How to write LinkedIn messages with AI

A good LinkedIn message is short, specific and about the other person. For a connection request that means one line on why you are writing and one line on what you want, short enough to fit LinkedIn's character limit for a note. For a reply to a recruiter it means answering their question first, then saying what you need to know. For a follow-up it means one reminder, one easy next step, and nothing that sounds like a complaint.

AI can write all three well, but only if you give it the facts it cannot guess: who the person is, how you found them, and what you actually want. Without those, every model falls back on the same polite, general message that recipients have learnt to skip. This guide gives you a structure for each type, a short example, and a way to use AI so the result still reads like you wrote it.

Why most AI LinkedIn messages get ignored

People on LinkedIn receive a lot of messages, and many of them are written by a tool. The ones that give themselves away share the same habits:

• A compliment with nothing behind it. "I was really impressed by your profile" says the writer did not read it.
• The ask hidden at the end. Three sentences about yourself before the reader learns why you wrote.
• Too long. A LinkedIn message is read on a phone, between other things. Anything past four or five lines gets skimmed.
• Stock phrases. "I hope this message finds you well", "I'd love to pick your brain", "let's connect and explore synergies".
• Invented detail. When a model is not given a reason to write, it tends to make one up, such as a post you never read.

The fix for all five is the same: give the model the real reason, and ask for less.

Connection requests

A connection request note has a character limit, so the structure is tight.

1. Why them. The specific thing: a talk you saw, a mutual contact, a post, the same field.
2. Why now, or what you want. Keep it small and honest. "I'd like to follow your work" is fine.
3. Stop. No sign-off paragraph, no list of your achievements.

Example:

Hi Priya, I heard your talk on warehouse routing at the logistics meetup last week and I'm working on something similar at a smaller scale. Would be glad to be connected and follow what you post.

Two sentences. It names a real event, says who the writer is in half a line, and asks for something easy. If you have no real reason to connect, a blank request is better than a made-up one.

Replying to a recruiter

Recruiter messages usually ask one thing: are you interested, or are you free to talk. Answer that first.

1. The answer. Interested, not interested, or interested with a condition.
2. The one or two things you need to know. Salary range, location, whether it is remote, what the team does.
3. A next step. A time that suits you, or "send me the job description".

Example, when you are open to it:

Thanks for getting in touch, Mark. I'm open to hearing more. Before we set up a call, could you share the salary range and whether the role is hybrid or fully in the office? If it fits, I'm free Tuesday or Thursday afternoon.

If you are not interested, one or two lines is enough, and it is worth keeping the door open. There is a whole page on saying no kindly: how to decline politely.

Following up

A follow-up on LinkedIn works like a follow-up email, only shorter. Leave it five to seven working days, then send one message.

1. Remind them in one line what you sent before, without restating it.
2. Make it easy to answer. A yes or no question, or a specific time.
3. Give them a way out. It lowers the pressure and people answer more readily.

Example:

Hi Tom, just bringing my note from last week back up. Would a 15-minute call on Friday work to talk about the design role? If the timing's wrong, no problem at all.

Avoid "just checking in" on its own, and never follow up more than twice. For the longer email version, see how to write a follow-up email.

How to write LinkedIn messages with AI

Any capable model can do this if you brief it properly. What goes into the prompt matters far more than which model you pick.

1. Paste their message or the relevant bit of their profile. Only what matters: their role, the post, the job they are hiring for.
2. Say who they are to you. A stranger, a former colleague, a recruiter who wrote first.
3. Give every fact. Where you met, what you want, your availability. Tell it to use only these.
4. Set the length. "Under 200 characters" for a connection note, "four lines at most" for a message.
5. Ask for one version and only the text. No options, no explanation.
6. Read it as the recipient would. Cut anything you would not say to their face.

A prompt you can adapt:

"Write a LinkedIn connection note to [name], [their role]. The reason I'm writing: [real reason]. What I want: [small ask]. Only use these facts, do not add any. Under 200 characters, plain and friendly, no compliments, no 'I hope this finds you well'. Give me only the note."

The last line matters more than it looks. Most general assistants wrap the message in commentary and offer three versions, and you then have to pick through them. Asking for only the note saves that step.

Doing it with Hemmingway

Hemmingway is built for exactly this kind of writing. Its model, Hemmingway-1, was trained to give you the message itself, one version, the length the situation calls for, and to stick to the facts you gave it. In our own tests of everyday messages and emails, CommunicationBench, it scored highest of the models we compared, and on our own Human-Likeness test a judge picked its replies as written by a person more often than any other model's. How that was measured.

For LinkedIn, paste the message or the profile details into the web app or the desktop app, say what you want to say, and copy the result back. If you tell it who someone is in People, it remembers that the next time you write to them.

Try HemmingwayDownload the app

It can still be wrong while sounding sure, so check any name, date or job title before you send.

A short checklist before you send

• Does the first line say why you are writing?
• Is every detail true, including the ones the AI wrote?
• Could you read it aloud to them without wincing?
• Is it short enough to read on a phone without scrolling?
• Is there one clear thing for them to do next?

Common questions

Is it okay to use AI for LinkedIn messages?

Yes, as long as what you send is true and you would stand behind it. The problem people notice is not that a tool helped, it is a generic message with invented detail. Give the AI the real facts and read every message before it goes out.

How long should a LinkedIn connection message be?

LinkedIn caps the note on a connection request at a few hundred characters, and on some accounts less, so one or two sentences is right. Say why you are writing to that person and what you would like. Anything longer is better sent as a message once you are connected.

How do I reply to a recruiter on LinkedIn if I'm not interested?

Thank them, say plainly that you are not looking right now, and leave the door open if you mean it. Two lines is enough. You do not need to give a reason.

How long should I wait before following up on LinkedIn?

Five to seven working days is a reasonable gap for a first follow-up. Send one short reminder with an easy question, and stop after a second follow-up if there is still no answer.

Can AI write a LinkedIn message that doesn't sound like AI?

Yes, if you give it specific facts, ask for a short message, and ban the stock phrases. A model trained for everyday writing, such as Hemmingway-1, starts closer to how a person writes, but reading it aloud before you send still catches the rest.