AI prompts for LinkedIn posts that sound like you

LinkedIn AI prompts produce pasteable copy only when you load four things the model cannot guess: role and audience variables, writing samples, field character limits, and a banned-word list. Skip those and you get a generic third-person bio. The templates below run as written in ChatGPT, Claude, or Gemini.

What makes a LinkedIn AI prompt actually work?

A LinkedIn prompt works when the model knows more about you than it knows about LinkedIn. A bare instruction gives the model nothing about you, so it fills the gaps with generic profile language. Five components move the output toward yours:

• Fill-in variables: Every prompt below carries bracketed slots like [TARGET_ROLE] and [AUDIENCE] so the same template serves a job seeker, a founder and a demand gen director without rewriting.
• Named model and workspace: The template states where it runs, because a prompt pasted into a fresh chat and the same prompt run inside a Project with your voice file loaded produce different drafts.
• Voice samples: Two or three paragraphs you wrote, plus your best-performing posts, give the model a specific target to imitate.
• Character limits: The prompt states the ceiling for the field it writes. Connection notes cap at 200 characters and the About section runs to 2,600 characters. LinkedIn’s Help pages give no number for the headline, but a member-authored 2025 Pulse post lists 220 characters on desktop and 240 on mobile, so write to 220.
• De-cliché instruction: A list of words the model may not use, plus a first-person requirement, strips out much of the stock phrasing before you touch the draft.

Paste “write a LinkedIn About section for a growth marketer” with nothing else and the model has no profile details to distinguish the draft from generic copy. Add two paragraphs of your own writing and a first-person instruction with the banned-word list, and you direct the same model to open on a thing you did.

You build voice by loading a structured profile, sample library, and reusable project instructions every session. Point each template at that file through the [VOICE_PROFILE] slot. The anti-cliché section shows a filled-in version. Build that file with an AI brand voice prompt, then drop it into an AI content workflow you run every week.

Which LinkedIn jobs do these AI prompts cover?

Each row below is a job, the prompt pattern that handles it, and the model pick. The picks rest on documented workspace features, explained in the model section, since no controlled head-to-head on LinkedIn copy exists in the research base:

Can one prompt rewrite a whole LinkedIn profile?

The master prompt runs as an intake interview before it writes anything, which is the step that separates a profile draft you can paste from one you throw away. The model asks for your inputs and voice samples, then produces headline, About, experience entries and a skills list in one pass with the limits applied. Run it inside a Project or workspace where your voice file already sits.

You are a LinkedIn profile editor. You write from an intake interview, never from assumptions.

STEP 1. Intake. Ask me the questions below one at a time. Wait for each answer before asking the next. Where I have pre-filled a value, confirm it and move on.

- Target role or roles I want to be found for: [TARGET_ROLE]
- Industry, company stage and function: [INDUSTRY], [COMPANY_STAGE], [FUNCTION]
- The people I want contacting me after they read this: [AUDIENCE]
- Three outcomes I am proudest of, each with a number, a timeframe and the constraint I worked under: [OUTCOME_1], [OUTCOME_2], [OUTCOME_3]
- My last three roles, with title, company, dates and the two things each role focused on: [ROLE_1], [ROLE_2], [ROLE_3]
- Two paragraphs I wrote that sound like me (an email, a post, a doc): [VOICE_SAMPLE_1], [VOICE_SAMPLE_2]
- Words and phrases I never use: [BANNED_WORDS]
- Anything I will not claim, even if it would help: [OFF_LIMITS]

STEP 2. Draft. After the interview, produce all four sections together:

1. HEADLINE. One line, 220 characters maximum. Lead with [TARGET_ROLE] vocabulary a recruiter would type, then one specific value statement with a number from my outcomes. No pipes stacked with adjectives.

2. ABOUT. 2,600 characters maximum, first person throughout. Structure: open on one concrete thing I did (two sentences), then the pattern behind it (one short paragraph), then three proof points with numbers, then one line on who should message me and about what. Match the sentence length and rhythm of my voice samples.

3. EXPERIENCE. For each role, two to four lines in Challenge, Action, Result form. Each Result line carries a number. Keep each description under 2,000 characters.

4. SKILLS. Ten skills, ordered by how often they appear in job descriptions for [TARGET_ROLE]. Use the exact phrasing a recruiter would search, not my internal terms for the same work.

STEP 3. Audit. List every sentence in the draft that could appear on a stranger's profile unchanged. Rewrite each one with a detail from my interview answers or flag that I need to supply one.

Rules for all sections: no words from [BANNED_WORDS], no third person, no adjective without a number or an example behind it, no claims beyond what I gave you in the interview.

The audit step is the part most people skip, and it is the reason their drafts still read as generic after a decent interview. It instructs the model to look for filler it might otherwise leave in the draft.

How do you prompt for a LinkedIn headline?

Headline prompts should generate variants around the role keywords a recruiter types, with a single value statement attached, rather than one polished line. You pick from eight, then edit the winner. Recruiter title filters read your Experience section titles, and skills searches read the Skills section plus keywords in the sections where skills usually appear, so the headline’s job is partly search and partly the first thing a human reads.

Using [VOICE_PROFILE], write 8 LinkedIn headline variants for someone targeting [TARGET_ROLE] in [INDUSTRY].

Constraints:
- 220 characters maximum per variant. Show the character count after each.
- Each variant leads with a title or role phrase a recruiter would type into search: [TARGET_ROLE], [ADJACENT_TITLE_1], [ADJACENT_TITLE_2].
- Each variant includes exactly one value statement drawn from these outcomes: [OUTCOME_1], [OUTCOME_2].
- Vary the structure across the eight: two role-first, two outcome-first, two that name the audience I serve ([AUDIENCE]), two that name the problem I solve.
- No words from [BANNED_WORDS]. No stacked adjectives. No emoji.

After the eight, tell me which two would surface best for a recruiter searching [TARGET_ROLE] and why, based on term placement alone.

Write to 220, the stricter of the two figures.

How do you prompt for a LinkedIn About section?

The About prompt explicitly requires first person, a story-then-proof structure and a cliché ban. The 2,600-character ceiling is LinkedIn’s own published figure. Open on the strongest two sentences, since a reader may stop there.

Using [VOICE_PROFILE], write my LinkedIn About section.

Inputs:
- Who I am and what I do now: [CURRENT_ROLE_ONE_LINE]
- One specific thing I did that explains how I work: [SIGNATURE_STORY], including the constraint, what I did and the number at the end
- Three proof points with numbers: [PROOF_1], [PROOF_2], [PROOF_3]
- Who should message me and what about: [AUDIENCE], [REASON_TO_MESSAGE]

Structure, in order:
1. Two sentences that open on the signature story. No preamble about my years of experience.
2. One short paragraph on the pattern behind that story, in plain language.
3. Three proof points, one sentence each, numbers included.
4. One closing line naming who should reach out and why. No "let's connect".

Constraints:
- First person throughout. Never refer to me by name or as "he", "she" or "they".
- 2,600 characters maximum. Show the count.
- Match the sentence length and vocabulary of my voice samples in [VOICE_PROFILE].
- No words from [BANNED_WORDS]. No sentence that would be true of most people in my role.

How do you prompt for experience bullets?

Experience prompts work as a reusable instruction you paste once per role, built on the Challenge, Action, Result shape. Bullets on LinkedIn are typed characters inside the single Description field, so the prompt controls the formatting. The 2,000-character description figure comes from the same member-authored Pulse post as the headline limit, since LinkedIn’s Help article for Experience gives no number.

Convert my notes for one role into LinkedIn experience lines using Challenge, Action, Result.

Role: [TITLE] at [COMPANY], [DATES]
Raw notes: [NOTES_ON_WHAT_I_DID]
Numbers I can defend: [METRICS]

Output: two to four lines. Each line follows this shape in one sentence of 30 words or fewer:
- Challenge: the situation or constraint when I arrived
- Action: the specific thing I did, with a verb that names the work (built, cut, renegotiated, hired, shipped)
- Result: the number, and the timeframe

Rules:
- Every line ends on a Result with a figure from [METRICS]. If a line has no number, mark it [NEEDS NUMBER] rather than inventing one.
- Total description under 2,000 characters.
- No words from [BANNED_WORDS]. No "responsible for". No "helped".
- Match the register in [VOICE_PROFILE].

