AI prompts for LinkedIn posts

The truncation problem: your first two lines are the whole post

LinkedIn's feed shows roughly the first 3 lines of a post — sometimes fewer on mobile — before cutting to a "see more" link, and the majority of viewers scrolling a feed never tap through. This means the actual creative unit worth obsessing over is not the full post; it is the visible fragment, which has to work as a complete, compelling thought on its own, whether or not anyone reads further.

Testing the opening two lines in isolation, before writing the rest of the post, catches the single most common AI-generated LinkedIn failure documented by creator-economy researchers and social platforms alike: a warm-up sentence that adds nothing and burns the limited space before truncation on iOS, Android and desktop alike.

A real story beats a general lesson every time

The core input that determines whether an AI-assisted LinkedIn post reads as authentic or generic is not the prompt's phrasing — it is whether real, specific source material was supplied. A prompt asking for "a post about the importance of good leadership" with no actual story attached will produce a competent-sounding, entirely generic post, because the model has nothing specific to draw from and defaults to the statistical average of leadership content it has seen.

The instruction to state the takeaway directly, without hedging, matters because AI models trained toward balanced, cautious language often produce a takeaway so qualified it says nothing — "leadership is complicated and depends on many factors" is true and useless. A real opinion, stated plainly, is what makes a post worth engaging with.

What actually reads as "written by AI," beyond the obvious clichés

Certain phrases have become recognizable AI tells — "in today's fast-paced world," excessive em-dashes, the three-item list format repeated in every paragraph, a concluding "In conclusion" — and forbidding them in a prompt is a reasonable first step. But the deeper tell is genericness itself: a post that could have been written about any company, any leader, any situation, with the nouns swapped out and nothing else changing.

This audit catches what a forbidden-word list cannot: a post entirely free of clichéd phrases can still be hollow, and a post using a slightly awkward phrase can still be genuinely specific and worth reading.

Early engagement matters more on LinkedIn than on most platforms

LinkedIn's distribution has historically weighted early engagement — comments and reactions in the first 60-90 minutes — heavily in determining how widely a post subsequently gets shown, a pattern LinkedIn's own creator guidance and independent analysis of the platform's behavior have both described. A post that draws no engagement in that early window tends to stop being shown regardless of its underlying quality, which means the opening hook and the closing question both carry real, measurable weight beyond just reader interest.

A closing question asking "have you ever dealt with a difficult client?" invites a one-word reply or none at all. A closing question asking "what's the longest you've let a difficult client relationship run before addressing it directly?" invites a specific, real answer — and specific answers generate the kind of comment thread the platform's distribution logic rewards.

Using AI to structure a post you already know the substance of

The strongest use case for AI in LinkedIn writing is not idea generation — it is structural editing of substance you already have. Voice-to-text a rough version of a real story, or write a messy first draft, then use AI to tighten pacing, cut redundancy, and sharpen the hook without changing the underlying facts or opinion.

The explicit instruction not to add unincluded claims matters here specifically because editing prompts are where fabrication creeps in quietly — a model asked to "improve" a post will sometimes add a plausible-sounding statistic or example to strengthen the argument, which introduces exactly the kind of unverified claim that damages credibility if a reader questions it.

Formatting for the algorithm and for a scrolling reader at once

Short paragraphs, generous line breaks, and readable structure serve two purposes simultaneously: they make a post easier to skim on a phone, and they tend to correlate with stronger engagement in the platform's own observed patterns, likely because posts that are easy to read get more completed reads and more comments. This is a case where formatting for the reader and formatting for the algorithm point in the same direction.

The instruction against imposing list formatting on narrative content addresses a specific over-application of "good LinkedIn formatting" — bullet points work well for genuinely list-shaped content and read as artificial and choppy when forced onto what is actually a story.

Comment replies are a different, shorter task

Replying to comments on your own post is a distinct writing task from the post itself — shorter, more conversational, and ideally specific to what the individual commenter actually said rather than a templated thank-you. A generic "Great point, thanks for sharing!" reply reads as dismissive precisely because it could apply to any comment on any post.

Running each comment through this kind of prompt individually, rather than batch-generating generic replies, is more time-consuming but produces the specific, responsive engagement that keeps a comment thread — and by extension the post's distribution — active longer, since LinkedIn's early-engagement window (roughly the first 60-90 minutes, per the platform's own observed distribution behavior) rewards exactly this kind of back-and-forth.

