What the quality review checks before you approve a draft

Drafts Solvgent writes from a brief go through a quality review before they reach your queue. The review scores the draft and decides whether it passes. It never publishes anything. A person still approves every post.

What gets scored

Each draft written from a brief is scored automatically on four dimensions:

• Brand voice: does the draft sound like your brand?
• Readability: can someone read it quickly and easily?
• Engagement: is there a reason to stop and respond?
• Factual grounding: does the draft stick to the facts you gave Solvgent?

Each dimension gets a score from 1 to 10. The review also gives an overall score on the same scale.

Which drafts are reviewed

The review runs on drafts written from a brief, including drafts written on a schedule. Versions made by repurposing a post for another network, or by refreshing an older post, don’t go through this review today. Every draft, reviewed or not, still waits for a person to approve it.

When a draft passes

A draft passes when the overall score is at least 6 and factual grounding is at least 7. Both conditions must be met. Grounding has the higher bar because a wrong price or a wrong opening hour does more damage than a flat sentence.

When a draft fails

If a draft fails and the review names an issue, Solvgent tries one corrective rewrite that works on that issue.

If the review can’t run, the draft is kept and marked unreviewed. It is not marked as passed. Read an unreviewed draft closely before you approve it, because nothing has checked it.

What grounding means

Grounding means sticking to the facts you gave Solvgent in Brand Brain. Those facts are knowledge entries such as products, prices and hours. If a draft states a price that is not in Brand Brain, the grounding score should be low.

Specific entries are easier to check against. Keep your entries current, so the review checks against facts that are still true.

A quick example

Say you run a bakery. Your draft says the shop opens early on Saturdays, but your Brand Brain entry for hours says something else. The draft describes a fact you never gave Solvgent, or it contradicts one you did. Factual grounding is the score meant to catch this kind of mismatch. Like any automated check, it can miss things.

The fix is simple. Correct the entry in Brand Brain if it is out of date, or edit the draft so it matches the entry. Either way, read the post again before you approve it.

How to read the result

Start with factual grounding. A low score there needs your attention first, because you can check facts quickly.

Then read the overall score and the other dimensions. A draft can pass on its overall score and still be weak in readability or engagement. Edit those parts before you approve.

Why a person still decides

The review is automated, and it can be wrong. It can pass a draft that needs work, or flag one that is fine. Treat the scores as a guide, not a verdict.

Nothing publishes without approval. Every post waits in the approvals queue until someone on your team approves it. Use the scores to decide where to look first, then read the whole draft.

This split is deliberate. The scores tell you where to look. Judgment about tone, timing and context stays with the people who know the brand.