Without the [NEEDS NUMBER] flag, a model may fill a missing result with a plausible-sounding percentage, and LinkedIn holds you responsible for the accuracy of anything a model helped you write.

How do you prompt for LinkedIn keywords and search?

Keyword prompts take a job description in and return the terms a recruiter would search, mapped to the profile field where each term belongs. LinkedIn states that adding extra keywords does not automatically improve your appearance in search, so the prompt’s job is alignment of vocabulary, with no stuffing.

Here is a job description for the role I want: [JOB_DESCRIPTION]
Here is my current profile text: [HEADLINE], [ABOUT], [EXPERIENCE_TITLES], [SKILLS_LIST]

Step 1. Extract from the job description the 15 terms a recruiter would most likely search for this role: job titles, skills, tools and certifications. Rank them by how many times the description repeats or emphasizes each one.

Step 2. For each term, mark whether it appears in my profile exactly, appears as a synonym, or is missing.

Step 3. For each synonym or missing term, recommend one field to place it in (Experience title, Skills section, About, or headline) and write the sentence or title where it would sit. Use each term once. Do not add a term I cannot substantiate from my experience.

Step 4. Flag any term in my current profile that is internal jargon a recruiter outside my company would never type.

The recruiter-screening simulation is the second prompt, and it exists because recruiters now have a tool that pastes the job description directly. LinkedIn’s AI-Assisted Search lets a recruiter paste a job description or intake notes instead of building a Boolean string, and LinkedIn’s January 2026 engineering post describes people search as reading intent rather than matching keywords. The model cannot see LinkedIn’s ranker, so the simulation finds gaps between the job description’s vocabulary and yours, which is the part you control.

Act as a recruiter who has pasted this job description into a search tool: [JOB_DESCRIPTION]

Here are two candidate profiles. Profile A: [CURRENT_PROFILE]. Profile B: [REVISED_PROFILE].

Rank the two profiles against the description. For each, list the required qualifications it matches, the preferred qualifications it matches, and the ones it misses. Quote the exact profile text that produced each match. Then tell me the single change to Profile B that would close the largest remaining gap.

How do you prompt for posts, hooks, and calendars?

Post prompts need your best-performing posts as input so the model can learn your voice rather than LinkedIn’s house style. On the format question, carousels and native documents lead across the benchmarks: Socialinsider’s 2026 study of company pages put native documents at 7.00% engagement by impressions in 2025 against 4.50% for text. LinkedIn’s own guidance for AI visibility is to keep posts between 200 and 300 words and publish at least two to three times per week.

Run the hook prompt on an existing draft so the model rewrites only the opening.

Here is a post draft: [DRAFT]
Here are my five best-performing posts: [BEST_POSTS]

Write 10 opening lines for the draft. Keep each under roughly 150 characters so the opening stays compact. Vary across these types: a specific number, a decision I made, a thing I was wrong about, a question I get asked, a contrast between two approaches I tested. Match the rhythm of [BEST_POSTS]. No words from [BANNED_WORDS]. No "unpopular opinion". No emoji.

The post prompt then writes the body to LinkedIn’s 200–300 word guidance.

Using [VOICE_PROFILE] and [BEST_POSTS], write a LinkedIn post about [TOPIC] for [AUDIENCE].

Structure: hook in roughly the first 150 characters, one concrete example with a number, the mechanism behind it in plain language, and one line that invites a specific reply (a question about the reader's own number, not "thoughts?").

Constraints: 200 to 300 words. One idea per paragraph, paragraphs of one to three sentences. No words from [BANNED_WORDS]. No lesson list at the end. No hashtags unless I give you them.

The carousel prompt outputs slides for a native document upload.

Turn this post into a native document carousel: [POST]

Output 8 to 12 slides. Slide 1 is a title of eight words or fewer plus one subtitle. Each middle slide holds one idea in 25 words or fewer, with a bold headline line and one supporting line. The final slide states the single action to take next and who I am in one line. Mark which slides need a chart or screenshot.

The 30-day calendar prompt produces the plan, and it needs your post history to avoid repeating what already ran.

