How to cold DM on Twitter (without getting flagged)
March 24, 2026
My first 20 Twitter DMs got 1 reply. It was “no thanks.”
My last 20 got 7 replies, 3 conversations, and 1 booked call. Same offer. Same type of person. Different approach.
The difference was a 10-day warm-up sequence I now run before every DM. Here’s the full breakdown.
Why Twitter DMs are different
Twitter DMs sit in a different part of the brain than email or LinkedIn. People check Twitter to scroll, not to work. That means your DM is competing with memes, threads, and notifications – not calendar invites and status reports.
The bar for attention is low, but the tolerance for anything salesy is near zero. A DM that reads like a LinkedIn InMail will get screenshot-roasted on the timeline. A DM that reads like a human being starts conversations.
Some platform realities:
- DMs from people you don’t follow land in a “requests” folder. Most people never check it.
- If they follow you back, your DM goes to their primary inbox. This is everything.
- Twitter’s spam detection flags accounts that send the same message to multiple people. Vary your copy.
- The daily DM limit for newer accounts hovers around 20-30 per day. Older accounts get more room. Hit the limit and you’re locked for 24 hours.
The 10-day warm-up sequence
This is the day-by-day sequence I use before DMing anyone on Twitter. It takes time. That’s the point.
Days 1-2: Follow them. Like 2-3 of their recent posts. Not a burst – spread it across the day. You’re registering your avatar in their notifications.
Days 3-5: Reply to a post with something useful. Not “great thread” – an actual observation, a related datapoint, or a respectful pushback. One reply that adds value beats 10 likes.
Days 6-8: Quote-tweet or reply to a second post. At this point, they’ve seen your name 4-6 times. You’re moving from stranger to familiar. If they reply to your reply, even better – you’re in a conversation before the DM ever happens.
Days 9-10: Send the DM. By now, they’ve seen your name in their notifications. You’re not cold. You’re lukewarm. And lukewarm gets a 31% reply rate vs 8% for true cold.
The DM templates
Template 1: The thread reference
DM:
Your thread on [topic] – the part about [specific point] –
that's exactly what I've been testing at [your company/project].
Getting different results though. Mind if I share? No pitch.
This works because it’s specific and offers value. You’re not asking for their time – you’re offering information. I’ve used this template 34 times. 9 replies. 26% reply rate.
Template 2: The content compliment with a hook
DM:
Been following your stuff on [topic] for a few weeks.
Your take on [specific take] changed how I'm approaching
[your related problem]. Quick question – are you still
seeing [specific result they mentioned]?
The question at the end is doing the work. It’s easy to answer, relevant to them, and shows you actually read what they posted. Not a compliment for the sake of a compliment – a compliment that leads somewhere.
Template 3: The direct ask
DM:
Hey – building [your project] and you know [their area of
expertise] better than most people I follow. One question:
[specific, narrow question]. Totally fine if you're too
busy. Appreciate the content either way.
Use this only after the full warm-up. The direct ask works when you’ve already established presence – otherwise it reads as presumptuous. When the warm-up is done, this gets a 22% reply rate. Without it, 5%.
What gets you flagged
Twitter’s spam system looks for patterns. Here’s what triggers it:
- Sending the same message (or near-identical) to 5+ people in a short window
- DMing 10+ people who don’t follow you back within a few hours
- Including links in DMs to people who haven’t followed you. Twitter suppresses these hard.
- New accounts (under 30 days) sending more than 10 DMs per day
I got locked out for 24 hours in month 1. The mistake: I sent a batch of 12 DMs in one afternoon with the same opener. Varied the middle and end, but the first line was identical. Twitter caught it.
The fix is simple. Write each DM individually. Use templates as skeletons, but change the specific references every time. It takes longer. It works.
What doesn’t work
A few patterns I tested that consistently failed:
The cold pitch. “Hey, I built [product]. Think it’d be useful for you.” 47 sends, 2 replies – both negative. Don’t mention your product by name in the first DM. Ever.
The voice note. Twitter lets you send audio DMs. I tried 15 of them. 1 reply. People don’t want to listen to a stranger’s voice in their DM inbox. It feels invasive.
The mega-DM. Anything over 4-5 lines. If the DM requires scrolling in the mobile preview, it’s dead. I tested long vs short on similar prospects – short outperformed by 3x on reply rate.
The numbers
Over 3 months, I sent 186 Twitter DMs for Scouter and Prospect Organic.
- Read rate: 52%
- Reply rate (all DMs): 18%
- Reply rate (with warm-up): 28%
- Reply rate (true cold, no warm-up): 6%
- Average time to reply: 3.4 hours
- DMs that led to a call: 11
The read rate tells you something important. Nearly half your DMs aren’t even getting opened. That’s why the warm-up matters – it gets you out of the request folder and into the primary inbox.
For comparison, cold email typically gives you more volume but less engagement per send. LinkedIn has a higher read rate but lower reply rate. Twitter is the highest-friction channel to start, but the conversations it creates are the most genuine.
The system
Here’s what I actually do each week:
- Identify 5-8 people I want to reach
- Start the warm-up sequence for each (Days 1-10)
- Send 5-8 DMs per week to people who’ve completed the warm-up
- Track reads and replies in a simple spreadsheet
That’s 20-30 DMs per month. Not a high volume play. This is a precision channel, not a scale channel. If you need volume, use email. If you need quality conversations with specific people, use Twitter DMs with the warm-up.
The warm-up is the whole game. Skip it and you’re shouting into a request folder nobody checks. Do it and you’re starting a conversation that was already halfway started.