How does GEOScope calculate its score?
GEOScope scores a website across four weighted categories — robots and llms.txt access, structured data, content structure, and semantic HTML — then rolls them into a single 0–100 grade. Each category checks one specific, verifiable pattern that crawler documentation and published research associate with a page getting cited by ChatGPT, Perplexity, or Google AI Overviews, rather than a subjective quality judgment.
Why does robots.txt and llms.txt access make up 30% of the score?
This category is worth 30% of the total score, the largest single share. It parses robots.txt against 11 known AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and others — to confirm each one can actually fetch the page, since a blocked crawler cannot cite content it was never allowed to read. It also checks llms.txt and llms-full.txt for the heading, summary line, and linked-section structure the llms.txt convention expects.
What counts as complete structured data?
This category carries 25% of the score. It parses every application/ld+json block and checks which schema.org types are present — Article, FAQPage, Organization, BreadcrumbList, HowTo — against a list of recommended fields per type. A block with fewer than 60% of its fields filled in scores worse than no structured data at all, because thin schema misleads more than it helps a system extracting facts from the page.
Why does content structure matter as much as technical markup?
This category, also 25% of the score, implements structural patterns from published generative-engine-optimization research: question-phrased subheadings, paragraphs that answer before they explain, tables over dense prose, a 30-to-70-word paragraph band, and a baseline density of concrete statistics. None are hard requirements — they're the shape that answer-extracting systems tend to quote most cleanly.
How much does semantic HTML and metadata affect the score?
The remaining 20% covers semantic HTML and page metadata: whether main, article, and section landmarks are used correctly, whether headings skip levels, whether the title and meta description sit inside their recommended length ranges, and whether a sitemap.xml exists and parses. These are the same landmarks screen readers rely on, which is not a coincidence — machine readers, human or artificial, benefit from the same structural cues.
| Category | Weight | What it checks |
|---|---|---|
| Robots & LLMs.txt | 30% | AI-crawler access in robots.txt, plus llms.txt/llms-full.txt structure |
| Structured Data | 25% | JSON-LD presence and completeness of recommended schema fields |
| Content Structure | 25% | Question-style headings, answer-first paragraphs, tables, paragraph length, statistics |
| Semantic HTML & Meta | 20% | Landmarks, heading hierarchy, title/meta length, sitemap.xml |
Frequently asked questions
The four questions below cover what this score does and doesn't claim, since a numeric grade invites more confidence than any heuristic check should get.
Does a high GEOScope score guarantee my page will be cited by ChatGPT or Perplexity?
No — GEOScope's score reflects published technical and structural signals correlated with AI-crawler access and citation-friendliness. It never calls ChatGPT, Perplexity, or any other AI system to test a real citation, so treat it as a heuristic proxy for citability, not a guarantee of AI search visibility for any specific query.
Why does GEOScope check robots.txt for AI bots specifically, instead of just Googlebot?
Because AI answer engines crawl with their own named user agents, separate from traditional search indexing, and an older robots.txt written before AI crawlers existed usually has no rule for them at all. A missing rule defaults to allowed under the standard, but an explicit disallow — often added defensively — silently blocks citation-worthy content from every AI crawler on the list.
Is a thin FAQPage or Organization schema better than none at all?
Not according to this checker. Schema with under 60% of its recommended fields populated scores as a fail, worse than the page having no structured data. The reasoning: a nearly-empty schema block signals a template that was never filled in, which can actively mislead a system extracting facts from it — better to omit the block than ship it half-done.
How is GEOScope's own /geo page scored, given it's a scanning tool rather than an article?
The same four categories apply to any page type, tool or article. Because self-scanning a subpath like joram.eu/geo currently resolves robots.txt, llms.txt, and sitemap.xml checks against the joram.eu domain root rather than this page's own path, those three specific checks reflect the parent domain's files, not this repository's — a known limitation, not a scoring exemption.