Is the score computed by an AI?

No — deliberately. The score is deterministic: the same article always produces the same number, computed from measurable components. An AI grading an AI would just be a second opinion you can't audit.

Treat it as a pre-flight check rather than a grade to chase. A high score means structure, verification coverage, length, and linking are all in order; a low score points at exactly which of those needs attention before you publish.

Can a well-written article score badly?

Yes — if its claims couldn't be verified. Fact-check coverage is the heaviest component, so an elegant article full of unverifiable statistics scores lower than a plain one that checks out. That's the intended behavior.

The AI Content Quality Score, Explained

A number you can argue with

Plenty of tools stamp content with a quality score. The usual problem: the number is a black box, so you can't tell whether 84 means "publish confidently" or "the model felt good today."

SEOtopilot's score is different in one structural way: it's deterministic and decomposable. The same article always scores the same number, and the number is a weighted sum of components you can inspect individually. You don't have to trust the score — you can audit it.

The components

Fact-check coverage — weighted heaviest

What share of the article's factual claims were verified against a live source in the fact-check pass? Verified claims raise the component; contradicted claims actively lower it. An article with no checkable claims at all sits at a deliberate neutral — fine, but unremarkable.

This component carries the most weight on purpose. Structure and length affect rankings; unverifiable claims affect your reputation. The score reflects that priority.

Structure

Did the draft actually deliver the outline's plan? Every planned section present, headings in a valid hierarchy, no skipped levels. An article that silently dropped two sections scores against it here.

Length adequacy

Measured against the outline's target word count — not against a mythical "2,000 words ranks best" rule. A tight 900-word answer to a simple query scores fully when 900 words was the plan.

Heading depth

A modest reward for genuine subsection structure — scannable content with real H3s under its H2s, rather than a wall of same-level headings.

How to use it in practice

The score appears with every article in the review queue, next to the margin annotations. The workflow it enables:

Sort by score. Review the weakest articles first — that's where your editing minutes matter most.
Read the components, not just the total. A 68 driven by low fact-check coverage is a different job (source or cut the claims) than a 68 driven by missing sections (regenerate the outline).
Regenerate freely. Fixing a weak component and re-running costs nothing — regeneration is always free, per our pricing.

What the score is not

It's not a ranking prediction — nobody can honestly sell you one of those. It's not a substitute for reading the article. And it's not a gate: a low-scoring article can still be published if you judge it right, exactly as the review workflow leaves every publish decision with you. The score's job is to point your attention, not to replace it.