How is the GEO score calculated?

The GEO score is a weighted average of 60 scored checks in 8 categories (ruleset 2026.10). A pass earns full points, a warning half, and a fail none. Checks that don't apply to a page are left out rather than counted against it.

What caps an engine's score?

An engine's score is capped at 10 when a check that blocks it fails. For example, a robots.txt rule that disallows OAI-SearchBot caps the ChatGPT score, because ChatGPT search cannot read the page at all.

Which training crawlers are ignored?

Training crawlers such as GPTBot, ClaudeBot and Google-Extended are reported for information only. Blocking them is a valid choice that does not affect AI search citations.

AI crawler access (20%)

CheckPointsBlocks engine
Page responds with HTTP 2003chatgpt, perplexity, claude, google, copilot
robots.txt is reachable1–
OAI-SearchBot (OpenAI) can crawl this page4chatgpt
PerplexityBot (Perplexity) can crawl this page4perplexity
Claude-SearchBot (Anthropic) can crawl this page3claude
Googlebot (Google) can crawl this page4google
bingbot (Microsoft) can crawl this page3copilot
Other AI search crawlers are allowed1–
User-triggered AI fetchers are allowed2–
AI training crawlers (informational)info–
AI crawlers aren't blocked by a firewall or CDN3–
Page is indexable (no noindex)4chatgpt, perplexity, claude, google, copilot
Content is snippet-eligible2google
HTTPS with a short redirect chain1–
Fast, lightweight HTML response1–
robots.txt declares a sitemap1–

Indexability & discoverability (10%)

CheckPointsBlocks engine
XML sitemap lists this page4–
Sitemap lastmod matches the page's modified date1–
Canonical URL is set correctly3–
Last-Modified or ETag header1–

Renderability without JavaScript (15%)

CheckPointsBlocks engine
Main content is in the HTML (no JavaScript needed)6–
H1 and opening paragraph are in the raw HTML3–
Structured data is in the raw HTML2–
Healthy text-to-HTML ratio2–
Markdown version advertised (bonus)info–

Content structure & answer-first writing (20%)

CheckPointsBlocks engine
One H1 and a logical heading hierarchy3–
Headings match how people ask questions2–
Sections open with a direct answer4–
Passages make sense on their own3–
Short introduction before the first section2–
Lists or tables for scannable facts2–
FAQ section2–
Enough substantive content3–
Readable sentence length1–
Specific facts, numbers and dates3–
Quotes or attributed statements1–
Cites external sources2–
No keyword stuffing1–
Defines the main subject near the top2–

Structured data (12%)

CheckPointsBlocks engine
Valid JSON-LD structured data3–
Organization or Person entity with sameAs3–
WebSite entity1–
Schema type matches the page2–
Article has dates and an author2–
FAQ schema matches visible text1–
Breadcrumb structured data1–
Product offers include a price1–
Schema headline matches the H11–

Entity clarity & trust (10%)

CheckPointsBlocks engine
Brand name is consistent3–
Visible author byline2–
About and Contact pages are linked2–
Visible publish or update date1–
Links to official brand profiles2–

Freshness (5%)

CheckPointsBlocks engine
Content updated within the last year3–
Dates agree across sources1–
Copyright year is current1–

Metadata & previews (5%)

CheckPointsBlocks engine
Descriptive title tag2–
Meta description1–
Open Graph and Twitter tags1–
Page language is declared1–

llms.txt & agent files (bonus) (bonus)

CheckPointsBlocks engine
llms.txt2–
llms-full.txt1–