Scoring method
GEO Cool (GEO库) offers a familiar single-page GEO audit: you give a URL; we return a score, per-item evidence, and a fix order. Rules do not come from a public 176-check list; they follow our knowledge-base acceptance criteria.
Knowledge base as of . Deep analysis.
Sampling
HTML, compression, HTTP version, and crawler probes all target the URL you submit. robots.txt, sitemap.xml, llms.txt, and /ai/ discovery files are read from the site root, because that is where those resources live. We do not silently audit another article linked from the homepage.
Base score
| Axis | Max | Items |
|---|---|---|
| Citation reach | 35 | Explicit allow 15, live reach 12, sitemap 8 |
| Structured data | 30 | Valid JSON-LD 10, key types 8, Q&A 5, entity 7 |
| Citable content | 20 | Answer first 8, table or authority link 7, time + price date 5 |
| Technical health | 15 | No nesting 5, lang + canonical + social 6, compression + HTTP/2 4 |
Confidence, micro-adjust, cap
- Fetch failures are unknown—not counted as 0 in the numerator; they only lower confidence.
- Discovery files such as llms.txt add at most +5. High-signal injection at most −5. Combined ±5.
- Any remaining P0 caps the final score at 74.
Bands
90+ excellent, 75–89 good, 50–74 pass, below 50 needs work.
Web vs CLI
Judges are the same. The web UI uses a 12-second fetch timeout and audits only the submitted URL. The CLI defaults to 20 seconds and may follow a discovered article when only a homepage is given.
Primary heading, semantic tags, meta placement, and security headers enter the issue list without owning a share of the base 100. Image dimensions are hints only.
Entity types: schema.org/Organization. Classic search background: Wikipedia · SEO.