A blocked draft's card on Approvals shows the reason recorded by the review step and, where there is one, the first concrete fix. There are four kinds of block.
1. Reviewer policy block
The Reviewer reads every draft against your brand rules and platform policy. A regulated medical, financial or legal claim is blocked outright; softer problems (an absolute guarantee, a superlative, an off-brand tone) are sent back for a rewrite instead.
2. Claims that need a source
Every factual claim in a draft is checked against the Claims list on Brand. A claim the list supports is fine; an unsupported one is softened where a safe wording exists, otherwise it is surfaced on the draft as "claims need source" for you to verify. While your Claims list is empty, unsupported claims are flagged for you rather than blocked, so you decide. Adding your real, evidenced claims to Brand is what makes future drafts pass cleanly.
3. Too similar to a post you already have
Each draft is compared with your earlier posts. A near-duplicate is re-composed automatically from a different angle. In the weekly batch, a run that still collides after its retries is dropped and shows on the Weekly plan as a post without a draft.
4. The safety linter
Deterministic rules run before the Reviewer and block regardless of anything else: medical efficacy claims ("cures", "miracle cure"), political campaigning, timestamped production notes left inside a video post's text, and platform limits such as tweet length.
What auto-fix already tried
Before a draft is blocked, the pipeline re-composes it with the Reviewer's feedback up to three times, rewrites unsupported claims in up to three softening passes, and re-runs every check on each attempt. A draft you see as blocked is one that couldn't be fixed within that budget.
What to do
An already-blocked draft is not retried on its own. Edit the copy on the draft, use Revise with an instruction, or regenerate it from Compose. If the block was a claim, add the claim with its evidence on Brand first.