LinkedIn DM templates that get replies (and why templates are the problem)

The reason LinkedIn DM reply rates are low is the templates. A message that works gets screenshotted, published in a listicle, and sent by ten thousand people within a quarter, at which point the recipients have learned its shape and it stops working. That cycle is why every list of “LinkedIn DM templates that get replies” is describing something that used to get replies. What survives copying is structure, not script: one specific observation only true of this person, one sentence explaining why you are in their inbox, and one ask cheap enough to say yes to. Below are six structures written that way, the real character limits, and an honest account of which parts of this nobody can measure, including us.

Do LinkedIn DM templates actually work?

They work in the way a lockpick works on a lock nobody has changed. The first thousand people to send “I noticed you are hiring three SDRs, are you finding the ramp time is the problem?” got replies. The hundred thousandth did not, because by then the recipient had received the same sentence eleven times that month.

This is not a moral argument. It is a mechanical one. Outreach templates live in a market where the returns to a tactic fall as adoption rises, and adoption rises fastest for the tactics that get published. So a template list is a leading indicator of what will stop working, and the freshest templates are the ones nobody has written up.

There is external evidence for the decay. Expandi's 2026 benchmark report, built on 13,218,869 connection requests and 6,730,447 outbound messages sent through its platform between May 2025 and April 2026, reports that the reply rate on notes attached to connection requests fell from 3.5% to 2.2% over twelve months. That is a population of automated senders, which is exactly the population that adopts templates fastest, and it is the population where you would expect decay to show up first.

So the useful question is not which template to copy. It is which parts of a message are copy-resistant.

What makes someone reply to a LinkedIn message?

Three things, in descending order of importance, and only one of them is about wording.

1. They already know who you are. A message from a name the person has seen in their feed, in a comment thread, or at an event is a different object from a message from a stranger. This is the largest single factor and it is decided before you write a word. It is also the whole subject of our piece on outreach that does not feel cold.
2. You said something only true of them. Not their industry, not their job title, not their company size. Something they wrote, shipped, argued about, or changed. This is what cannot be templated, because acquiring it requires reading.
3. The ask is cheap.“Do you have 15 minutes” is expensive. “Is that still how you think about it?” is nearly free, and it starts the conversation that leads to the 15 minutes anyway.

Everything else that gets discussed in template articles, the greeting, the emoji, the sign off, the use of first names, is noise around those three.

What is a realistic LinkedIn message reply rate?

Between about 3% and about 20%, depending on whose study you read, and the spread is not because some senders are twenty times better. It is because the studies measure different things.

A 10.4% platform average and a 19.98% agency average are not in conflict. One is what happens when software sends at volume; the other is what happens when a person researches each recipient. The gap between them is roughly the value of doing the research, which is the most useful thing in the table.

We take these apart properly, including why the denominators differ and what you should actually expect, in our meta-analysis of LinkedIn outreach benchmarks. The short version: use these to calibrate your own trend line, never as a target.

What is the difference between a DM, an InMail, and a connection note?

Three different mechanisms with three different limits, and template articles routinely conflate them. All of these come from LinkedIn's own help pages.

Note the contradiction in the InMail row. LinkedIn's two help pages give different body limits, 1,900 and 2,000 characters, and both are live. We are flagging it rather than picking one, because picking one is how bad numbers spread. In practice it does not matter, since the evidence says you should be nowhere near either.

The economics are worth stating plainly: a message to an existing connection is free and unlimited. An InMail costs a credit you paid for. If the connection request is likely to be accepted, accepting it first is strictly cheaper. InMail earns its cost when the person will not accept invitations at all, which is common at senior levels.

How long should a LinkedIn DM be?

Shorter than you want it to be. The best evidence here is LinkedIn's own, from an analysis of tens of millions of InMails sent by corporate recruiters between May 2021 and April 2022:

• The response rate for the shortest InMails, under 400 characters, ran 22% higher than average.
• The response rate for the longest, over 1,200 characters, ran 11% below average.
• InMails sent individually saw response rates roughly 15% higher than those sent in bulk.

