Transparency
Methodology & scoring
RankForge is an automated diagnostic and prioritization assistant. It finds technical and content issues fast and tells you what to fix first — it is not a source of truth for whether a page will rank. Here is exactly how the score works and where its limits are.
The score is a model score, not a Google score
A RankForge score of 82 means “82 according to RankForge's model,” not “Google considers this site 82% optimized.” Third-party tools have no access to Google's ranking systems. Per Google's own guidance on third-party SEO tools, no external tool can guarantee rankings or reflect Google's internal signals. Use the score to track your own progress over time, not as a ranking prediction.
Every finding is labelled by evidence
We keep proven facts visibly separate from measurements and opinions, so polished prose never reads as more certain than the evidence supports.
Proven from the HTML, rendered DOM, response headers, or source code we actually fetched. Example: “the page ships no canonical tag.” These are facts you can re-check in the evidence we cite.
A number from an instrument. We label lab (one synthetic headless-Chromium run from one location) separately from field (real-user CrUX / Search Console data). Example: “LCP was 3.4s in this test environment (lab).”
A model's judgement, not a hard fact — e.g. the AI design review's read of your visual hierarchy. Useful, but weaker than verified or measured evidence.
A recommended change (an opinion about what to do), not a detected defect.
Category weights
The overall score is a weighted average of category scores. Each category score starts at 100 and loses points per issue: penalty = weight × severity multiplier, where severity multipliers are critical ×1.6, warning ×1, notice ×0.5.
| Category | Weight |
|---|---|
| Technical SEO | 20% |
| On-Page SEO | 20% |
| Content Quality | 14% |
| Structured Data & Social | 10% |
| Performance | 12% |
| UX & Web Vitals | 14% |
| Code & Architecture | 10% |
Not every issue is a ranking lever. Each finding is tagged with what it actually affects — ranking, crawlability, presentation, social, accessibility, security, or performance — so a missing Open Graph tag is never confused with a crawl-blocking robots rule.
What a crawler cannot know
RankForge identifies many possible contributors to how a page performs. It cannot measure these, and no on-page audit can:
- ✕Backlink strength, quality, or authority
- ✕Brand and branded-search demand
- ✕Query-specific intent and SERP competition
- ✕Real-user performance across devices, networks, and countries
- ✕Google's indexing decisions and quality assessments
- ✕How helpful your content is versus specific competitors
For those, pair RankForge with Search Console (connect it in Settings for real query data), PageSpeed Insights, and the Rich Results Test.