Skip to content

Site Pulse

Grade your site on what AI search actually rewards

AI engines can only recommend what they can read. Site Pulse crawls your pages and scores each one across the signals that decide AI readability: content quality, structure, metadata and technical health.

Free snapshot crawl. No card required.

What it checks

Our own homepage, taken apart exactly as the crawler read it. Every number below is lifted from the scan embedded further down this page.

Site Pulse scan of baselinelabs.ai - 25 August 2026 - 50 pages crawled - 22 modules

baselinelabs.ai 200 - self-canonical
ReachabilityPass

Status code, redirect chain, TLS, security headers. If the fetch fails, nothing below it can be graded. This scan recorded HTTP 200, an HTTP to HTTPS redirect of 308, and a TLS certificate valid for 77 more days.

Head, the part no reader sees
<title> 54 characters
<meta name="description"> 131 characters
<link rel="canonical"> self
MetadataPass

Title and description presence, length and uniqueness across the crawl, canonical, Open Graph. This page passed. Across the full crawl the same module raised 25 warnings, 22 of them descriptions outside the useful length band.

Crawler rules
robots.txt - 28 retrieval bots allowed
live probe - GPTBot, ClaudeBot, PerplexityBot, Google-Extended 200
llms.txt - 10,086 bytes, 47 links
Bot accessPass

What robots.txt permits is only half the answer, so the scan also fetches your site as each bot and records the real status code. A CDN can turn away an agent your robots.txt welcomes.

Structure
headings - 7 total, exactly one H1
hierarchy - valid, no skipped levels
landmarks - 3 semantic elements
StructurePass

One H1, headings in order, landmarks that separate body from chrome. Structure is how a model locates the answer inside a page instead of skimming past it. Three of the 50 crawled pages did raise an H1 warning.

What the model actually ingests

871 words of body copy. The extractor stripped 64 noise elements and reduced 194,372 bytes of HTML to 5,288 characters of readable text, an estimated 1,009 to 1,322 tokens.

Content clarityWarn

The content-to-markup ratio came out at 2.3%, so almost all of this page's bytes are markup rather than text a model can read. It is the most common warning in the crawl, raised on 49 of 50 pages.

Page weight and delivery
189.8 KB HTML 7 origins, no preconnect HTTPS CSP + HSTS
Technical healthWarn

Weight, render-blocking assets, mobile layout, security headers, tracking. No synchronous scripts sit in the head and every security header is set, but seven third-party origins load with no preconnect or dns-prefetch hint.

92.7/100
Fifty pages, 22 modules, one score

Every page in the crawl gets the treatment above. This run returned 1,396 passing findings, 143 warnings and no failures. The score is a weighted pass-ratio over the crawled sample, averaged across categories.

Then: what to fix first
  1. 1 Third-party origins loaded with no resource hint50 of 50 pages crawled 50
  2. 2 Content-to-markup ratio below the readable threshold49 of 50 pages crawled 49
  3. 3 Meta description outside the useful length band22 of 50 pages crawled 22

Ranked by how many crawled pages raise each warning. Every warning in this run carries equal weight, so the count is the ranking. Your own report ranks the same way over your own crawl.

What you'll get

See your site the way an AI engine does

Run a free scan

Site Pulse

Crawl your site and grade every page on the things that matter for AI search: content quality, structure, metadata, and technical health.

Manage
George
Online
0%