From keyword to 30 usable short-form content angles without drowning in junk ideas

What this solves

AI can generate many ideas, but most are bland. Ask a raw LLM for "content ideas about AI tools for solo creators" and you get back twenty variations of "5 AI tools you need," half of which are invented products. The real problem is not volume — it's signal-to-noise. Generic angles cost you time at every stage: they waste your editing budget, they test poorly with audiences, and they dilute your content identity.

This workflow adds scoring and rewrite passes so the final list is usable. It transforms a firehose of half-baked concepts into a curated set of platform-native titles ready for editorial review. The key innovation is the scoring rubric — not just "is this good?" but four specific dimensions that map directly to content performance.

Real-world example: turning "AI for creators" into 30 distinct posts

A newsletter operator ran this workflow on a single keyword: "AI tools for solo creators." The raw LLM output produced 50 angles — 37 were identical rewordings of "X AI tool saves time." After scoring and filtering, the top 10 included: "The AI TTS voice that sounds like you (not a robot)," "I replaced my VA with a local LLM and saved $1,200/month," and "Why your Xiaohongshu drafts are getting 0 saves (AI audit checklist)." Each one was specific, novel, credible, and conversion-optimized.

The operator reported that the scored list needed only 20 minutes of human review versus 3+ hours of manually brainstorming and filtering. The 30 approved angles produced 9 newsletter signups in the first week — a 3x improvement over their previous manually-curated batch.

Test input

Reusable asset

Scoring rubric: how we separate gold from noise

Each generated angle is scored 1-5 on four independent criteria. An angle must average 3.5+ across all four to survive the cut. Here is what each dimension means in practice:

Specificity — Does the angle reference a concrete tool, workflow, or outcome? "Best AI tools" scores 1. "How I use Whisper to transcribe client calls in 3 minutes" scores 5.

Novelty — Would a creator in this niche have seen this take before? "AI saves time" scores 1. "Why running a local LLM on an M4 Mac Mini beats paying for ChatGPT Pro" scores 5.

Credibility — Can the claim be backed with a test, screenshot, or data point? "AI is amazing" scores 1. "I tested 4 AI meeting summarizers against human notes — here is the one that got 94% accuracy" scores 5.

Conversion fit — Does the angle naturally lead to a newsletter signup? A listicle that ends with "subscribe for more" scores 2. A case study framed as "I solved [pain] with [method] — full walkthrough inside my newsletter" scores 5.

Step-by-step walkthrough

Step 1. Paste your keyword and platform into the prompt. Generate 50 raw angles in one pass. Do not stop to judge — volume is the goal at this stage.

Step 2. Run the scoring pass. Using the same AI session, ask it to score each angle on the four dimensions above. Request a CSV-like output: angle | specificity | novelty | credibility | conversion_fit | average.

Step 3. Apply the 3.5 average threshold. Anything below gets archived. This typically cuts the list from 50 to 12-18 surviving angles.

Step 4. For the survivors, request a rewrite pass. The AI should convert each angle into a platform-native title matching the character limits and style of your target platform. For Xiaohongshu: question-based hooks, emoji use, 20-25 characters. For LinkedIn: problem-solution framing, professional tone, 50-70 characters.

Step 5. Human review. Check each angle for truthfulness, personal fit, and platform compliance. This takes 15-20 minutes for 30 angles. Approve or reject. The final output is a ranked editorial queue.

Expected output

A ranked list of angles with reason codes and rewritten titles ready for editorial review. Each entry includes the original angle, its dimension scores, the platform-adapted title, and a "why this works" note. You walk away with 30 ready-to-publish post concepts that are distinct from each other, platform-appropriate, and designed to drive your specific conversion goal.

Why this matters for creators

Short-form content is a volume game, but volume without structure is noise. Every generic angle you publish costs you: it trains your audience to scroll past your posts, it eats the creative energy you could spend on high-impact pieces, and it buries the posts that actually convert. This workflow externalizes the filtering effort to AI while keeping editorial judgment where it belongs — with you.

A creator publishing 3 posts per week can sustain a 10-week backlog from a single 45-minute workflow session. That means no more "what do I post today" panic. It means your content calendar is pre-validated against the metrics that matter for growth.

Limits and notes

Needs human review before publishing. The score is a filter, not a guarantee of virality. The biggest risk is false positives — angles that score well on paper but don't resonate with your specific audience. Run a 3-post A/B test on any new format before committing to a series. Also note that the LLM's novelty detection is relative to its training data, not to your specific niche. If you cover an obscure topic, manually overrule scores that seem inflated.

Related reading

More on AI workflows and productization: