AI Tools & Software

How to Write Cold Emails With AI That People Actually Answer

Abstract blue chat graphic with the label Write Cold Emails With Ai

You can write cold emails with AI in minutes, but speed is not the problem with most cold outreach. The problem is that most cold emails are generic, and AI makes generic emails easier to send at scale. The emails that get answered are short, relevant and clearly written by someone who did a little homework. This guide shows how to use AI for the research, drafting and follow-ups without sounding like a template, plus the legal basics from the FTC's CAN-SPAM guidance, checked on 15 October 2026. It is written for small businesses, freelancers and founders doing their own outreach.

What makes a cold email work

Before any prompt, get the fundamentals right:

  • The right person. A great email to the wrong person still fails.
  • A real reason. Why them, why now? A recent hire, launch, review, job post or problem you can solve.
  • One clear ask. A short call, a yes or no question, or permission to send something useful.
  • Brevity. If it does not fit on a phone screen, cut it.
  • Honesty. No fake "Re:" subject lines, no pretend familiarity.

AI helps with each of these, as long as you give it real information.

Step 1: Research with AI, then verify

Use AI to speed up research, not replace it:

  1. Summarize the prospect's website or recent posts into three bullet points about what they do and what they care about.
  2. Find a trigger: a new product, location, hire, award or public challenge.
  3. Draft a hypothesis of the problem you might solve for them.

Then check every fact yourself. An email that mentions something wrong is worse than one that mentions nothing. For deeper prospect research, see our guide to AI competitor analysis, which uses similar techniques.

Step 2: Write cold emails with AI using a clear prompt

Give the AI the context it needs and strict constraints:

> Write a cold email under 90 words. > Me: [who you are, what you offer, one proof point]. > Prospect: [name, role, company, one verified detail]. > Reason for writing: [the trigger]. > Ask: [one simple question]. > Tone: plain, friendly, no hype, no exclamation marks. > Do not invent any facts about the prospect. Do not use the phrases "I hope this finds you well" or "quick question".

Generate three versions, choose the most natural, and edit it until it sounds like you.

Subject lines

Ask for five subject lines under six words that describe the email honestly. Avoid clickbait; CAN-SPAM requires subject lines that are not misleading about the content.

Step 3: Personalize the first line properly

The first line decides whether the email is read. Good first lines reference something specific and relevant:

  • "Saw you opened a second location in Austin last month."
  • "Your post about slow invoice approvals matched what we hear from other agencies."

Weak first lines flatter or fake familiarity. Ask AI for several first-line options based on your verified detail, and pick one you can stand behind.

Step 4: Plan follow-ups

Most replies come after a follow-up. Plan two or three:

  1. Follow-up 1 (3–4 days later): a short nudge with one new piece of value, such as a relevant example.
  2. Follow-up 2 (a week later): a different angle or a resource.
  3. Break-up email: a polite close that makes it easy to say "not now".

Ask AI to draft each as a reply to the original thread, under 60 words, adding something new each time.

Step 5: Follow the rules

The FTC's CAN-SPAM compliance guide sets out the US basics for commercial email, and it makes no exception for business-to-business messages:

  • Accurate header information: the "From", "To" and "Reply-To" details must identify you.
  • Honest subject lines that reflect the content.
  • Identify the message as an ad where it is one.
  • Include a valid physical postal address.
  • Tell recipients how to opt out, and honor requests promptly.

The FTC lists penalties of up to $53,088 for each email in violation. Other countries, including in the EU and Canada, have stricter consent rules, so check before emailing internationally. This is general information, not legal advice.

Step 6: Protect your deliverability

Even a perfect email fails if it lands in spam:

  • Send in small volumes from a properly configured domain.
  • Verify addresses before sending.
  • Avoid sending the identical text to hundreds of people; real personalization helps here too.
  • Watch replies and bounces, and stop when signals turn bad.

Tools that help

  • A general assistant such as ChatGPT or Claude for research summaries and drafts. Our ChatGPT Projects guide shows how to store your offer, tone rules and examples so every draft starts from the same brief.
  • An email assistant in your inbox for faster replies. See our guide to an AI email assistant.
  • A CRM to track who you contacted and when, so follow-ups do not slip.

Example: before and after

Generic AI draft: "I hope this email finds you well. Our innovative solutions help businesses like yours achieve their goals. Would you be open to a quick call?"

Edited version: "Saw you're hiring two more bookkeepers in Denver. Most firms we work with hit that point when month-end reconciliation starts eating weekends. We set up automated bank matching for three Denver firms this year. Worth a 15-minute look next week?"

The second version is still AI-assisted. The difference is the verified trigger, the specific problem and the honest proof point you supplied.

Measuring results

Track reply rate, positive reply rate and meetings booked per 100 emails. Change one variable at a time, such as the first line, offer or ask, and compare. Over a few weeks, your own numbers will tell you which AI-assisted approach works for your audience.

Related reading

For more tools grouped by job, browse our AI tools by profession guide.

FAQ

Is it legal to send cold emails?

In the US, commercial email is governed by the CAN-SPAM Act. The FTC's compliance guide requires accurate header information, non-deceptive subject lines, a valid physical postal address and a way to opt out, and it makes no exception for business-to-business email. Other countries have stricter consent rules.

Can AI write a whole cold email campaign?

AI can draft the emails, but the research, the offer and the final review should be yours. Fully automated, unreviewed outreach tends to sound generic and damages your sender reputation.

How long should a cold email be?

Short. Most effective cold emails fit on a phone screen: a relevant opening line, one clear reason you are writing, and one easy question.

How many follow-ups should I send?

Two or three spaced over a couple of weeks is common. Each should add something new, and every message must honor opt-out requests.

What should I never let AI do in cold emails?

Invent facts about the prospect, fake familiarity, make claims you cannot support, or remove the opt-out and address details required by law.

Conclusion

To write cold emails with AI that get answered, use AI to speed up research and drafting, then do the human parts: verify facts, write a real first line, make one clear ask and edit until it sounds like you. Follow CAN-SPAM basics, protect your sending reputation, and measure replies rather than volume.

Sources and methodology

Legal basics come from the FTC's "CAN-SPAM Act: A Compliance Guide for Business" (https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business), checked 15 October 2026. This article is general information, not legal advice. No reply-rate statistics are claimed.