DMs that read like you know them.

Personalized cold DMs get replies because they prove you looked. They open on something the lead actually posted, speak to the problem they described and end on one easy question. Written by hand, each one takes minutes of research, so most outreach falls back on templates and gets ignored. Nico AI, Gemlead's AI DM writer for X, does that research for every lead in seconds and writes one message for that one person, in your voice and tied to your offer.

Below: what templates cost you, how Nico AI writes each DM and what the research says. Every figure links to its source, listed at the end.

What generic cold DMs cost you

A template saves you five minutes and costs you the lead. The numbers:

• Fewer replies. In Hunter's 2026 report, sequences sent to 21 to 50 people averaged 6.2% replies, against 2.4% for sequences sent to 500 people or more. Small and specific beats big and generic.
• A bad reputation. 73% of B2B buyers actively avoid suppliers who send irrelevant outreach, in a Gartner survey of 632 buyers. They remember your name.
• Higher expectations. 73% of customers expect companies to understand their unique needs and expectations, according to Salesforce's State of the Connected Customer, a survey of 13 020 consumers and 3 916 business buyers.
• X's rules. X's authenticity policy lists bulk, aggressive unsolicited DMs and identical DMs as platform manipulation. The same paste to a hundred people puts your account at risk.

So why does everyone still use templates? Time. Sellers devote nearly a full day of their workweek to prospecting, and 48% still say they lack the bandwidth to do adequate cold outreach (Salesforce State of Sales 2026). Personal research does not scale by hand. On X, that hurts twice: your leads posted their problem in public, and a template throws that away.

How Nico AI writes a personalized DM

Nico AI is Gemlead's outreach agent. He writes every DM in your campaigns, once Simon AI has found the lead and Dana AI has scored it.

1. Your persona is his brief. It says who buys, what they struggle with, what you offer them, the competitors they already use and the profiles to skip. Simon AI drafts it from your X account and your website. You edit it.
2. Only hot leads get a DM. Dana AI scores every lead from 0 to 100 against that persona, and campaigns write to the hot ones, 60 and up by default. Her Contact block on each hot lead says when they are usually online.
3. The deep read. Before writing, Nico AI reads the lead's bio, latest posts and activity: what they sell, what they post about, what gets in their way.
4. One DM for one person. He writes the message in your voice and ties it to your offer. When the lead was found through a post or a reply less than three days old, the DM speaks to it in one sentence. A lead found because they just followed a competitor is never told so.
5. The right version. In the Warm-up flow, an Interaction test checks whether the lead reacted to your like or your comment. If they did, they get a warmer DM. If not, a cold DM and a bridge comment. If their inbox is closed, Nico AI leaves a comment inviting them to write instead.
6. One follow-up, then stop. The campaign templates add a follow-up DM three days later. The moment the lead replies, Nico AI stops and the thread moves to Sam AI's inbox.

Each AI-written DM that is sent costs 5 credits. A fixed message you write yourself for a step costs nothing. To test angles, run several campaigns on the same persona and compare them.

Template or Nico AI: the same lead, two DMs

Say a founder posts that trial users of their app go quiet after day two, and asks how others fix it. The template:

Hi FIRSTNAME, love your content! We help SaaS founders grow revenue with our all-in-one platform. Do you have 15 minutes this week for a quick call?

What a deep read produces:

Saw your post about trial users going quiet after day two. Is it the onboarding emails, or do they never reach the first win in the app?

The second is shorter, about them, and ends on a question they can answer in five words. It could only go to that person.

Why personalized DMs work

Personalization is the variable the data keeps pointing to:

• More replies. Emails with two personalized fields got 56% more replies than generic ones, 5.6% against 3.6% (Hunter, 2026). LinkedIn recruiter InMails sent one by one got roughly 15% more responses than InMails sent in bulk (LinkedIn).
• AI can write the personal part. In a Scientific Reports paper by Matz and colleagues, messages that ChatGPT tailored to a person's psychological profile were significantly more influential than non-personalized ones, across four studies and 1 788 participants. The studies tested product ads and climate appeals, not DM reply rates, but they show the writing itself can be personal at scale.
• Interest beats a meeting request. Gong Labs' study of 304 174 cold emails found that asking about the recipient's interest performed better than asking for a meeting. Nico AI's DMs end on a question, not a calendar link.
• Fresh beats perfect. Companies that tried to reach a lead within an hour were nearly seven times as likely to qualify it as those that waited longer (Harvard Business Review, 1.25 million leads). A DM about a post from this morning lands while the problem is still on their mind.

Most of these studies measure email and LinkedIn, the closest public benchmarks: no large study measures X DMs on their own. Gemlead's own figure is a 60%+ average reply rate on its DMs, and it comes from the whole sequence, the warm-up and the timing as much as the words. The rules behind each message are in how to write cold DMs on X that get replies.

Writing DMs by hand vs with Nico AI

Who it is for

• B2B founders. You sell a SaaS, an app or a service and do the sales yourself. A lead who posts about churn, a tool they hate or a hire they cannot make gets a DM about exactly that, and you get your mornings back.
• Agencies and consultants. Your buyers describe their problem before they look for a provider. Nico AI opens on that problem, in your words, on each X account you connect: up to four on Business.
• Coaches and creators. People ask for advice in public. Keyword search finds them while the question is fresh, and the DM picks up the conversation they started.
• B2C sellers. Sports bettors asking who to follow, players complaining about their current tipster, fans replying to a rival service. With a B2C persona, Competitor comments and Keyword search bring these people in, and Nico AI writes to each one about what they said.

Getting the best DMs from Nico AI

1. Feed the persona. Nico AI only knows what you tell him. Say what you sell, to whom, what hurts before they buy and what result you deliver. A vague persona gives vague DMs.
2. Read the first 50. Check the tone and every claim before you let a campaign run on its own, as recommended in what AI SDRs can and cannot do.
3. Test two angles. Run two campaigns on the same persona with different messaging, and keep the one that gets more replies.
4. Stay inside X's limits. Set each campaign's Daily leads limit at or under what your X plan allows for message requests.

FAQ