Every article gets a single quality score from 0–100 the moment it's generated. We bucket that score into three labels you see on the cards:

What the score is actually measuring

• AI-search readability. Whether the article has the clean structure ChatGPT, Perplexity, Claude, and Google AI Overviews need to quote it (a clear answer near the top, scannable H2s, FAQ pattern).
• Brand voice match. How closely the writing matches your saved voice fingerprint (rhythm, sentence length, word choice).
• Crawlability. Word count, heading hierarchy, internal links, structured data.
• Banned-phrase check. Generic AI-sounding language ("delve into", "in today's world", "game-changing") drags the score down.

Why aren't all my articles Great?

The writer aims for "Good or better" on the first pass — that's the bar where an article is genuinely useful and publishable without rework. Hitting "Great" consistently usually means one of these is dialed in:

• You've uploaded writing samples on the Voice page so the brand voice is well-defined.
• Your customer questions are specific and have an answerable angle (vague questions hurt the score).
• You've added competitor URLs in Settings → Writing options so the writer knows what to cover.

A "Rough" or "Good" article isn't broken — it's a draft you can ship in five minutes of cleanup. Use the editor's Preview button to see what it'll look like live before publishing.