For posts drawing dozens or hundreds of comments, individual replies to every single one becomes impractical, and a reasonable triage prioritizes replying personally to comments from people worth building a real professional relationship with, while a shorter, still-specific acknowledgment covers the remainder — the goal is avoiding a wholly generic reply pattern, not achieving perfect individual attention at any volume.

Disclosure and reputation: what the platform requires versus what your audience expects

LinkedIn's Professional Community Policies do not require disclosure that a post was AI-assisted, and there is no platform-level penalty for using AI in drafting. What does carry real reputational risk is a broader pattern documented in general public opinion research: a 2023 Pew Research Center survey found substantial public wariness toward AI-generated content specifically in contexts involving personal or professional trust, which is relevant to LinkedIn precisely because the platform's value proposition is built on individual professional credibility.

The practical implication is not a disclosure requirement but a substance requirement: content that is genuinely yours — your real opinion, your real experience, your real voice, even if AI helped structure it — does not carry this risk, because the trust concern is about hollow, impersonal content passed off as personal insight, not about tool usage itself.

A useful self-check before publishing: would you be comfortable if a close colleague asked "did you actually experience this, or did you just write it because it sounded good"? A post you could answer honestly and specifically to that question has cleared the bar this section is describing; one where the honest answer requires hedging suggests the underlying substance was thinner than the finished draft makes it appear.

Building a repeatable voice profile instead of prompting from scratch each time

Rather than re-explaining your tone and style preferences in every prompt, building a reusable voice reference — a short document describing sentence length preferences, whether you use humor, how you typically open and close posts, and 2-3 examples of posts that sound authentically like you — saves time and produces more consistent output than a fresh, unguided prompt each time. Both Anthropic and OpenAI's own published prompt-engineering documentation recommend saved, reusable context for any recurring task, and a personal voice guide is a direct application of that general practice.

This voice-profile prompt is worth running once and saving, rather than repeating in every subsequent post prompt — the output becomes a reusable reference that keeps future AI-assisted drafts consistent with an established voice rather than drifting toward the model's own default style each time.

Update the voice profile periodically rather than treating it as fixed forever — every 6 to 12 months is a reasonable cadence — since a professional voice evolves over time, particularly after a role change, a shift in industry focus, or simply growing more comfortable expressing an opinion publicly, and a voice guide built from posts written two years ago may no longer match how someone actually writes today.

Article-length LinkedIn content follows different rules than short posts

LinkedIn's native long-form Articles feature and its newsletter format serve a different reading pattern than the short feed post — readers who click into an article have already committed to reading, which removes the truncation pressure that governs short-post hooks but raises the bar for whether the full piece justifies that commitment. A prompt for article-length content should be told explicitly which format it is writing for, since the two call for structurally different openings.

The different opening instruction matters because a short-post-style punchy hook, stretched across an 800-word article, tends to overpromise relative to what the piece actually delivers — article readers respond better to a clearly stated scope than to a hook optimized for a truncated feed preview.

LinkedIn's newsletter format also sends a notification to subscribers on publish, unlike a standard feed post, which means article-length content reaches an audience that opted in specifically for longer-form thinking from that author — a meaningfully different, higher-intent reader than the passive-scroll audience a short post reaches through the main feed algorithm.

Carousel and document posts need a different prompt than plain text

LinkedIn's document/carousel post format — a swipeable series of slides — rewards a different structure than plain text: each slide needs to work as a near-standalone unit while building toward a connected argument, since readers can swipe at variable speed and some will only see the first few slides.

Flagging slides that need more room, rather than forcing every point into an arbitrary word count, prevents the common carousel failure of cramming a paragraph's worth of nuance into a slide format built for a single clear statement.

What to check before publishing any AI-assisted LinkedIn post

Four checks. First, the specificity test: does the post contain at least one detail — a number, a name, a date, an outcome — that could only apply to this exact situation, or could the nouns be swapped for any company's story with nothing else changing? Second, the claim check: does every statistic, example, or reference to an event trace to something real you actually supplied, rather than something the model added while "improving" the draft? Third, the voice check: read it aloud — does it sound like something you would actually say in a conversation, or does it sound like a press release?

A fourth check worth adding before any post referencing company performance, a client outcome, or a specific metric: confirm the number is current and accurate as of the September 2026 publish date or later, not a figure pulled from an earlier Q1 draft, an outdated pitch deck, or a round estimate the model treated as precise. LinkedIn posts referencing outdated or inflated figures are searchable and screenshot-able indefinitely, and a specific wrong number does more reputational damage over time than a vaguer, honest claim would have.