Skip to content

Using the Content Coach

Nine named checks on a draft, five measured against a network's published policy and four judged by a model. Advice, never a gate.

Updated

The Content Coach reviews a draft against nine named checks and tells you, for each one, what it found and why. Five of them are measured against a network's published policy; the other four are a model's judgement, and every row says which kind it is.

It is advice and never a gate. Brand Guard decides what may be published; nothing here blocks anything, and nothing here forecasts how a post will do.

BEFORE YOU START

  • The "edit" or "create" permission. Without it the form is not shown at all.
  • An AI text provider, if you want the four judgement checks to run. The five measured checks run either way.
  • A Business DNA, if you want the brand voice check to mean anything. Without one the model is told it has nothing to judge brand voice against.
  • Nothing else, and no credits: the Coach reserves and settles nothing.

STEPS

  1. Open Create and choose "Content coach", or go to /app/content-coach.
  2. Paste the copy into "The draft".
  3. Pick the "Network". Leave it unchosen and the length, hashtag and link checks report themselves not judged, because every network publishes a different limit and there is nothing to measure against.
  4. Choose "Review this draft" to see the result without storing anything, or "Review and record it" to write the review down so it can be read later beside what the post actually did.
  5. Read the table. Each row gives the dimension, whether it was Measured or Judged by a model, the verdict, why, and what to change.

WHAT YOU SHOULD SEE

A summary chip and a nine-row table. The summary is the worst dimension that could be judged, never an average and never a score. Verdicts are Strong, Adequate, Worth changing or Not judged.

WHAT THIS WILL NOT DO

  • It does not give a score. There is no number, and the summary is deliberately the worst judged dimension rather than a blend.
  • It does not block anything. Approval and publishing are unaffected by what it says.
  • It does not predict performance.
  • It does not quietly pass a check it could not run. A dimension that could not be judged says "not judged" and gives the reason.
  • Reviewing stores nothing at all. Only "Review and record it" writes anything down.

IF IT DOES NOT WORK

  • "No AI provider is configured in this environment." means the four judgement checks cannot run. They report as not judged; the five measured checks are unaffected.
  • "The model could not be reached for this review." or "The model replied in a form this review could not read." mean the same thing for those four rows.
  • "That draft was not reviewed. Nothing was stored - check your permissions and retry."
  • "Both actions need a draft: type or paste the copy you want reviewed."
  • "Recorded at ..., but this draft matches no stored content version" means the review is kept but can never be checked against an outcome. Save the copy as a content version and record again if you want it checkable.
  • "Not enough evidence yet: N of this organisation's posts have a recorded measurement, and 8 are needed" means the read-back section stays ungraded until then. Nothing has been lost.

COMMON QUESTIONS

Why does the length check say it cannot judge?

Either you did not name a network, or the product holds no researched posting policy for the one you named, or that network publishes no caption ceiling. All three are different and the message says which.

What is the hashtag guide of eight?

That is our own editorial view, not a rule any network publishes, and the message says so when it applies it.

What does "What happened next" show?

Reviews you recorded, beside what those posts actually measured, and whether the advice pointed the right way. It needs eight measured posts in the organisation before anything can be called typical.

  • Using the Viral Coach

    Comparing a draft with your own past posts, and importing reference signals by hand. Nothing here predicts reach.

  • Writing and approving content

    The content editor - saving versions, submitting for approval, approving, and why an author cannot approve their own work.

  • Writing a Brand Brain

    How one brand's tone, audiences, banned phrases and claim rules are recorded, and which of them are actually enforced.