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.