Best AI Prompts for LinkedIn

Why does LinkedIn punish AI-sounding posts so aggressively?

LinkedIn's product team made a public decision in late 2025 to down-rank posts flagged as likely AI-generated. In October 2025, LinkedIn's VP of Global Editorial Laura Lorenzetti told Entrepreneur the platform was rolling out measures to detect and limit the reach of AI-generated comments. Accounts that posted high-volume AI content started seeing sharp drops in distribution. Third-party LinkedIn-marketing blogs put the impact at roughly 30% lower reach for flagged posts, though LinkedIn itself has not published a specific number. The mechanism is silent. There is no notification. Posts that previously reached 5,000 impressions reach 200. Accounts that previously drove inbound recruiter messages stop hearing from anyone. The penalty compounds across posts; one flagged piece does not just lose its own reach, it lowers the baseline distribution for everything that follows.

The fix is not to abandon AI. It is to use AI for the parts that benefit from structure (research, framing, draft generation, voice analysis) and to deliver in your own words. The 7 prompts below are designed for that workflow. Every one has the LinkedIn-specific constraints baked in: no em-dashes in every paragraph, no triple-adjective phrases, no rhetorical-question openers, no “the truth is…”, no “let me share something…”. These are the exact patterns LinkedIn's detection layer was tuned for.

⚠️ The signal you cannot afford to miss

If your LinkedIn reach has dropped 60-90% in the last 6 months without an obvious explanation, the most likely cause is AI-pattern flagging on recent posts. The fix starts with the editing pass: How to Edit AI Out of Your Writing. Run it on your last 10 posts before publishing the next one; the cleanup compounds.

What are the seven LinkedIn prompts?

1. The profile-rewrite prompt

Here is my current LinkedIn profile: [paste headline, about, experience].

Here is what I want to be known for now: [paste 1-2 sentences].

Here is my actual resume (the ground truth on what I have done): [paste].

Rewrite the profile to position me for [target audience: recruiters / clients / peers / partners]:

– Headline (220 chars max, opens with a specific claim not a job title)

– About section (2,000 chars max, first sentence is a hook a stranger would read, no “passionate about” or “results-driven”)

– 3 most recent experience bullets reframed for this positioning

For each rewrite, mark what changed and why. Do not invent experience I do not have.

Why this works: The headline is the highest-impact real estate on the whole profile. Most people fill it with their current job title, which is the lowest-information line possible (the recruiter already saw their job title in the search results). The “opens with a specific claim” constraint is what makes the headline earn the click. The “mark what changed and why” output is the audit trail that catches AI-padding before it ships.

2. The weekly post template

This week I want to post about: [topic, in your own words, 2-3 sentences].

The specific lesson or insight I want to share: [paste your raw thinking].

My audience: [recruiters / peers / clients / industry / mixed].

Draft 3 versions of the post:

– Version A: opens with a specific scene or moment (not a question, not “here's what I learned this week”)

– Version B: opens with a candid admission (something I did wrong, something I changed my mind about)

– Version C: opens with a counterintuitive claim (something that pushes against received wisdom in this space)

Each version: 4 short paragraphs, 600 characters max total. No em-dashes. No bullet point lists. No “tag a friend who needs this” close. Includes one piece of specific detail only I would know.

Tell me which version is most likely to be flagged as AI-generated and why. Suggest the rewrite.

Why this works: Three versions force a real choice between framings. The “tell me which is most likely flagged” self-check is the LinkedIn-specific safeguard; AI can identify its own patterns when explicitly asked. The “one piece of specific detail only I would know” line is what separates a generic AI-sounding post from a post that reads like an actual professional reflecting on actual work.

3. The comment-framework prompt

Here is a LinkedIn post in my industry that I want to comment on: [paste].

Here is what I actually think about this post: [paste your raw take, even if it's 2 sentences].

Draft three comment options:

– A: Adds a specific data point or example from my own experience

– B: Constructively disagrees with one point in the post, gracefully

– C: Asks a sharper question that invites the original author to go deeper

Each comment: 2-3 sentences max. No “great post!” opener. No “thanks for sharing!” close. Sounds like a real professional, not a network builder.

Tell me which one is most likely to earn a reply from the original author.

