Cold outreach has a bad reputation, and most of it is deserved. Inboxes are full of messages that open with flattery, list generic services, and ask for 'a quick call' with a stranger. AI makes it easy to send hundreds of these, which is exactly why they fail: recipients can tell when a message was written for everyone. The freelancers who get replies from cold outreach do something different. They personalise one specific detail, and they use AI to speed up the research, not to fake the relationship.

Use the first prompt for research. Give the AI the prospect's website or recent post and ask: 'Summarise what this business sells, who it sells to, and one recent change or project visible on their site. List only facts found in the text, no guesses.' That summary becomes your personalisation material. When you reference something real, like a new service page or a recent launch, your message stops sounding like a blast.

Use the second prompt for the draft itself, and keep it short. A useful prompt reads: 'Write a cold outreach email under 120 words to a small business owner. Open with one observation about their business: [paste observation]. Say I help similar businesses with [service]. Name one specific outcome I could improve for them. End with one low-pressure question, not a call request.' The low-pressure ending matters: a question like 'is this something you are working on right now' gets more replies than a calendar link from a stranger.

Use a third prompt for the follow-up, and write it at the same time as the first message. Most replies come after the second or third touch, but most freelancers never send one because they run out of things to say. Prompt: 'Write a follow-up email to a prospect who did not reply. Add one useful idea for their business, no guilt, under 80 words.' A follow-up that adds value instead of asking 'just bumping this up' is a follow-up worth sending.

The rule that holds all three prompts together is simple: personalise one line every time, by hand. Let AI draft the structure and the research summary, but write the observation sentence yourself, in your own words, from something you actually noticed. That single hand-written line is what separates your message from the hundred AI blasts in the same inbox. Speed is the AI's job. Sounding like a real person is yours.

Finally, keep your volume honest. Ten well-researched, personalised messages a week will outperform a hundred generic ones, and they will not damage your reputation or get your domain flagged. Track replies by message type, keep the prompts that earn conversations, and retire the ones that do not. Cold outreach is a numbers game only in the sense that better messages are the numbers that count.

The best cold email is sometimes the second contact, not the first. Before you write, spend a few minutes becoming a familiar name: engage thoughtfully with a prospect's public post, share something of theirs with a genuine comment, or answer a question they asked publicly. This is not flattery or manipulation; it is the online equivalent of introducing yourself at an event before pitching. When your email arrives, the name is already faintly familiar, and familiar names get opened.

Use AI to support this too, but keep it honest. Prompt for a short, specific comment on a prospect's article or post: 'Write a two-sentence comment on this post that adds one observation or question. No praise for its own sake, no pitch.' If the comment would embarrass you with your name on it, do not post it. Warm outreach works because it is real attention, and real attention is scarce. Ten minutes of genuine engagement beats ten generic emails, and it builds a reputation that cold blasts slowly destroy.