Diverge-Then-Select Content Generation
Generate multiple angles before selecting and drafting the strongest
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 90%
This workflow makes AI brainstorm before it writes. Given a topic, the model first creates six to eight meaningfully different angles, such as thought-provoking, curiosity-driven, contrarian, or debate-provoking approaches. It exposes the alternatives to the user and then selects—or helps the user select—the angle best matched to the topic and intended audience response. Only after that decision does it draft the full post. Separating divergence from convergence reduces the chance that the first statistically obvious framing becomes the final answer. The user should still review whether the options are genuinely distinct and whether the chosen angle is accurate, useful, and aligned with the brand rather than merely provocative.
Origin
Extracted from Marketing Against The Grain
Core principles
- 01Separate idea generation from final drafting
- 02Create enough alternatives to escape the first obvious answer
- 03Make the option set visible before selection
- 04Choose an angle that fits both topic and content format
- 05Treat the model's selection as a recommendation, not a verdict
How to run it
- 1
Frame the creative problem
State the topic, audience, platform, and reaction the content should produce. Include any factual or brand constraints.
Pro tip Define a desired audience response such as reflection, curiosity, or constructive disagreement.
Watch out Without constraints, the model may generate cosmetic rather than strategic variety.
- 2
Generate multiple angles
Ask for six to eight distinct approaches before any full draft is written. Require a concise hook or premise for each.
Pro tip Use rhetorical categories to encourage genuine divergence.
Watch out Several paraphrases of the same idea do not constitute distinct options.
- 3
Display the brainstorm
Make the model show all candidate angles and briefly explain why each might work. Do not let it silently discard alternatives.
Pro tip A visible option set allows the human to notice ideas the model would not select.
Watch out Automatic selection can hide the most original candidate.
- 4
Apply selection criteria
Compare the options for relevance, originality, credibility, audience fit, and potential engagement. Choose one angle deliberately.
Pro tip Let the model recommend an option, but retain the final decision.
Watch out Do not select provocation that undermines accuracy or trust.
- 5
Draft the chosen angle
Create the complete content using only the selected premise and the required format. Preserve the strongest wording from the brainstorm where appropriate.
Pro tip Restate the chosen angle in the drafting instruction to prevent drift.
Watch out The model may blend discarded options back into the draft.
- 6
Iterate the execution
Critique the resulting content for generic language, weak evidence, or a diluted hook and revise it through focused feedback.
Pro tip Generate alternative hooks if the body works but the opening does not.
Watch out A strong angle does not guarantee a strong final draft.
In the wild
A marketer gives AI a leadership topic and requests several possible treatments, including thought-provoking, curiosity-led, and deliberately challenging approaches. After reviewing the options, the marketer selects the sharpest credible angle and asks for the complete post.
→ The draft begins from a deliberate content strategy rather than the model's first generic framing.
A founder asks AI for eight announcement angles: customer problem, founder story, contrarian belief, product demonstration, quantified result, market shift, behind-the-scenes build, and customer quote. The founder scores them for evidence and audience relevance before drafting.
→ The announcement uses the strongest supportable narrative rather than a default feature list.
Common mistakes
Generating disguised duplicates
Changing the hook wording without changing the underlying premise creates an illusion of choice.
Letting AI select invisibly
The model may discard an original but unconventional idea before the user can evaluate it.
Choosing engagement over truth
A provocative angle that cannot be supported can damage credibility even if it attracts attention.
Is it for you?
Best for
Creators who need several viable hooks or rhetorical approaches before drafting a post.
Not ideal for
Simple factual messages where option generation adds complexity without improving the answer.
From the transcript
“I was actually getting it to brainstorm for itself. So basically, it brainstorms six to eight different options for that LinkedIn post, shows you the…”
“I was actually what I was trying to do here was say the different types of posts that work really well on LinkedIn and then…”
From the episode
Generate 10x Views On A LinkedIn Post With These GPT-4o Prompts