Why this works: Comments are the lowest-effort, highest-return move on LinkedIn. A useful comment on a high-reach post puts your voice in front of that author's entire audience. The “great post” opener marks you as a network-builder bot; the specific-data or constructive-disagreement comment is what gets your account read by people who matter. The “most likely to earn a reply” output is the calibration that turns commenting from broadcast into conversation.

4. The recommendation request

I am asking [name] for a LinkedIn recommendation.

Our relationship: [paste: what we worked on together, when, what they saw of my work].

What I want to be known for in their words: [paste].

Draft a recommendation request:

– 5 sentences max

– Opens with what I am asking, briefly

– Names the specific projects or moments they would most likely recall

– Suggests 2-3 specific qualities they could speak to (so they have a starting structure if they need one)

– Offers to reciprocate if appropriate

– Includes a graceful out (they can say no without weirdness)

Do NOT pre-write the recommendation for them; the prompt is for asking, not ghost-writing their voice.

Why this works: The most common failure mode in recommendation requests is the over-prescription pattern, where the requester writes most of the recommendation and asks the recommender to “just review and post.” That produces recommendations that all sound identical and get devalued. The “starting structure” framing gives them the prompt without writing the words.

5. The headline experiment

My current LinkedIn headline: [paste].

The 3 audiences I want it to land with, in priority order: [list].

My actual differentiator (in plain words, not corporate-speak): [paste].

Generate 10 alternative headlines:

– 5 that lead with what I do (verb-first)

– 3 that lead with who I serve (audience-first)

– 2 that lead with a specific claim or number (proof-first)

Each headline: under 220 characters, no triple-adjective phrases, no “passionate about”, no “results-driven”, no “thought leader”.

For each, mark which audience it best fits and which one I should test for the next 2 weeks.

Why this works: The headline is the only LinkedIn surface that is testable. Swap, observe, swap again. Ten alternatives gives you real range. The “verb-first / audience-first / proof-first” categorization is the structural taxonomy that most people skip and that determines which kind of inbound the headline produces.

6. The DM template

I want to DM [name, role, company].

One specific reason I am reaching out (not “I'd love to connect”): [paste].

One thing I have to offer them or one specific ask: [paste].

Draft three DM options:

– A: 3 sentences, leads with the reason, ends with a yes/no question

– B: 5 sentences, leads with a piece of value (a link, a referral, an insight)

– C: 2 sentences, the candid one-shot ask

For each, mark when (kind of recipient + kind of relationship) it works best.

Hard constraint: no “Hope this finds you well.” No “I came across your profile and was impressed.” No “I think you and I could create incredible value together.” Those are the DM equivalents of spam.

Why this works: LinkedIn DMs have the lowest reply rate of any outreach channel in 2026 because they are flooded with templated AI messages. The hard-constraint forbidden-phrase list is the single most actionable part of any LinkedIn DM advice. The candid 2-sentence variant (option C) is consistently the highest-reply-rate variant for recipients who are tired of long pitches; it works because it respects their time.

7. The rebrand sequence (career-change posts)

I am changing my career direction from [old] to [new].

Why I am making the change (in 2-3 sentences of plain truth): [paste].

The audience I want to reposition with: [recruiters in new field / current network / both].

Design a 6-post sequence that re-introduces me in the new direction:

– Post 1: the announcement (what I am changing and why, not a manifesto)

– Post 2: the bridge (what I am bringing forward from the old work)

– Post 3: the learning curve (what I am learning, candid about the gap)

– Post 4: the first project / public artifact in the new direction

– Post 5: the reflection (one thing I see differently now)

– Post 6: the ask (specific opportunities I am looking for)

Each post: 300-500 characters, posted 5-7 days apart. No “thrilled to announce”. No “embarking on a journey”.

Why this works: Career-change announcements get ignored when they come as a single “I am pivoting!” post, but earn distribution when they unfold as a sequence that shows the work happening. The 6-post arc is paced for LinkedIn's algorithm (5-7 days between posts is enough cadence to keep you visible without saturating). The “no thrilled to announce” is the line that separates a sincere career-change post from the templated AI-generated version.

What is the worst thing you can do with AI on LinkedIn?

