Can an AI agent read your site, or does it leave with a blank page?
Most AI agents and answer engines read a site as raw text: no browser, no JavaScript, no patience. What they cannot see they cannot cite, and what they cannot cite they cannot recommend. The Barometer takes the same reading they do and scores what they can actually see: one URL, sixteen checks in four families, a score out of 100. Nothing is stored.
You leave with a score out of 100, a reading on the dial, each of the sixteen checks passed, warned or failed, and what to fix first.
Sixteen checks in four groups, each weighted by how much reach a site loses when it fails, not by tidiness.
Can agents get in and move around? The first file every crawler reads. It should welcome them and point at the sitemap. GPTBot, ClaudeBot, PerplexityBot and the rest are named separately in robots.txt. Blocking them is a decision a site is entitled to make, and it settles whether those assistants can quote it at all. The index of everything worth finding, with dates so crawlers know what changed. One stray noindex tag and the page is invisible everywhere. Tells machines which address is the real one, so the signals do not split. Agents walk a site by following hrefs. Navigation that only exists once JavaScript runs is a dead end to them.
Does the page explain itself? The line agents quote first. 15 to 70 characters lands cleanly everywhere. The one-sentence answer to what this site is. 50 to 160 characters. A single top heading gives the page one unambiguous subject. One attribute that tells every parser what language it is reading. Without it, mobile rendering is guesswork and crawlers notice. Most agents do not run JavaScript. Content that only exists after a framework boots starts from a blank page.
Can machines quote you accurately? JSON-LD is how machines learn prices, services and FAQs without guessing. og:title and og:image decide how every shared link looks.
Is anything written for agents? A plain-text briefing for AI assistants at /llms.txt. Few sites have one yet. A title, a summary and a linked list of pages make the file liftable; a bare paragraph gets skimmed.
How the Barometer reads a site, and what happens to the URL.
The page is fetched server-side and read as raw HTML, plus robots.txt, the sitemap and llms.txt. No browser, no JavaScript execution: the view most crawlers and AI agents start from.
robots.txt is read past the blanket rule at the top. A site can welcome Googlebot and still turn away GPTBot, ClaudeBot and PerplexityBot by name, and that block is invisible to an ordinary SEO audit.
The reading runs, the report renders, and the URL goes nowhere else.
The Barometer reads the site. Three free diagnostics read the function, in your own Claude or ChatGPT.
Does every piece of marketing work make the next one easier, or does each team start from scratch?
Run the diagnosticCan everyone on your team get the same quality from your agents, or only the person who wrote the prompt?
Run the diagnosticWhich of your marketing jobs should run overnight, which should wake you up, and which should never leave the room?
Run the diagnostic