Scoring version v1.0
Methodology
CiteCheckup is a deterministic audit of one public web page. It evaluates observable content and structure against seven rules-based dimensions. It does not access a private AI index, and its score is not the probability that an AI system will cite the page.
100 points in total
How the score is composed
20 / 20 / 15 / 15 / 10 / 10 / 10
Clear positioning
Looks for a useful page title, description, headings, and an explicit connection between the page and the target topic.
Topic relevance
Checks whether the supplied topic and its component terms appear with enough context to make the page's subject clear.
Answer-style content
Recognizes question headings and practical answer patterns such as explanations, steps, and guides.
Comparison / alternative content
Looks for decision-support language around comparisons, alternatives, reviews, features, and pricing.
FAQ & structured content
Rewards genuine question-and-answer content and detects whether the extracted page includes structured data.
Crawlability & content quality
Considers successful retrieval, useful content depth, and basic page metadata available to the extractor.
Brand clarity
Checks for clear, consistent use of the supplied brand in the title, headings, and page body.
Retrieval
Two extraction paths
The checker tries the Jina Reader service first, then falls back to a direct HTML fetch when needed. These paths are not equivalent: reader output and approximate HTML parsing can expose different text, metadata, or structured-data signals. Running the same URL through a different path can therefore change the score.
Language
English-oriented matching
Topic tokenization, prompt patterns, and several content signals are designed for English and ASCII text. Results for other languages, scripts, word forms, or synonyms may be incomplete because matching is rules-based rather than semantic.
How to interpret a result
Treat the score as a checklist for improving page clarity and structure. A higher result means the extracted page matched more of the published rules; it does not prove discovery, ranking, mention, or citation by any AI product.
When an OpenAI search-model snapshot is available, it is reported separately. That snapshot is a one-time observation, can vary between runs, and does not alter the deterministic readiness score.
Changelog
v1.0
Initial published methodology: seven deterministic dimensions totaling 100 points, with structured data treated as a general page signal rather than a promise of AI citations.