AI writing tools aren’t new. But when ChatGPT showed up in December 2022, everything changed. Oh man, did everything change.

This isn’t a mere AI writing tool. It can understand complex prompts and deliver remarkable — downright frightening — results. It uses well-structured sentences and writes with confidence and logic. It can even take stabs at writing humor.

So when I saw ChatGPT in action, I had two thoughts.

One, my entire career is ruined and I’d better learn how to weld or something because it’s like five months away from stealing my job and maybe my wife.

And two, I wonder how this works for cold email.

I ran several tests using ChatGPT to figure out what it can and can’t do for cold emailers.

Here are five great ways ChatGPT (and its future descendants) can aid your cold email copywriting.

And… there are also five big ways ChatGPT alone won’t be able to take over your entire cold email operation. You can (and should) read up on those here as well.

ChatGPT Cold Email: Table of Contents

• 5 Ways ChatGPT Can Aid in Your Email Copywriting
• 5 Crucial Cold Email Elements Where You Can’t Rely on AI
• ChatGPT and Cold Email: Takeaways and Next Steps

5 Ways ChatGPT Can Help You with Cold Email Copywriting

Here are five ways you can start using ChatGPT immediately in your cold email copywriting.

1. Write the first draft of your email copy (in different styles)

I’m not going to bury the lede here: ChatGPT can absolutely handle the first drafts for most cold emails.

No more staring at a blinking cursor wondering how to dive in.

For this experiment, I decided I would be a lead gen agency promoting our services to real estate agents in Bethesda, Maryland.

My first prompt to ChatGPT was:

write a cold email promoting my lead generation agency's services to real estate agents in Bethesda, Maryland

The result:

I don’t love it. It’s too wordy and messed up a crucial fact. (I’m not a fellow real estate agent, I’m from a lead gen agency. Well, in reality, I’m neither. But in this example case I’m the latter.)

So I prompted ChatGPT to tighten it up and to get the fact right.

can you make that email shorter, and also remove the part about me being a fellow real estate agent?

That’s much better. Now that I got that squared away, I can have ChatGPT try writing the email a few other ways.

can you make it funnier? (I highlighted its jokes in yellow.)

I also tried out other prompts like can you make the email feel more urgent?… can you write the email with a sense of scarcity, letting them know we're not taking on many more clients?… and can you make the email just 4 sentences?

You can keep refining your email ad nauseam, trying different styles, tones, and lengths. And by the time you’re done, you should have a great first draft.

I wouldn’t recommend sending this email as-is. It’s too generic to be effective. (Or, at least, to hit maximum effectiveness.)

But as a jumping off point? ChatGPT can write great first drafts to save you lots of time.

And if you decide you do want to roll with ChatGPT’s result unedited? At least you probably won’t get caught.

Here was the result when I checked the first cold email with an AI and plagiarism detector:

2. Give you subject line ideas

ChatGPT can also give you subject lines to try out with your email. And it can bang out those subject lines at scale.

please write 10 possible subject lines for that email

Those aren’t all winners but there are three or four in there that caught my eye has having potential.

And as with the email copy, I can direct ChatGPT on tone.

This is, hands down, my favorite thing ChatGPT can do for cold emailers.

I got my hands on 40 possible subject lines in less than a minute. Not all are winners, but each list of 10 has a few potential candidates. (Plus if I’d tried to write 40 subject lines myself, it would’ve taken an hour and not all would be winners there either.)

And if I wasn’t happy with this 40, I could keep on generating to find better and better subjects. At least until I ran into the dreaded ChatGPT Too many requests, please slow down message.

And that’s not all: ChatGPT can also generate subject line + preview text pairs.

It took me a lot of work on prompts to get it right (ChatGPT really wanted to write long preview text), but these are impressive results:

3. Give you A/B testing possibilities

So how do you choose from the subject lines and email messages you’ve generated?

With this many possibilities in front of you, A/B testing is pretty much essential.

Here’s how to use GMass’s spintax to run quick split tests on multiple email messages.

(You can also use this same technique on subject lines. Yes, you could test subjects and email messages together. But I always recommend against testing multiple variables in single A/B tests.)

In the Gmail compose window, connect your Google Sheet to GMass and enter your subject line.

Then type {{spin}} and paste in the first variation of your email.

Type {{variation}} and paste in another variation of your email. You can repeat this over and over to test different copy options for this campaign. (Just make sure your list is big enough to send each variation to a nice-sized sample.)

Once you’re done adding variations, type {{end spin}}.

Now go into the GMass settings for the campaign. Go to the Advanced section and check the box next to A/B Test.

Your options are:

Percentage to send before making a decision. In the screenshot above, the first variation would go to 20% of your list and the second variation would go to a different 20%. That makes up the 40% total to send to before making a decision. After that, the remaining 60% of your list would get the “winning” variation.

Decide when? You can choose how long to run your test before declaring a winner.

Decide how? If you choose automatic, GMass will send your winning variation to the remainder of your list at the end of the testing period. And you can decide how GMass will determine the winner: Using opens, clicks, or replies. If you choose manual, you’ll receive an email with the A/B test results where you can decide which variation to send to the rest of your list.

