# Content Score Content checks whether your page is actually **shaped to answer the question** it's about - clear, current, and deep enough to be worth citing. --- ## Why it exists - AI assistants favor content that gets to the point, stays current, and covers a topic in enough depth to be a genuinely useful source. - A page can load perfectly (passing Technical) and still lose out here if it's thin, stale, unclear, or hidden behind JavaScript the crawler never runs. --- ## What's checked 13 checks feed this score. Four of them - answer-first, question-style headings, original data, and named author - are judged by an AI reading the page the way a person would, not by simple keyword matching: | Check | Severity | What it verifies | |---|---|---| | Readable content | Critical | The main content is in the raw page (not something only visible after JavaScript runs - many AI crawlers skip that step). | | Page title | Warning | There's a clear, descriptive title. | | Meta description | Warning | A 50-160 character summary exists for AI to use as a snippet. | | Main heading (H1) | Warning | The page has exactly one clear main heading. | | Heading order | Warning | Headings follow a logical outline, not a jumbled structure. | | Publish / update date | Warning | A visible, parseable date lets AI judge how current the page is. | | Content freshness | Warning, graded | Full credit if updated within 60 days, tapering to 0.7 within 180 days, 0.4 within a year, 0.2 within two years, and 0 beyond that. | | Answer-first | Warning, AI-judged | The page opens with a direct, self-contained answer AI can quote. | | Question-style headings | Warning, AI-judged, graded | Headings are phrased the way real customers actually ask questions. | | Original data | Notice, AI-judged | The page includes original research, surveys, or benchmarks worth citing. | | Named author | Warning, AI-judged | Content is attributed to a real, credentialed person - a trust signal AI weighs. | | Content depth | Notice, graded | Word count judged against a target band for the page's type - thin pages earn partial credit, not a flat pass/fail. | | Tables & lists | Notice | Structured formats let AI extract answers cleanly (skipped entirely on page types where it doesn't apply). | - Answer-first specifically means the first ~150 words contain a direct, 40-60 word, quotable answer - not a marketing lead-in. - Named author means a real Person with credentials (a job title, qualifications, a linked profile) - an "Organization" or a generic team byline doesn't count. --- ## Depth bands by page type GrackerAI classifies each page's type automatically - from its schema markup, Open Graph tags, URL, or (as a last resort) its own content shape - and grades word count against a target range for that type, not one flat number for every page: | Page type | Target word count | |---|---| | Article, Recipe, Review | 600-4,000 | | FAQ / Q&A page | 400-3,000 | | Product | 200-1,500 | | Service | 200-1,200 | | Listing / homepage | 150-800 | - Inside the band earns full credit. - Below it, credit scales down linearly toward zero. - Above it, credit tapers off but never drops below 0.4 - going long isn't itself a fault. Several checks are skipped entirely (not failed) on page types they don't apply to: - **Publish date, freshness, and named author** only apply to Article/Recipe/Review pages. - **Answer-first** applies to those plus FAQ/Q&A pages. - **Question-style headings** applies to Article/Recipe plus FAQ/Q&A. - **Original data** only applies to Article pages. - **Tables & lists** applies to Article, Product, Recipe, and Dataset pages. A Product page, for example, is never penalized for missing a named author or a stale-content warning - those checks simply don't run on it. --- ## What moves the needle most Same weighting every score uses: - **Critical** issues (content hidden from crawlers entirely) count 5 points. - **Warning**-level issues (a missing H1, stale content, no answer up top) count 2. - **Notice**-level polish (missing tables, no original data) counts 1. The highest-severity fix is always the biggest lever. --- ## How to raise it - Make sure the real content renders in the raw HTML, not only after JavaScript executes. - Write a clear title and a 50-160 character meta description. - Use exactly one H1, and keep headings in a logical H1 → H2 → H3 order. - Show a visible publish or last-updated date - and actually keep it current. Credit starts decaying after 60 days and keeps dropping through the two-year mark. - Open with a direct, 40-60 word answer to the page's main question within the first ~150 words - before any marketing lead-in. - Phrase headings the way a real person would ask the question. - Add original data, research, or a named expert author where you can. - Go deep enough on the topic to be genuinely useful, not a thin overview. - Use tables and lists wherever the content is naturally structured. --- ## Next Steps - [AEO Audit overview](/docs/using-gracker/aeo-audit): Page Health, Site Health, and running an audit - [Technical score](/docs/using-gracker/aeo-audit/technical-score): can AI actually fetch and read the page? - [AEO score](/docs/using-gracker/aeo-audit/aeo-score): structured data and sourced claims