Using [BEST_POSTS], [TOPICS_I_OWN] and [AUDIENCE], build a 30-day LinkedIn calendar at three posts per week.

For each post: date, format (text, native document, image, video), the topic, the angle in one sentence, and which of my past posts it builds on or must avoid repeating. Weight the mix toward native documents and image posts. Each week includes one post that reuses a top-performing past topic from a new angle. Output as a table.

Feed the model your top posts by saves and reposts. Richard van der Blom’s research found saves and reposts resurface posts later, so those are the posts still working for you.

How do you prompt for outreach and recommendations?

Outreach prompts stay personal when they carry a named [SHARED_CONTEXT] slot the model must use, and they fail when that slot is empty. Connection notes cap at 200 characters, and the same page states that free members can personalize up to three requests per month while Premium members can personalize all of them. InMail subjects run to 200 characters, and LinkedIn’s own pages disagree on the body limit, so write to the 1,900-character figure to be safe.

Write a LinkedIn connection note to [NAME], [THEIR_TITLE] at [THEIR_COMPANY].

Shared context I have with them: [SHARED_CONTEXT]
Why I want to connect, in one honest sentence: [REASON]

Rules: 200 characters maximum, show the count. Open on [SHARED_CONTEXT], not on who I am. No "I came across your profile". No ask beyond connecting. If [SHARED_CONTEXT] is empty, stop and tell me to find one.

The InMail prompt keeps the [SHARED_CONTEXT] rule and uses the longer body limit.

Write a LinkedIn InMail to [NAME], [THEIR_TITLE] at [THEIR_COMPANY].

Shared context: [SHARED_CONTEXT]
What I want from them: [SPECIFIC_ASK]
What they get: [WHAT_THEY_GET]

Output a subject line of three words or fewer and a body under 1,900 characters. The body opens on [SHARED_CONTEXT], states [SPECIFIC_ASK] in the second paragraph with a specific time or decision, and closes in one line. Use [VOICE_PROFILE] for register. No words from [BANNED_WORDS]. No "I hope this finds you well".

The recommendation request prompt reverses the usual flow and drafts the recommendation itself, so the person you ask can edit rather than start from blank.

Draft a recommendation request to [NAME], who was my [THEIR_RELATIONSHIP] at [COMPANY] from [DATES].

Include: a two-sentence note asking for the recommendation, and a 60-to-90-word draft recommendation they can edit, written in their likely voice, covering [PROJECT_WE_SHARED] and one specific result ([RESULT]). Make clear in the note that the draft is optional and they should change anything that does not match their memory.

How do you prompt for a headshot and banner?

Headshot prompts should specify photographic parameters rather than adjectives, and the output must remain your likeness. LinkedIn’s community policies bar profile photos that are not your likeness, which rules out a generated face that looks like an idealized stranger. Direct the shot with lighting and crop instructions, and keep your own photos as the source.bar profile photos that are not your likeness, which rules out a generated face that looks like an idealized stranger. Use AI to direct the shot and control its lighting and crop, with your own photos as the source.

Profile photos upload at 400×400 pixels minimum to 7680×4320 maximum, PNG or JPG, 8MB, and the cropping tool is square. Personal banners are 1584×396 pixels at 8MB. LinkedIn notes that photographs work better than logo graphics, and that the banner sits behind your profile photo.

Using my uploaded photos as the subject, produce a professional headshot.

Camera: 85mm lens, shallow depth of field, subject in sharp focus with a softly blurred background.
Framing: head and shoulders, eyes in the upper third, 1:1 crop, 2000×2000 pixels.
Lighting: soft key light from the front-left, gentle fill, no hard shadows under the eyes.
Wardrobe and setting: [INDUSTRY_PERSONA], for example "Series B SaaS operator, plain dark crewneck, out-of-focus office with warm practical lights".
Expression: neutral-to-warm, closed mouth or slight smile, direct eye contact.

Do not alter my facial features, skin tone, age or hair beyond lighting and color correction. Output three variants with different background blur strengths.

Run the banner prompt after the headshot so the two match in color temperature.

Design a LinkedIn personal banner, 1584×396 pixels, 4:1 aspect ratio.

