsing category in the default config. PageSpeed Insights inherits within two weeks. Every site owner running a Lighthouse report now gets agent-readiness signals whether they asked for them or not. Here's what changed, why scoring intentionally isn't a single number, and the fastest fix paths for the audits most sites fail.

BridgeToAgentEditorial team

Chrome just made Agentic Browsing default in Lighthouse

On May 7, 2026, Google quietly shipped Lighthouse 13.3.0. One line in the changelog flipped a switch the rest of the SEO and dev tools industry has been waiting on:

The Agentic Browsing category is now part of the default config.

Two weeks later — that's right now — PageSpeed Insights inherits it. Chrome 150 DevTools follows on the next stable release. The category that was behind an experimental feature-flag yesterday is the audit every site owner who clicks "Run Lighthouse" sees today.

If you've been treating agent-readiness as a "we'll get to it" item, you have about 14 days before your client, your boss, or your audit-running customer asks why your Agentic Browsing signals look bad. We've shipped a free llms.txt validator so the most common failure mode is one paste away from a fix.

This post is the news + the tool + the honest read on what changed.

The 30-second version

What changed. Lighthouse 13.3.0 (May 7, 2026) moved the Agentic Browsing category from experimental into the default config. PageSpeed Insights gets it within two weeks of release; Chrome 150 DevTools soon after.

Scoring changed too. Unlike the other Lighthouse categories, Agentic Browsing doesn't ship a single weighted 0-100 score. The spec team explicitly chose actionable signals over a single number — "the standards for the agentic web are still emerging, the current focus is to gather data and provide actionable signals rather than a definitive ranking."

What we shipped today. A free llms.txt validator at bridgetoagent.com/tools/llms-txt-validator. Paste a file or enter a URL; get per-line parse errors, link-rot detection, and a pass/fail aligned to the Lighthouse audit. Also ships as @bridgetoagent-com/llms-txt-validator on npm.

What to do. Run Lighthouse on your homepage now, before your customers do. The two audits most sites fail are llms-txt-present and agents-json-present. The validator handles the first; the $49 kit handles both plus the other four.

What Lighthouse 13.3.0 actually changes

Two changes worth understanding separately.

Default-config means surface area

Before 13.3.0, Agentic Browsing was a category you had to deliberately enable. Most teams didn't. Most third-party Lighthouse runners (PageSpeed Insights, web.dev/measure, the dozens of SEO-audit SaaS tools that wrap Lighthouse) didn't include it because it wasn't on by default.

That's the change. Now any Lighthouse run — from any of those surfaces — returns Agentic Browsing audits alongside Performance, Accessibility, Best Practices, and SEO. The signal volume goes from "developers who opted in" to "anyone who runs an audit." The audit was always there. The visibility was the missing piece. Visibility just shipped.

Scoring is intentionally not a single number

The second change is more interesting and gets less press. Per the Chrome team's documentation: "the Agentic Browsing category does not have a weighted average score from 0 to 100. Because the standards for the agentic web are still emerging, the current focus is to gather data and provide actionable signals rather than a definitive ranking."

This is a smarter choice than it looks. A weighted score for Performance makes sense — the inputs (LCP, CLS, INP) are stable, the weights are calibrated against measurable user outcomes, the number is comparable across sites. The inputs for Agentic Browsing aren't there yet. WebMCP annotations are emerging. llms.txt adoption sits at about 10% across the SE Ranking 300k-domain sample published in early 2026 — large enough to matter, too uneven to weight against a baseline.

Practically: when you run Lighthouse 13.3.0 on your site, you'll see per-audit pass / fail / warning markers. No "Agentic Browsing: 67/100" headline number. That's by design, not by oversight, and it means our previous claims about target scores need to be read as guidance rather than as a normative number.

What the audits actually check

For completeness, the audits in the default config:

  • llms-txt-present — is /llms.txt reachable as plain text?
  • llms-txt-well-formed — does it parse against the llmstxt.org reference?
  • agents-json-present — is /agents.json reachable as valid JSON?
  • agents-json-actions-typed — does every declared action have typed parameters?
  • sitemap-discoverable — is /sitemap.xml valid and referenced from robots.txt?
  • agent-runbook-present — is /agent-instructions.md reachable?
  • auto-discovery-links — does the homepage <head> contain <link rel="alternate"> references to the three kit files?
  • schema-org-density — Schema.org JSON-LD blocks above minimum density on homepage and primary product / article pages?
  • webmcp-annotations — emerging; low weight, expected fail on most sites in 2026

The full spec is at developer.chrome.com/docs/lighthouse/agentic-browsing/scoring.