A/B testing is this easy with GMass. So with the volume of possibilities you’ll get from ChatGPT, it makes sense to always test at least a few of them in every campaign.

4. Writing automated follow-ups

We hear it all the time: No one likes writing auto follow-ups. You put everything you had into that initial email.

The follow-ups then become a challenge to see how many different ways you can say, “Hey did you see my original email? Are you interested?”

So offload your follow-ups onto ChatGPT.

It’s a lot easier to use theirs (with some edits) than start from scratch on your own.

I prompted can you write 3 short follow-ups to the email that will go to people who do not respond?

The emails were too generic for my liking so I tried a few other prompts. And ultimately, I found a prompt that got ChatGPT to write a sequence using the strategy I wanted.

Please write three follow-up emails to send to people who don't respond. Ramp up the urgency with each email and make sure the final email lets them know it's our last time reaching out.

Much like the initial campaign email ChatGPT created, I wouldn’t use these emails without editing and tweaking them first. However there’s a solid foundation to work from here so these will absolutely save me some time.

To add those follow-ups to GMass, open the settings for your campaign and go to the Auto Follow-up section.

You can set the trigger and timing for each follow-up. For example, I’m sending the first one after two days if I haven’t received a reply; the second one after five days; and the third one after nine days.

5. Proofreading, gauging tone, and help for non-native speakers

So far I’ve covered ways ChatGPT can write emails and related copy. But it can also review your copy in ways other tools can’t.

There are tools out there to help with grammar and writing (your Grammarlys of the world and such).

ChatGPT can replace a lot of what they offer; here’s an email I fed it with three errors to see if it would catch them. It did — and even suggested a new clarifying sentence to add.

That takes ChatGPT beyond a copy editor; it’s also a developmental editor.

And ChatGPT has another, even more valuable function in the editing process: It can read your email and gauge its tone.

This could be quite valuable for non-native speakers — or native speakers who want a second opinion.

For example, I had ChatGPT evaluate the tone on one of the follow-up emails it produced. Here’s what it told me:

Then I tested it by writing a hostile email. And it correctly recognized I was being too aggressive.

Finally, I tried an email with an issue we hear about a lot from non-native English speakers: Being too formal. ChatGPT was able to recognize that this message was unnaturally formal.

I could see a lot of cold emailers using these editing functions going forward. Even if you aren’t going to use ChatGPT on the email copywriting side it’s a handy tool to have on the copy editing side.

Plus it never hurts to have a second set of eyes on an email. Even cold, unfixing robot eyes.

5 Crucial Cold Email Elements Where You Can’t Rely on AI

Alright. So we just witnessed ChatGPT write an email, dozens of subject lines and preview text options, and follow-ups. We even had it proofread and check our tone.

So are we all out of a job?

Not yet. We’re closer to being wiped out than ever before… but we’re not obsolete yet.

Right now, ChatGPT is like a warp zone in a video game. You can use it to skip a lot of the time-consuming grunt work — but it won’t defeat Bowser for you.

And in the case of cold email, the warp zone takes you past steps like writing multiple drafts, subject lines, and follow-ups. But the big boss at the end is creating a cold email that actually converts. And ChatGPT can’t do that for you.

Here are five crucial facets of cold email where you’ll need to take the reins from AI.

1. Writing complete, ready-to-send drafts

As we saw earlier, ChatGPT can produce email drafts and, depending on how you prompt it, can revise it and tweak those drafts in endless ways.

But even if you spend hours refining your prompts and coaching ChatGPT to write better and better messages, you’ll still want to get in there to tweak the copy at the end.

An effective cold email needs to do four things:

• Understand the recipient’s specific, personalized, pressing problem or need.
• Explain how you’re offering the perfect solution for that problem.
• Give relevant social proof as to why you’re the right person to provide that solution.
• End with a convincing call-to-action that prompts the recipient to click or reply.

ChatGPT can address all four of those in generic ways. But no matter how well you prompt it, it will never achieve a full understanding of who you’re emailing, what you offer, and who you are.

I tried getting as specific as possible when prompting ChatGPT to add social proof to an email. But the results still have a generic vibe. I’d be better off writing that section manually than trying to prompt my way to success.

ChatGPT also trends toward being way too wordy. You can prompt it to keep things shorter — but you’ll still find extraneous words throughout the copy.

You want every word of a cold email to matter.

So sentences like “Our team of experts uses a variety of proven tactics and strategies to target potential clients in your area and get them interested in your services” are antithetical to your mission.

That’s a whole lot of words just to cough out something painfully generic. What lead gen agency doesn’t do everything described in that sentence? That’s not a differentiator.

ChatGPT can lay the foundation for your cold email — but you need to take over from there.

2. Personalizing your cold emails the right way

Cold emails need personalization to work. That was true before ChatGPT and it’s true after.

And cold emailers have been trying to use AI for personalization for a few years now. There’s an entire niche of cold email AI line writers which… um… casually stalk your contacts’ website and LinkedIn, then write personalized opening lines.