Subject: [WHAT_I_DO_ONE_LINE]. Audience: [AUDIENCE].
Style: photographic, not illustrated. Match the color temperature of the headshot I uploaded.
Layout: keep the left 30% low-detail because the profile photo overlaps it. Place one short text line (six words maximum) in the right half: [BANNER_LINE].
No stock-photo handshakes, no city skylines, no abstract gradients.

What anti-cliché block do you append to every prompt?

The add-on is a block you paste under any prompt in this article, and it carries three things: the banned-word list, a filled-in voice profile, and the character limits for each field. Build it once, store it in your Project or workspace, and reference it as [VOICE_PROFILE] and [BANNED_WORDS] everywhere else. Below is a sample filled in for a B2B growth leader. Swap the values for yours.

=== VOICE PROFILE ===
Role: VP Growth Marketing, B2B SaaS, Series B.
Audience: CMOs and heads of demand gen at Series A to C companies, plus recruiters filling VP Growth roles.
Register: direct, operator-to-operator, first person, short declarative sentences mixed with one longer explanatory one per paragraph. Dry humor allowed once per piece. No exclamation marks.
Vocabulary I use: pipeline, payback period, attribution, cohort, ICP, PLG, sales-led, CAC.
Vocabulary I avoid: synergy, thought leadership, growth hacking, journey, north star.
Sample sentence 1: "We killed the webinar program in Q2 because the payback period had drifted past 14 months and nobody could tell me why."
Sample sentence 2: "Organic was 4% of new ARR when I started. Eighteen months later it was 31%, and most of the lift came from fixing twelve pages, not writing two hundred."

=== BANNED WORDS AND PHRASES ===
The usual filler: delve, game-changer, game-changing, passionate about, results-driven, thought leader, synergy, journey, empower, elevate, cutting-edge, innovative, dynamic, proven track record, the fast-paced-world opener, let's connect, excited to announce, humbled, I help [X] do [Y] as a headline formula, any sentence that begins with "As a".

=== LINKEDIN FIELD LIMITS ===
Headline: 220 characters (240 on mobile, write to 220).
About: 2,600 characters.
Experience description: 2,000 characters per role.
Connection note: 200 characters.
InMail: subject 200 characters, body 1,900 characters.
Post: 200 to 300 words, hook in roughly the first 150 characters.

=== OUTPUT RULES ===
First person only. Every adjective needs a number or example behind it. Flag any sentence that would be true of most people in my role as [GENERIC] instead of leaving it in. Show character counts for every limited field.

I’d argue the vocabulary-I-use line does more work than the banned list. A banned list removes the worst words. A used list tells the model which specific nouns to reach for when it would otherwise reach for an abstraction.

What does raw AI output look like versus the edit?

The examples below run one sample profile through three surfaces, with the before copy, the prompt, a typical raw output and the edited final. The raw outputs are illustrative of what these prompts typically return, not logged runs, and the numbers are placeholders.

Headline

• Before: “Marketing Leader Passionate About Growth Helping B2B Companies Scale”
• Prompt: The headline template above with [TARGET_ROLE] set to VP Growth Marketing, [INDUSTRY] to B2B SaaS and [OUTCOME_1] to the organic ARR figure from the voice profile.
• Typical raw output: A typical top variant reads “Results-Driven VP Growth Marketing Driving Pipeline Through Data-Backed Strategy Building High-Performing Teams”, which clears the character count and fails the banned list twice.
• Edited final: “VP Growth Marketing, B2B SaaS Took organic from 4% to 31% of new ARR in 18 months Hiring senior demand gen leads”. The edit kept the role phrase for search, replaced three adjectives with one number, and ended on what the reader can do.

About

• Before: The section opened in the third person with a line about a seasoned marketing executive with over a decade of experience driving growth.
• Prompt: The About template with the voice profile loaded and the signature story set to the webinar cancellation.
• Typical raw output: A typical draft opens correctly on the story, then drifts by paragraph three into a list of competencies (“I bring expertise in demand generation, lifecycle marketing, and cross-functional leadership”).
• Edited final: The edit cut that paragraph, moved the three proof points up, and rewrote the close from “Let’s connect to discuss how I can help your team” to “If your organic pipeline share is under 10%, message me and tell me the number.”

