How to write cold DMs on X that get replies: what the data says

A cold DM on X gets a reply when it is short, about the person rather than your product, ends on one easy question, and gets one or two follow-ups that add something. That is not a hunch. Studies covering tens of millions of cold messages put the average reply rate between 3 and 5%, and point to what the best senders, above 10%, do differently. Below: the benchmarks, X's own DM rules, five rules backed by data, and templates to copy.

No large public study measures X DMs on their own, so the figures here come from cold email and LinkedIn messages, the closest equivalents. Every figure links to its source, listed at the end.

What reply rate to expect from a cold DM

Three large studies give a realistic baseline:

The Backlinko study covers link building outreach, not sales, and dates from 2019, so use the 3 to 5% range as your baseline for prospecting. The gap between the average and the best senders is the interesting part: the same few habits show up across these datasets, and all of them carry over to X.

How DMs work on X in 2026

Before writing anything, know where your message lands and what X allows:

• Who can message whom. Someone you do not follow can DM you only if you accept messages from everyone, or if they are verified and you accept messages from verified users. Their message goes to your Requests, not your main inbox (X Help Center).
• The daily limit. An account can send 500 DMs a day, a limit X may lower temporarily when the site is busy (X limits).
• Automation. X's automation rules forbid unsolicited DMs sent in bulk or in an automated way. Automated DMs are allowed only to people who asked to be contacted or clearly showed they want to be, with a way to opt out (X automation rules).
• Identical messages. X's authenticity policy lists bulk, aggressive unsolicited DMs and identical DMs as platform manipulation (X authenticity policy).

In practice: write each message for one person, send at a human pace, and never paste the same text twice. That is also exactly what gets replies, as the five rules below show.

Rule 1: keep it short

The datasets point the same way:

• The best performers in Instantly's 2026 benchmark keep their cold emails under 80 words.
• In Gong and Outbound Squad's analysis of 28 million emails, messages of 100 words or fewer, in three or four sentences, got the most replies.
• Hunter's study of 34 million emails found the highest average reply rate, 4.5%, for emails of 20 to 39 words.
• LinkedIn's data on recruiter InMails shows messages under 400 characters getting 22% more responses than average.

One nuance from the same Hunter study: emails under 100 words averaged 3.9% against 3.7% for longer ones. Length alone does not make a message work. Short just leaves no room for the parts that hurt, like the company history and the feature list.

On X, that means two to four sentences, roughly 25 to 60 words. If it needs scrolling, it is an email.

Rule 2: write it for one person

Generic messages are easy to spot and easy to ignore. The data:

• In Hunter's 2026 report, emails with two personalized fields got 56% more replies than generic ones, 5.6% against 3.6%.
• The same report found that sequences sent to 21 to 50 people averaged 6.2% replies, against 2.4% for sequences sent to 500 people or more.
• LinkedIn recruiter InMails sent one by one got roughly 15% more responses than InMails sent in bulk (LinkedIn).

X gives you an advantage email does not: the person already told you what is on their mind. If they became a lead because they asked for a recommendation or complained about a tool, your first line writes itself. Quote their problem in their words, then ask about it.

Personalization is not "Hi FIRSTNAME, love your content". It is one line that could only have been written to this person.

Rule 3: ask about their interest, not for a meeting

The first message has one job: start a conversation. Three findings explain what kills it:

• Pitching. In Gong and Outbound Squad's data, pitching reduced reply rates by as much as 57%.
• Asking for a meeting. Gong Labs' study of 304 174 cold emails found that asking about the recipient's interest performed better than asking for a meeting, measured in meetings actually booked.
• No question at all. Emails asking one to three questions were 50% more likely to get a response than emails with none, in Boomerang's data. That study covers all kinds of email, not only cold ones, but the pattern matches the others.

Also pick the right person. Gong found that C-level executives are 30.2% less likely to reply to cold emails than other employees. On X, the person who posted about the problem is usually the one feeling it, whatever their title. Message them, not their CEO.

A good closing question is easy to answer in a few words: "Is that still a problem?", "How are you handling it today?", "Worth a look, or not a priority right now?"

Rule 4: follow up once or twice, then stop

Most replies come from the first message, but a short sequence still pays:

• In Instantly's benchmark, 58% of all replies came from the first message.
• In Hunter's 2026 data, sending three messages instead of one more than doubled total replies, 6.8% against 3.3%. Following up three times or more lowered the average again.
• Backlinko's study found that a single follow-up brought 65.8% more replies.

So write your first DM as if it were the only one, then plan two follow-ups at most, a few days apart. Each one should add something: a useful resource, a sharper question, a result from someone in their situation. "Just bumping this" adds nothing and reads like a sequence.

Rule 5: timing matters less than relevance

People spend a lot of energy on the perfect hour. The data says it barely matters:

• Hunter's 2026 report found no meaningful effect of the sending day on reply rates.
• On LinkedIn, 65% of InMail responses arrived within 24 hours, and messages sent on Saturday got 8% fewer responses than average (LinkedIn).

On X, the timing that does matter is the post's. Someone who asked for a tool this morning is choosing this week. Message while the post is fresh, and answer their reply fast. In a Harvard Business Review study of 1.25 million leads, companies that tried to reach a lead within an hour were nearly seven times as likely to qualify it as those that waited longer.

Cold DM templates for X

Replace the words in capitals. Each one stays under 50 words. The first three open with the lead's own post and end on one question; the follow-ups add something or close the loop.

They asked for a recommendation:

Saw you're looking for a CATEGORY that handles THEIR NEED. Happy to point you to a few options, ours included. What matters most to you, SPEED or PRICE?

They complained about a competitor:

Your post about COMPETITOR and THE PROBLEM hit home, we hear it a lot. Is it a dealbreaker yet, or more of an annoyance for now?

They described the problem, not a tool:

Read your thread on PROBLEM. Curious how you're handling THE HARD PART today, manually or with a tool?

First follow-up, a few days later:

One more thought on PROBLEM: ONE USEFUL TIP OR LINK. Happy to share how other ROLES handle it if that's useful.

Last follow-up:

Last note from me, I don't want to clutter your inbox. If PROBLEM becomes a priority, I'm one message away.

Mistakes that kill replies

• The same text for everyone. It gets fewer replies, and X treats identical DMs as manipulation.
• Opening with your product. Your first line is about them. Your product comes up when they ask.
• Asking for a call in the first message. Ask about their interest first.
• A wall of text. If it needs scrolling, cut it in half.
• Five follow-ups. Two, each with something new, then stop.
• Messaging the biggest names. Executives reply less, and big accounts get hundreds of requests. The person who posted the problem an hour ago will read yours.

FAQ

Scale it without the spam

Everything above works by hand: find the post, write the line, ask the question, follow up twice. It just does not scale past a few dozen leads a day. That is what Gemlead is for. It finds the people posting about your problem on X, scores them against your ideal buyer, and writes each DM from what that person actually said. Read how to find leads on X for the prospecting side, and what AI SDRs can and cannot do before you hand any of this to a machine.