Why we shipped a validator now

The single most common failure when sites do attempt llms.txt is malformed files. Free generators ship something; the something doesn't always parse. The Lighthouse audit llms-txt-well-formed catches this, but tells you it failed without telling you where or why.

The validator at bridgetoagent.com/tools/llms-txt-validator closes that gap. Paste raw text, upload a file, or enter a URL. Get back:

  • Per-line parse errors against the llmstxt.org reference spec
  • Required-section presence (# Title, optional ## Start here / ## Reference / ## Optional)
  • Link-reachability check on every URL referenced (parallel HEAD requests, flagged on 4xx / 5xx)
  • A summary banner: pass / pass-with-warnings / fail

It's free. No signup. No email gate. The output is yours; the source code is MIT-licensed at github.com/bridgetoagent/llms-txt-validator and ships as @bridgetoagent-com/llms-txt-validator on npm if you'd rather wire it into a build pipeline.

If you want a validator for agents.json too, that's shipping in about three weeks. And if you'd rather skip writing either file by hand: the $49 BridgeToAgent kit generates both — plus agent-instructions.md plus the three <link rel="alternate"> tags for auto-discovery-links — from your actual site in under two minutes. The validator is the free side; the kit is the do-the-work side. Both make sense; pick what fits.

The honest read on llms.txt adoption

A counter-narrative has been gaining traction in 2026 — that llms.txt adoption is too low to matter and that the major AI crawlers (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended) overwhelmingly skip the file and crawl HTML directly.

That's partially true and partially misleading.

Partially true. Adoption sits at roughly 10.13% across the SE Ranking 300k-domain sample. Fortune 500 adoption is at 7.4% per ProGEO research. Major brands have not rushed to ship. Crawlers that bypass the file in favor of HTML are real.

Partially misleading. Three things the skepticism narrative tends to skip:

  1. Chrome just made it a default Lighthouse audit. That's the whole signal-volume change. Whatever the crawler adoption looks like in May 2026, the audit existence raises the cost of not having the file — the audit appears in every Lighthouse run, including audits run by your customers, your prospects, your potential partners. Failing the audit is a visible-to-everyone signal, regardless of whether ClaudeBot reads the file today.

  2. Crawler adoption is not the right metric. The audience for llms.txt is the agent doing the curated read, not the crawler doing the bulk scrape. ChatGPT browse mode, Claude with web access, Perplexity, Atlas, Comet — these read a handful of URLs per session. llms.txt is the file that tells them which handful. Crawler skip-rates don't measure agent-read behavior.

  3. The signal compounds with the other audits. llms.txt alone doesn't fix Agentic Browsing. llms.txt + agents.json + agent-instructions.md + <link rel="alternate"> + per-page Schema.org density does. Six of the nine default-config audits move at once when the kit ships. Evaluating llms.txt in isolation misses the system-of-files argument.

The honest position: the file isn't a silver bullet, but as of May 7 it's part of the default Lighthouse audit your customers, prospects, and partners now see automatically. The cost of shipping it is low. The cost of being the only site in your category that hasn't is rising.

What to do this week

  1. Run Lighthouse on your homepage now. Chrome 130+ should already have Agentic Browsing available; in 13.3.0 it's on by default. Web-based check: open pagespeed.web.dev/analysis/{your-url} once PageSpeed Insights inherits the 13.3.0 config (within two weeks of May 7 — should be live by the time you read this).
  2. If llms-txt-well-formed fails, paste the file into bridgetoagent.com/tools/llms-txt-validator. The error report tells you the line and the reason.
  3. If llms-txt-present fails, you don't have one yet. Hand-write a minimal version (the spec is at llmstxt.org; a 5-page site takes about 20 minutes), or run the audit to see how the kit generates one from your actual DOM.
  4. Don't wait for webmcp-annotations to pass. Lighthouse marks it as emerging; the WebMCP spec is still moving. The audits worth investing in this week are the seven stable ones.
  5. Re-run Lighthouse in 7 days after any change. Both PageSpeed Insights and DevTools cache audits; a re-run is how you confirm the file is being parsed correctly.

The window before "passing Agentic Browsing" stops being a differentiator and starts being table stakes is closing. Lighthouse default-config is the inflection. Six months from now this audit is normal, expected, and unremarkable. Right now it's a signal your competitors haven't shipped against yet.

Related

Keep reading

All posts →