Post

• Before: A 600-word reflection on marketing attribution with a lesson list at the end.
• Prompt: The post template with [BEST_POSTS] set to the sample’s five most-saved posts and [TOPIC] set to killing the webinar program.
• Typical raw output: A typical draft lands around 240 words with a usable hook, appends hashtags, then closes on “What’s your take on attribution? Share below!”
• Edited final: The edit kept the body, cut the hashtags the model added anyway, and replaced the close with “What is your payback period on webinars right now? If you don’t know, that was my problem too.”

Across all three, the prompt handles structure and limits, including vocabulary, and the human edit should focus on the close, where models may default to generic invitations.

Which other prompt angles are worth trying?

Three more patterns extend the same system without new context. Comment prompts take a post you want to engage with plus your voice profile and return a two-sentence comment that adds a specific detail or a number from your own work, with the instruction that any comment the model could have written about a different post gets cut. Cold DM sequences take [SHARED_CONTEXT], [SPECIFIC_ASK] and a three-touch cadence, with each message under 300 characters and the third message closing the loop rather than following up again.

The angle with the longest payoff is the reusable project file. Store the voice profile, the banned list, your five best posts and your target job descriptions in one place and point every prompt at it. ChatGPT Projects are available on every plan with per-project instructions and file uploads, and they are the safer home than a custom GPT, because personal accounts can no longer create new GPTs and OpenAI has scheduled their retirement. Claude Projects hold the same instructions and files, and Gemini’s equivalent is in transition: Skills are replacing Gems, with Gems support for personal accounts ending in November 2026. Build the file as plain text so it moves between all three.

Store the voice profile in a shared project file so the next run starts from the same inputs. For a weekly working example of this kind of operational prompt work, subscribe to the newsletter.

Which model should you run these LinkedIn prompts in?

Pick the model by the workspace feature the task needs, since no controlled head-to-head on LinkedIn copy exists in the sources behind this article. Anthropic’s current guidance is to start with Claude Opus 5.5 for most work, using Sonnet 5.5 as the faster, lower-cost option and Fable 5.1 for demanding reasoning. OpenAI’s flagship is GPT-6 Astra, introduced September 3, 2026, with GPT-6 Sol added later that month for paid ChatGPT plans. Google’s Gemini 3.8 Flash is available to AI Pro and Ultra subscribers, and 3.1 Pro remains the newest Pro-tier model. Model names change often, so the prompt templates stay model-agnostic, and only the table and this section name current versions.

The task-by-task picks:

• Profile rewrite: Run the master prompt in Claude Opus 5.5 inside a Project, because the intake interview plus voice file plus three roles of notes is the longest-context task on the list, and Opus 5.5 is Anthropic’s recommended starting model for most work.
• Posts and calendars: Run them in a ChatGPT Project with your past posts uploaded, because Projects exist on every plan including free, and, if you turn on the project-only memory setting, project memory draws only on conversations inside that project.
• Outreach and experience bullets: Use Sonnet 5.5 or GPT-6 Sol. These are short-output tasks where the voice file matters more than reasoning depth. Use Sonnet 5.5 or GPT-5.6 Sol. These are short-output tasks where the voice file matters more than reasoning depth.
• Gemini: Use it when your source material already sits in Drive, and plan to migrate the setup to Skills as Gems wind down.

Taplio sits in a different category. It describes itself as an all-in-one LinkedIn growth tool that writes posts in your voice, schedules them at your best time and tracks engagement, and it connects to Claude and ChatGPT through MCP. AI writing is on its Growth and Pro plans only. On output quality, one Product Hunt reviewer reported that the AI drafts were often irrelevant and that ChatGPT gave better results directly, on a 4.1/5 average from 14 reviews. Treat Taplio as the scheduling and analytics layer and run the writing through the templates here.

To settle the question for your own voice, run the headline and About prompts through all three models with the same voice file and score each draft on four lines: banned-word count, number of sentences the audit step flagged as generic, characters over limit, and how many edits you made before pasting. The model with the lowest total on your material is your model.