Post AI drafts verbatim. The LinkedIn detector flags pattern-matches at scale. Every post should go through a human editing pass (your own voice, your own examples, your own rhythm).
Schedule 5 AI-drafted posts a week. High-volume AI output is the single fastest way to get throttled. Quality at lower frequency outperforms volume.
Use AI to draft sympathy posts (death, illness, layoffs). Sincerity is the whole point of those posts. AI involvement, even editing, is the wrong call here.
Comment with AI on every post in your feed. LinkedIn now flags this pattern too. Comment less frequently with more substance.
Skip the edit on lead-magnet posts. The post that drives the most inbound traffic is also the most exposed if it pattern-matches; the consequence of a flagged high-reach post is a baseline-distribution drop that takes weeks to recover from.

What if your LinkedIn is the primary source of new business?

If LinkedIn drives leads or recruiting outcomes, the 7 prompts are a candidate for the ladder. Save them as Claude skills. Bundle them as linkedin-presence that runs the weekly cycle: generate a post draft on Monday based on your week, draft 3 high-value comments Tuesday-Wednesday on industry posts in your feed, prep one DM on Thursday. The setup is one Sunday afternoon; the time saved over a quarter is multiple hours per week. The bigger benefit is consistency: your LinkedIn presence stops being a sporadic effort and becomes a system.

📊 The Prompt-to-Workflow Ladder

Tier 1: the prompts (this post). Tier 2: the skill (one per LinkedIn surface). Tier 3: the plugin (linkedin-presence bundle). Tier 4: the workflow (Monday-post draft + 3 comments + 1 DM auto-prepared every week). When to climb →

What are common questions about AI on LinkedIn?

How do I know if I've been throttled?

The signal is silent. Watch your impressions per post over a 30-day rolling window. A drop of 60-90% with no change in posting cadence or topic is the most common indicator. LinkedIn does not notify you; you only see it in the analytics.

Can I recover from a throttle?

Yes, but slowly. The throttle lifts as your recent posts pass the detection layer cleanly. Six to eight clean posts (no AI tells, human-edited) over four weeks is the typical recovery pattern reported by third-party LinkedIn-marketing communities; LinkedIn itself has not published recovery guidance. Posting through the throttle with more AI-pattern content makes it worse.

Should I disclose AI use in my posts?

Not for posts you edit and rewrite. Disclosure becomes important when the content is substantially AI-generated and the topic is one where authorship matters (e.g., expert opinion, original research, lived-experience claims). Most LinkedIn posts do not require disclosure if you have done the editing pass.

What about LinkedIn newsletters and articles?

Same rules, higher stakes. Long-form LinkedIn content is more exposed to the AI-pattern detection layer because there is more text to flag. The editing pass for long-form is mandatory, not optional.

Where do these prompts come from?

They are the LinkedIn-presence section of the larger AI Prompt Library. The Library has over 500 prompts across 33+ categories, including the full career-visibility stack: LinkedIn, personal newsletter, conference speaking, podcast guesting, and the executive-update prompts for being known in your company without being on social.

Sources to read next?

✏️ Before you post

The single most important LinkedIn quality move in 2026 is the AI-tells edit before publishing. The full 29-pattern catalog is in our cornerstone guide. For anyone posting weekly or more, the open-source Humanizer skill (Claude Code / OpenCode, MIT license, 20K+ stars) is the fastest way to run all 29 patterns against every draft before it ships. Our pre-publish prompt and 5-pass checklist: How to Edit AI Out of Your Writing →

🎯

The AI Prompt Library · $39

over 500 tested prompts including the full LinkedIn stack

The seven LinkedIn prompts above are a free preview. The full Library has the career-visibility stack: LinkedIn newsletters, conference speaker pitches, podcast guesting outreach, awards / press submissions, and the executive-update prompts.

Get the Library →

🤝

1-on-1 Custom AI Tutorial with James · $99

Audit your LinkedIn with me, then rebuild

A 1-hour private call. We audit your last 30 days of posts for AI-pattern signals, rewrite your headline and About section live using prompt 1, and you walk out with a weekly LinkedIn cadence that the algorithm will actually distribute.

Book Tutorial →

Get Smarter About AI Every Morning

Free daily newsletter. Built for people who want to use AI well, not chase every model.

Free forever. Unsubscribe anytime.

Gemini 4 Argon: Who Gets It, When

ChatGPT Dots: What They Do

ChatGPT Voice Can Now Use Plugins