DEV Community

Cover image for 1,800+ sites audited. The median GEO score is 59 out of 100. Here's what that tells us.
Juan Camilo Auriti
Juan Camilo Auriti

Posted on

1,800+ sites audited. The median GEO score is 59 out of 100. Here's what that tells us.

GeoReady has audited 1,800+ sites as of October 17, 2026. Each audit is a real run of the GEO Optimizer engine against a live URL, scoring 0 to 100 across 9 categories. The dataset is self-selected (people who ran an audit), not a random sample of the web. But it is the largest open dataset of AI-readiness scores I know of, and the patterns are clear.

The headline numbers

  • Median score: 59/100. Half of all audited sites score below 59.
  • Average score: 56.5/100. Slightly below the median, pulled down by a long tail of low scores.
  • 58% publish an llms.txt file. More than half, but 42% do not have one.
  • 76% ship JSON-LD structured data. Three quarters have some form of schema markup.
  • 27 AI crawlers checked per site. The audit tests whether each known AI crawler can access the site via robots.txt.

The score bands

The 1,896 audited sites split across four score bands:

Band Score range Share What it means
Foundation 0-49 57.6% AI crawlers can reach the site but basic signals are missing
Good 50-74 28.2% Most fundamentals in place, gaps in content structure or schema
Critical below 25 12.4% Site is largely invisible to AI crawlers
Excellent 75+ 1.8% Strong across most categories, minor gaps only

57.6% of sites sit in the Foundation band. They are reachable but missing basic signals: no llms.txt, incomplete schema, missing canonical URLs, no published dates, no named authors. These are not advanced GEO tactics. They are the prerequisites.

Only 1.8% score Excellent. That is 34 sites out of 1,896. The bar for "AI-visible" is low. Almost no one clears it.

What the median site looks like

A site scoring 59/100 typically has:

  • robots.txt that allows most AI crawlers (gaining 8-10 of 10 possible points)
  • Some JSON-LD schema, usually Organization and WebSite, but missing FAQPage or Article (gaining 6-8 of 10 possible points)
  • A meta description and title tag (gaining 5-6 of 8 possible points)
  • No llms.txt or an invalid one (gaining 0-2 of 6 possible points in rubric v2)
  • Content with headings but no answer blocks or quotable statistics (gaining 8-10 of 15 possible points)
  • No published or updated dates visible on the page (losing Google AI readiness points)
  • No named author byline (losing Google AI readiness points)
  • Some brand entity signals but no Knowledge Graph presence (gaining 4-5 of 8 possible points)

The gaps are consistent. Most sites have the technical basics (robots.txt, some schema, some meta). Most sites miss the content and entity signals (answer blocks, dates, authors, Knowledge Graph). The technical layer is partially solved. The content layer is not.

What the calibration study adds

The calibration study I published last week found that the GEO score does not predict whether Perplexity cites a page (odds ratio 0.9625, p = 0.7172). That result does not invalidate the audit. It clarifies what the audit measures.

The audit measures citability: whether AI crawlers can reach, parse, and understand a page. The 59 median score says most sites are halfway there on citability. The calibration study says citability is necessary but not sufficient for citation. A site with a 59 score may or may not be cited. A site with a 20 score is almost certainly uncited, because the crawler cannot read it.

The practical reading: if your score is below 50, you have a citability problem. Fix the foundation first. If your score is above 70, your citability is solid, and your citation gap is about authority, third-party presence, and training data, none of which an audit can measure.

What changes a score

Based on the fix recommendations the tool generates across the 1,800+ audits, the highest-impact fixes are:

  1. Allow AI crawlers in robots.txt (if blocked). Sites that block GPTBot lose 10+ points and are invisible to ChatGPT's web search. This is the single most common critical issue.

  2. Add JSON-LD schema (if missing). Sites without any structured data lose 10 points. Organization and WebSite schema alone gain 6-7 points.

  3. Add published and updated dates. Sites without visible dates lose up to 4 points in the Google AI readiness category. The WashU study found dates correlate with AI Overview citation.

  4. Create llms.txt. In rubric v2, llms.txt is worth 6 points. Present and valid gains the full 6. Missing loses 6. It is the easiest fix, even if it is low-impact.

  5. Move content from JavaScript to HTML. Sites that render content client-side appear as empty shells to AI crawlers. This is not a points issue. It is a visibility issue. If the crawler sees nothing, the score is meaningless.

The dataset

The benchmark data is published on the GeoReady state of GEO page. It is self-selected, not random. Sites whose owners ran an audit are likely more aware of AI search than the average site. The real web probably scores lower.

If you want to see where your site sits, run a free audit on geoready.dev. No signup, no account. Paste a URL, get a score and a fix list. The CLI is on GitHub, MIT licensed, 1,137 stars, 2,000+ tests.

Top comments (0)