The pattern, not the origin
Google's own public stance is straightforward: it doesn't penalize content for being AI-generated. It penalizes content that's generic, low-effort, and formulaic — regardless of who or what wrote it. The problem is that AI-generated drafts tend to produce exactly that pattern by default, unless someone specifically edits it out.
Our AI-content-quality checks target that pattern directly, not the fact that a model was involved.
What we actually check
Vague attribution. Phrases like "studies show" or "experts believe" with no named source or date nearby. It's a pattern real AI drafts fall into constantly, and it's a real trust problem regardless of what wrote the sentence.
Sentence-rhythm variance. Human writing naturally varies in sentence length and complexity. AI output tends to be statistically more uniform. We compute this directly — coefficient of variation across sentence word-counts — and surface it as a rhythm tip, never as an accusation that a piece "is AI-written." That's a different, much riskier claim we deliberately don't make.
Missing first-person signal. Near-total absence of concrete experience markers — "I tested," "after using this for three weeks" — maps directly to the Experience pillar of Google's own E-E-A-T framework.
Unsourced statistics. Specific-sounding numbers with nothing backing them up nearby.
One important scoping note
These checks run in addition to the full general SEO checklist, not instead of it. Every piece of content — human-written or AI-assisted — gets the complete analysis. For well-written human content, these particular checks will mostly just come back passing. Rankbist works on any content; it's especially useful on AI-assisted drafts, which is a large and growing share of what gets published.