Three caveats you should apply before treating this as universal. It is recruiter outreach, not sales. It is InMail, which arrives with a subject line and a Premium badge. And the data window closed in 2022. But the direction matches every other published finding, and the 400 character figure is a useful working ceiling. That is about two thirds of a screen on a phone, which is where most people read messages.

The bulk versus individual finding is the more interesting one. It is not measuring writing quality. It is measuring whether a human made a decision about this specific recipient, and LinkedIn can see the difference even when the text is identical.

What are the message structures worth using?

Six structures rather than six scripts. Each one names the input you have to supply yourself, which is the part a template cannot give you.

1. The response to something they published

Input required: you actually read their post.

Say what you took from it, say what you did differently because of it, stop. No ask on the first message. This is the highest-reply structure we know of and it is also the slowest, because it only works when the reaction is real. The tell that someone faked it is that they describe the topic rather than the argument.

Worked example, made up for illustration: “Your post on why your team killed the weekly forecast meeting made me look at ours. We replaced it with a Slack thread and got 40 minutes back. Curious whether it held up for you six months on.”

2. The disagreement

Input required: a real position of your own.

Disagreement gets replies because it is rare and because most people cannot resist defending a claim. It fails badly if it reads as combative or as a debate invitation. The form that works is: here is the thing you said, here is the case where I have seen the opposite, what am I missing. That last clause does the work.

3. The specific, non-obvious observation

Input required: ten minutes on their website, changelog, or job board.

“I noticed you are hiring” is not this. “You have three open roles in support and none in engineering, which usually means the last release created more tickets than expected” is. It can be wrong. Being wrong in an interesting way still gets a reply, because you have demonstrated that you looked.

4. The introduction with a named referrer

Input required: permission from the referrer.

The strongest structure in the list and the least scalable. Put the referrer's name in the first sentence, not the third. Say what they said. Do not overclaim the relationship, because the recipient will check.

5. The direct commercial message

Input required: honesty.

Sometimes you are selling and pretending otherwise is worse than saying so. Name what you do in one sentence, name the specific reason you think it is relevant to them in one sentence, and make the ask a question rather than a meeting. People respect a clean pitch more than a fake curiosity. What they do not respect is a three message ramp that ends in the same pitch.

6. The re-open

Input required: a reason the timing changed.

For conversations that died. Not “circling back.” Something has to have changed: they shipped the thing, the funding round closed, the person they mentioned left. If nothing has changed, there is no message.

What should the first message after a connection is accepted say?

This is the single highest-volume message on LinkedIn and it is almost always wasted.

The default is a pitch, sent within minutes of acceptance, often automatically. It is the most reported behaviour on the platform and it has a compounding cost: every person who does it lowers the acceptance rate for everyone who sends an invitation afterwards. The second most common option is nothing at all, which is less offensive and equally useless, because a connection you never speak to is a name in a list.

What works is a short message that closes the loop on why you connected and asks something cheap. Three components, nothing else:

1. Acknowledge the accept in half a sentence. Not a paragraph of gratitude.
2. Name the specific thing that prompted the invitation. The post, the thread, the event, the referrer. This is the part that proves you are not running a list.
3. Ask one question they can answer in a sentence.Not a meeting. Not “what does your process look like.” Something narrow enough to be answered from memory.

Then let it sit. A non-reply is not a rejection at this stage, it is a person who is busy, and you have already achieved the main thing, which is that your name is now attached to a specific context in their memory. The failure mode is treating silence as a signal to send message two, three, and four.

How should you write an InMail subject line?

InMail is the only LinkedIn message type with a subject line, and LinkedIn documents it as a required field of up to 200 characters. Most people either waste it or over-write it.

The subject is doing one job: telling the recipient whether this is a mass send. That is what they are deciding in the preview pane, before they open anything. Subjects that read like marketing lose immediately. Subjects that read like an email from a colleague get opened.

• Use their words, not yours. If they wrote about warehouse throughput, the subject is about warehouse throughput.
• Keep it under about 60 characters. You have 200 available and almost no reason to use them. Preview panes truncate, particularly on mobile.
• No question marks in the subject if the body also opens with a question. Two questions before any statement reads as a survey.
• Never put your company name in it. It converts the message into an advertisement before it is opened.
• Do not use urgency.“Quick question,” “time sensitive,” and “following up” are the three most burned openers in B2B.

