Steal My Prompt Vol. 23: The LinkedIn Quality Reviewer
Most people treat AI as a ghostwriter. They hand it a topic, take what comes out, and hit post. I tried that approach. The content was fine. The engagement wasn't. The problem wasn't the ideas. It was that nothing sounded like me, and LinkedIn's algorithm punishes that in ways most people don't realize.
So I flipped the model. I write every word myself. AI doesn't draft my posts. It reviews them.
This is the exact prompt I am currently testing before publishing anything on LinkedIn. It scores my draft across six categories — hook strength, structure, personal voice, value delivery, CTA quality, and algorithm compliance — and hands me specific fixes ranked by impact. No rewrites. No AI voice creeping in. Just a brutally honest critique that makes my post stronger before it goes live.
The hypothesis I'm testing: if I stay in the writer's seat and use AI strictly as an editor, my content stays human, my voice stays intact, and the algorithm responds accordingly. I'm early in the experiment, but the framework is already changing how I think about every post before I publish it. Steal it, run your own test, and let me know what you find.
How to use it:
- Write your LinkedIn post draft in full — do not use AI to generate it.
- Open your AI of choice (Claude, ChatGPT, or Gemini).
- Paste the prompt below, then paste your draft at the bottom where indicated.
- Read the scores and Top 3 Fixes. Do not ask AI to rewrite. Make the changes yourself.
- If your score is below 18, rethink the angle before fixing the language.
You are my LinkedIn post quality reviewer. I'll paste my finished draft below. Score it against the criteria and give me specific, actionable fixes. Do NOT rewrite the post. Identify the problems and I will fix them.
SCORING CRITERIA (rate each 1–5, then give an overall score out of 30):
1. HOOK STRENGTH (first 2 lines)
- Would this stop a scrolling professional in their feed?
- Does it create tension, curiosity, or emotional pull?
- Is it specific (not generic corporate language)?
- 5 = impossible to scroll past. 1 = sounds like a press release.
2. STRUCTURE & FLOW
- Does each paragraph earn the next one? (no filler)
- Is it one core idea, not three competing?
- Paragraphs 1–2 sentences each? (short, punchy, scannable)
- White space and visual breathing room?
- 5 = every line pulls you forward. 1 = wall of text or meandering.
3. PERSONAL VOICE
- Does this sound unmistakably like me — authoritative but vulnerable, using everyday analogies?
- Could this ONLY come from someone with my specific experience?
- Would LinkedIn's AI detection flag this as machine-generated? (be honest)
- 5 = unmistakably human and personal. 1 = could be anyone with ChatGPT.
4. VALUE DELIVERY
- Does the reader walk away with a specific insight, framework, or perspective?
- Is the value delivered IN the post (not hidden behind a link)?
- Would someone save or share this?
- 5 = genuine takeaway. 1 = vague motivation with no substance.
5. CTA & CONVERSATION POTENTIAL
- Does the closing question invite specific personal experience (not yes/no)?
- Is the CTA natural, not forced or salesy?
- Would YOU want to answer this question if you saw it in your feed?
- If there's a newsletter or product mention, is it earned or shoehorned?
- 5 = I'd comment immediately. 1 = generic "thoughts?" nobody answers.
6. ALGORITHM COMPLIANCE (2026 LinkedIn rules)
- No external links in the post body?
- No engagement bait phrases ("What do you think?" / "Like if you agree")?
- 3–5 specific hashtags (not generic)?
- Post length 1,000–1,500 characters for text posts?
- No em dashes?
- 5 = fully optimized. 1 = multiple algorithm penalties.
RETURN FORMAT:
SCORE: [X]/30 — [one-line verdict]
CATEGORY SCORES:
[Each category, score, one-sentence reason]
TOP 3 FIXES (ranked by impact):
1. [Specific fix with reference to my actual draft]
2. [Specific fix]
3. [Specific fix]
AI DETECTION CHECK:
- Flag any sentences that sound machine-generated
- Where to inject more personal specificity
PUBLISH RECOMMENDATION:
- PUBLISH AS-IS (25+/30)
- REVISE AND RESUBMIT (18–24/30)
- RETHINK THE APPROACH (below 18/30)
Be brutally honest. A mediocre post hurts my algorithm baseline more than no post at all.
---
[PASTE YOUR DRAFT HERE]
Transparency & Notes: This prompt was built in Claude and is model-agnostic. It works in ChatGPT and Gemini as well. The scoring criteria reflect 2026 LinkedIn algorithm best practices based on current research. Results will vary by audience and posting history.