One source error should not become six posts

Repurposing multiplies mistakes too

Turning one long article into a LinkedIn post, a thread, a newsletter section, a video script and a few snippets is an obvious job for a model. The source exists, the channels are known, and the drafts are quick.

The danger is in the last step. A chain that publishes directly takes one factual error in the source, or one invented detail added along the way, and puts it in six public places at once. A correction is then one more post, seen by fewer people than the mistake.

Extract the claims once

The most useful step happens before any channel is written. Read the source once and extract its claims, each with the evidence the source gives for it. Every channel prompt then writes from that list, not from the original text.

This does two things. Channels cannot drift into saying things the source never said, because they only have the claim list to work from. And checking a draft becomes mechanical: does every statement in it map to a claim on the list?

Write each channel for its reader

A thread, a newsletter section and a community answer are not the same text at different lengths. Each channel has its own reader and its own failure, such as a thread that buries the point, or a community answer that reads like an advert. Separate prompts per channel, each with its own constraints, produce drafts that need less rewriting.

End at a queue, not a feed

The workflow's last step should write each draft to a review queue, never post it. A person reads it against a short checklist, edits or rejects it, and releases it. Rejecting a draft is a normal outcome, not a failure of the pipeline: a good example shows the drafts that were rejected and why.

Putting it into practice

The AI Content Repurpose Workflow is an importable n8n workflow built this way: it extracts claims and evidence first, checks the brand file, fans out to six channels with their own prompts, self-checks each draft against the claim list, and writes the results to a review queue. Nothing in it posts. It comes with a nine-point quality checklist and one article run through all six channels, including the two rejects.

Keeping the drafts in one recognisable voice is the subject of why tone adjectives do not make a brand voice.