Of course, those are all misconstruing “reciting facts about a person” with “personalization.” The most effective form of personalization is demonstrating you understand a person’s business — and their needs and problems.

ChatGPT can’t really do either of the types of personalization I’ve described.

It won’t do the whole stalking bit; ChatGPT doesn’t do internet research based on your prompts.

So there are no opening lines here about “I heard you on XYZ podcast and you provided some great insights” or “I see you attended the University of Kansas in 2005, must have been a great time while getting a great education.”

And it may try to address a recipient’s problem… but it’s just guessing. And, again, it trends generic in its guesses.

You don’t have to be generic. You can take the time to research an industry, business, and market to accurately ascertain a recipient’s problem and how to best address it.

Using the running example through this article, you could research the Bethesda, Maryland, real estate scene. You could find who the top agents are.

Then, you could reach out to the agents who aren’t on that list. Offer your services which will help them overtake the specific top agents in their market.

ChatGPT is wildly impressive, but personalization still needs to come from a person.

3. Following cold email copywriting best practices

There are lots of proven cold email formulas you can use for your copywriting.

You might like the Problem-Agitate-Solution template. Or the Attention-Interest-Desire-Action template. Or the Star-Chain-Hook method. Or one of the many other methods for writing concise, effective sales copy.

ChatGPT doesn’t write with that level of nuance. As I’ve said earlier, it’s wordy. It often lays out the information in the email without instinctively building toward the close.

I’ve tried to get it to follow copywriting best practices and it does its best… but still can’t quite nail them. For example, this is getting there but not what I’m looking for…

So… if you want ChatGPT to follow your preferred formula, you’ll need to work hard on different prompts. And even after you get ChatGPT’s best result — it will still take editing and finessing on your part.

4. Campaign strategy

ChatGPT can handle a lot of your writing but it’s not ready to take over on strategy.

You’ll still need to figure out the strategic aspects for your campaign like:

• Who to target
• How to build lists of those people to target
• What cold email formula to use
• If you should send auto follow-ups, and, if so, how to trigger and time them
• The right time to send your email
• Whether to throttle your sending speed
• Whether you should use images
• Whether to send your messages as replies to your last conversation with each recipient
• How to handle replies
• After prospects schedule calls, what to say to them to close a deal

ChatGPT will try to give you advice on these strategic decisions if you prompt it to do so… but the advice is generic.

Which is what you’d expect. It’s regurgitating middle-of-the-road answers from the hundreds of blog posts from which it learned.

Here are its thoughts on follow-up sequences.

And here are its thoughts on scheduling.

ChatGPT won’t make any definitive, well-reasoned strategic suggestions on your behalf.

You’ll still need to make those calls yourself to have a successful campaign.

5. Handling everything that comes after you send your campaign

So you used ChatGPT to help you write a great email campaign. You sent that campaign out to your list.

And now… you’re on your own. The robot has left the building.

You’ll need to handle everything that comes after you send a campaign.

Most important: Handling replies. Yes, in theory you could turn to ChatGPT to write responses to your prospects’ replies. But in most cases, when someone replies, they’re asking for specifics. And ChatGPT doesn’t do specifics.

Other things you’ll need to handle on your own are:

• Checking and analyzing your reports
• Making any edits to the campaign or the auto follow-ups
• Deciding to end or pause a campaign
• Coming up with ideas for follow-up campaigns based on recipient behavior (e.g., sending a new campaign to people who opened but didn’t reply)

AI has gotten good… but it’s not prepared to take over these crucial parts of campaign management.

ChatGPT and Cold Email: Takeaways and Next Steps

These two things are simultaneously true:

• You’d be crazy not to use ChatGPT to help with your cold email campaigns.
• ChatGPT is not at all ready to solely take over your cold email campaigns.

Prompts are everything with ChatGPT, so the better you get at writing prompts, the more it can do. And it can do a lot…

As we’ve discussed, ChatGPT can be a huge time saver on your first drafts:

• Writing email message copy and trying out different tones and styles.
• Writing high-volume subject lines (and preview text).
• Handling auto follow-up sequences since no one loves writing those.

And it can serve as your copy editor before you send:

• Checking your email for tonal issues.
• Checking your email for grammar, spelling, and other writing issues.

But you’ll still need to serve as a serious buffer between ChatGPT and your recipients:

• You need to revise the copy to make it less generic and wordy.
• You need to handle personalization.
• You need to really work with ChatGPT on cold email copywriting formulas and flows.
• You need to determine campaign strategy.
• And you need to handle everything post-campaign, especially corresponding with recipients who reply.

If you’re looking for more advice on ways to use ChatGPT with cold email, check out GMass’s community on Facebook.

And, of course, if you’re looking for a cold email platform to use after you and ChatGPT have whipped up some brilliant copy, make sure to check out GMass. It works directly inside Gmail and 99% of users report the best deliverability they’ve ever had.

You can get started with GMass by downloading the Chrome extension. There’s no credit card required to try it out — and you can send 50 emails a day during your free trial.

Use our quickstart guide to get rolling — you’ll be sending your first campaign (with or without ChatGPT’s help) in minutes.