The body limit is documented inconsistently by LinkedIn, at 1,900 characters on one help page and 2,000 on another, and the practical answer is that you should be under 400 either way.

What does the right message look like at each stage?

Message length and directness should track how much the person knows about you. Sending a stage-four message to a stage-one contact is what most bad outreach actually is.

The mistake almost everyone makes is compressing five stages into one message. The mistake experienced sellers make is stretching one stage across five messages. Neither is a template problem.

What should you never send in a LinkedIn DM?

• An immediate pitch after a connection is accepted. The single most reported behaviour on LinkedIn. It teaches people to stop accepting invitations, which makes the whole channel worse for everyone including you.
• A calendar link in the first message. It asks a stranger to schedule time with you before they know what you want.
• A voice note to someone you have never spoken to. It cannot be skimmed, which is the entire problem.
• “Quick question” as an opener. It is never a quick question, and everyone knows.
• Anything auto-sent.Beyond the reply rate, automated messaging is explicitly named in LinkedIn's User Agreement section 8.2 and is a documented trigger for account restrictions. We quote the clauses in our breakdown of LinkedIn automation and the terms of service.

How should you follow up without becoming a pest?

Once, with new information, then stop. That is the whole policy.

The argument for long sequences is that persistence pays, and in aggregate across thousands of prospects it can, on a spreadsheet. The cost is not on the spreadsheet. Ignored, pending, and spam-marked messages feed the same recipient-behaviour signals LinkedIn names as a trigger for restricting invitations, and a reputation for sequencing is durable in a way a single meeting is not.

A follow-up that works looks like: “You mentioned in your post last week that the migration finished. That was the thing I was asking about, so ignore my earlier message if it is no longer relevant.” It carries new information, it gives permission to decline, and it does not pretend the first message never happened.

What can nobody measure here, including us?

This is the section most template articles skip, so we will be specific about our own limits.

Our dataset is 12,988 public LinkedIn posts from 65 creators, scored by engagement rate. It contains reactions, comments, follower counts, and the visible text of each post up to the “see more” fold. It contains no messages, no InMails, and no reply data. We cannot tell you which DM opener performs best, and any article claiming to have measured that at scale is either working from one vendor's platform, one agency's campaigns, or nothing at all.

What our data does speak to is the thing that happens before the message. Across the cohort, the top decile of posts had a median of 72 comments against 5 in the bottom half, a far wider gap than reactions at 235 against 80. Comments are named individuals starting conversations with you, in public, unprompted. That is the same output outreach is trying to manufacture, arriving without a template and without a credit. Which is why we think the highest-value work on this channel is usually upstream of the inbox.

How do you write your own message in ten minutes?

1. Read their last five posts and their last five comments. Comments are more revealing than posts, because people argue in comments.
2. Write down one sentence you could only write about this person. If you cannot find one in ten minutes, they are the wrong person to message, not a person who needs a better template.
3. Write the message around that sentence. One line of context, the sentence, one question. Nothing else.
4. Cut it to under 400 characters. Delete the greeting, the credential, and any clause explaining why you are reaching out. The message is the reason.
5. Send it as an individual message.Not through a tool, not as part of a batch. LinkedIn's own data shows a 15% response gap between individually sent and bulk InMails.

Ten minutes per person is the actual cost of the higher reply rate, and it is why agency numbers beat platform numbers. If ten minutes per prospect is not economic for your deal size, that is a real finding, and the answer is to change the channel rather than to send worse messages faster. Our guide to LinkedIn lead generation works through what to do instead.

The honest summary on LinkedIn DM templates

Templates are a shortcut around the only expensive input in outreach, which is knowing something specific about the person you are messaging. That shortcut works right up until enough people take it, and then it inverts, because a recognisable template is now a signal that no thought went into the message. Use structures instead. Keep it under 400 characters. Make the ask cheap. Follow up once, with something new. And treat every reply rate you read, including the three we cited above, as a description of somebody else's population.

If you want the upstream version of this problem solved, start with our analysis of 34,012 LinkedIn posts and the practical follow-on in how commenting builds recognition before you ever send a message.