AI Content Repurposing That Doesn’t Create Generic Slurry

Our Testing Methodology

Common Mistakes That Create Low-Value AI Content

These patterns show up on almost every failed rollout we re-test in the workflow lab. Use the paired fixes when you evaluate your next tool.

Mistake

Prompting “turn this transcript into 10 LinkedIn posts” in one shot

Do this instead

Extract angles first (list of 5–7 claims), assign one angle per post, run channel constraint + anti-slurry filter on each

Mistake

Using AI summary as the source (Loom auto-summary, Notion summary only)

Do this instead

Work from full transcript with anchor quotes — summaries strip the specifics that differentiate you

Mistake

Shipping without a human pass because “AI got the grammar right”

Do this instead

15-minute edit for proper nouns, numbers, client confidentiality, and one sentence only you would write

Mistake

Same CTA and hook on every derivative asset

Do this instead

Map CTA to channel intent — awareness post ≠ newsletter ≠ lead magnet download

Mistake

Repurposing thin source material to hit a posting quota

Do this instead

One high-signal Loom per week beats five idea-list posts; skip weeks when signal is low

The Weekly Repurposing Framework (One Source → Many Assets)

1. Monday — Capture high-signal raw material

   1. Record one 15–30 min Loom doing real client work, or import one call into Descript and verify transcript accuracy on proper nouns.
   2. Mark 3–5 timestamped “anchor moments” — numbers, objections, or process steps worth extracting.

2. Tuesday — Extract angles (no drafting yet)

   1. Run anchor-quotes-first prompt in Claude AI; confirm at least 3 verbatim quotes with specifics.
   2. List 5–7 non-overlapping angles; kill any angle that fails “would a competitor say the same thing?”

3. Wednesday–Thursday — Batch drafts by tool chain

   1. Assign one angle per output type (newsletter, carousel, 2 social posts, 1 clip script) — max 5 derivatives.
   2. Run voice lock + channel constraint per asset; run anti-slurry filter on each before moving on.

4. Friday — Edit once, schedule, log winners

   1. Single 30-min edit pass: numbers, names, links, one “only I would write this” sentence per asset.
   2. Schedule distribution; note which angle/format got saves or replies — reuse that angle type next week, not the same hook.

Why “Repurpose Everything” Produces Slurry

Source has no proprietary detail

If the transcript could be any coach explaining “add value” and “be consistent,” no prompt chain saves you. Record while doing real work — screen share beats talking head.

One prompt, many formats

Mega-prompts optimize for volume, not information gain. Each format needs constraints, CTA, and a separate anti-slurry pass.

No extraction step

Jumping from raw transcript to polished post skips angle selection — the model fills gaps with generic marketing language.

Scheduling slurry to protect cadence

Posting interchangeable content trains your audience to ignore you. Better to ship two sharp pieces than eight forgettable ones.

Frequently Asked Questions

What is “generic slurry” in AI content?

Interchangeable marketing copy that could belong to any brand — correct grammar, zero specific examples, no information gain over the source. It fills calendars without building trust.

Is Loom enough as a raw material source?

Yes, if you work from the full transcript with anchor quotes — not the auto-summary alone. Walkthroughs where you show real work beat talking-head “tips” videos.

How many derivatives should one source support?

We cap at five quality pieces per flagship asset. More than that usually means stretching angles and diluting signal — the root cause of slurry.

Do I still need to edit AI repurposed content?

Always. Budget 15–30 minutes per batch for numbers, confidentiality, voice, and the anti-slurry filter. Grammar is the easy part; specificity is not.