How to Use AI for Cold Email That Gets Replies

AI helps you write, personalize, and follow up on cold email faster, but it does not fix a bad list or weak deliverability. Use AI (like ChatGPT) to draft and tailor short, relevant messages, then rely on tools like Instantly, Smartlead, and Apollo to clean data, warm inboxes, and send. Replies come from relevance plus authentication, not volume.

Disclosure: some links below are affiliate links. If you sign up through them we may earn a commission at no extra cost to you. It never changes our picks. See our affiliate disclosure.

Key takeaways

• AI speeds up writing and personalization, but a clean, verified list and proper authentication (SPF, DKIM, DMARC) matter more for getting replies.
• Google and Yahoo tightened bulk sender rules in February 2024: authenticate your domain, keep spam complaints under 0.3%, and offer one-click unsubscribe.
• Use ChatGPT to draft and vary emails, then personalize the first line or two with real research, not generic filler.
• Warm up new sending inboxes for one to two weeks before you send campaigns, and keep daily volume conservative.
• Cold email in the US must follow CAN-SPAM: a real physical address, a clear opt-out, and you must honor opt-outs within 10 business days.

Step 1: Build a clean list

Replies start with the right people, not clever copy. Before you write a word, define who you are targeting: job title, company size, industry, and region. A tight list of 200 relevant contacts beats 5,000 random ones, because irrelevant sends drive spam complaints, and complaints wreck your deliverability.

Pull verified contacts from a B2B database. Apollo combines a large contact database (reportedly 265M+ contacts) with filtering by title, seniority, company size, and intent signals, and it includes email verification. Instantly also offers a built-in lead database so you can find and verify contacts in one place. Whichever you use, verify every address to cut bounces. High bounce rates signal a low-quality list to inbox providers and can get your domain flagged fast. For current plans and limits, check current pricing on the tool's site.

Step 2: Write the email with AI

Cold email is short by design: a relevant reason for reaching out, one clear point of value, and a low-friction ask. AI is genuinely good at drafting this quickly and generating variations so you avoid sending identical text to thousands of inboxes, which providers can flag as spammy.

Use ChatGPT to draft, then edit hard. The output is a starting point, not a finished email. Keep it under about 120 words, cut buzzwords, and make the ask specific.

A short example prompt:

You are writing a cold email to a [job title] at a [industry] company with about [X] employees. Goal: book a 15-minute call about [specific problem you solve]. Write under 90 words, plain language, no hype, one clear question at the end. Give me three subject line options under 40 characters.

Then rewrite the winner in your own voice so it does not read like generic AI text. Reviewers and recipients both notice filler phrases, so trim anything that could apply to any company.

Step 3: Personalize at scale

Generic personalization ("I loved your work!") is worse than none, because it reads as automated. The goal is a genuine, specific first line or two that shows you did a small amount of homework: a recent product launch, a role they are hiring for, or a problem common to their industry.

AI helps here when you feed it real inputs. Paste a prospect's LinkedIn summary or company page text into ChatGPT and ask for one factual, non-flattering observation tied to your offer. Do this in small batches so you can sanity-check the output, because AI will invent details if you let it, and a made-up fact destroys trust instantly.

For sending platforms, Smartlead and Instantly support custom variables and spintax so each message pulls in the right name, company, and custom line. Some tools also offer AI personalization built in. Keep merge fields simple and always send yourself a test to catch broken variables like "Hi ."

Step 4: Warm up and send

This step decides whether your carefully written emails reach the inbox or the spam folder. First, authenticate your sending domain with SPF, DKIM, and DMARC. Since February 2024, Google and Yahoo require bulk senders to use SPF and DKIM plus a DMARC policy of at least p=none, and they expect spam complaint rates to stay well under 0.3% (aim for under 0.1%). One-click unsubscribe is also required for bulk senders.

New inboxes need a warm-up period, usually one to two weeks, so providers learn your domain sends real mail. Instantly and Smartlead both include automated warm-up that simulates normal email activity, and Smartlead's approach mimics human sending patterns and matches senders to recipients by provider. Send from a separate domain (not your primary company domain), keep daily volume per inbox low (often 20 to 40 emails), and ramp up slowly. Volume without reputation just burns domains.

Step 5: Follow up

Most replies come from follow-ups, not the first email. A short sequence of two to four messages, spaced a few business days apart, consistently outperforms a single send, because timing and inbox noise mean your first email is often missed rather than rejected.

Keep follow-ups even shorter than the original, and add a new angle each time rather than "just bumping this." AI is useful for drafting these variations quickly: ask ChatGPT for three follow-up options that reference the original without repeating it. Sending platforms like Instantly, Smartlead, and Apollo automate the sequence and stop it automatically when someone replies.

Stay compliant the whole way. US cold email falls under CAN-SPAM, which requires a valid physical postal address in the message, a clear opt-out, honest subject lines and headers, and honoring opt-out requests within 10 business days. Penalties can reach tens of thousands of dollars per violating email, so treat compliance as part of the workflow, not an afterthought.

A simple AI cold email workflow

Put it together and the process is repeatable:

1. Build a tight, verified list with Apollo or Instantly. Relevance first.
2. Draft short emails with ChatGPT, then edit them into plain, human language.
3. Personalize the opening line with one real, verified detail per prospect.
4. Authenticate your domain, warm up inboxes for one to two weeks, and send low volume from a separate domain.
5. Follow up two to four times with new angles, automate the sequence, and honor every opt-out.

AI removes the slow parts (writing, varying, and personalizing), while the sending tools handle deliverability and scale. Neither replaces the fundamentals: a good list, an honest offer, and respect for the inbox. Get those right and replies follow.

Related reading: Best AI Lead Generation Tools and ChatGPT Prompts for Small Business.