Learn / Email copy
How to write a cold email that actually gets replies
We've booked over 3,000 sales calls from cold email in the past year, and it has never been harder to land in someone's inbox and stay there. AI and sending tools made mass email close to free, so almost everyone is sending, inboxes are slammed, and you get ignored. That's only true if you write the same email everybody else is writing. So here is how to write a cold email that survives that inbox, in four steps: the offer, the subject line, the body, and the follow-up.
Every rewrite below is torn down from a real cold email that landed in my inbox in the last few weeks. What they sent, why it died, and the version I'd send instead.
Step 1: the offer, because nothing else matters if this is wrong
You can write the best email in the world and it still goes in the trash if the offer doesn't make an impression. This is where most people are held back without realizing it. Replies come in, the campaign feels like it's working, and then nobody books. Or they book and the conversation goes nowhere.
That's farming interest. Anyone can make an offer broad enough to get a yes, and that yes means nothing if the person was never going to buy. A real offer does the opposite: specific enough to filter out the people who don't want it, compelling enough to pull in the people who do.
The frame I use is Hormozi's value equation: dream outcome times perceived likelihood of achievement, sitting over time delay times effort and sacrifice. In plain words, four questions. Do they care about the thing? Will they believe it can happen? Does it happen soon? And how little do they have to do to get it? Get all four right and the prospect lands on one thought, which is the only reaction you're actually engineering: why not.
The offer does the work. The copy's only job is staying out of its way.
Three real offers, torn down and rewritten
Left is what hit my inbox, middle is the diagnosis, right is the rewrite. Note how little changes structurally: every fix is the same fix. Name the outcome, shrink their effort, move the risk onto your side.
| What they sent | Why it dies | The rewrite |
|---|---|---|
| LLM and Reddit SEO. “We help brands become more visible in early stage conversations.” The actual offer was buried behind a Fiverr link. | No mechanism, no end result, no proof, no risk removed. Visibility is not an outcome anyone can picture. | “If we could get your brand recommended by ChatGPT and Perplexity when buyers ask for the best in your category, just like we did for {similar company}, all you do is hand over your differentiators and a sign-off, and you only pay once you're showing up above them.” |
| Lead gen. “A quick overview of how agencies build a steady pipeline, 5 to 9 qualified calls monthly, without SDR hires or managing outbound themselves.” | Genuinely decent: low friction, an asset instead of a call, and it names two things founders hate doing. But there's no risk reversal, and 5 to 9 calls at a 10% close rate is not a closed deal. | “If we could book you 10 to 20 qualified calls with {dream client}, just like we did for {similar company}, all you have to do is approve the targeting and show up, and you only pay per qualified call.” |
| YouTube growth. “Saw your channel only has 340 subscribers and you've published 20 videos. I can help you monetize, increase your watch time, and increase your targeted audience.” | Opens by insulting the prospect, then specifies nothing. No number, no mechanism, no timeframe, no reason to believe any of it. | “We'll make your next video 100% free: ideated, scripted, edited, and packaged with a thumbnail. All you do is record.” Or the results version: “if it doesn't book you 5 calls in the next 30 days, full refund.” |
One caveat on the lead gen rewrite: pay per qualified call has been run into the ground, so you'd need a layer above it to sound like anyone other than the last six vendors who emailed them. More on building the thing you hand over first in the front-end offer breakdown.
Step 2: the subject line only has one job
There's one rule here: your subject line cannot look like it belongs to a cold email. Inboxes see the same pattern so often that people filter it out before they consciously decide to ignore you. Any trace of marketing language and it's dead on arrival.
Then people overcorrect. They strip out all the context, chase mystery, and end up with something like "Can it be achieved?", which gives the reader nothing to be curious about. Land in the middle: enough that it feels relevant to them, not so much that they know they're being pitched. The moment they can guess the offer from the subject line, the curiosity is gone and so is the open.
A web design offer came in with "FW:" on the front, faking a forward to feel internal. Starting a relationship with deceit is a choice. All lowercase gets the same internal feel honestly: "your new landing page", or sharper, "issue with your page", which implies a problem you happen to solve. The LLM visibility email used "one quick question". It works, but the reader clocks it as a cold email instantly. "your recent GPT results" does more with the same three words.
That's the shape: three to four words, lowercase, and it reads like it came from inside their company. The full breakdown, including the patterns that still open in 2026, lives on the subject lines page.
Step 3: the body is the offer, chunked
Here's the part people overbuild. Once the offer is right, it basically IS the copy. Translating one into the other is a readability job, not a writing job. The rewritten LLM offer, dropped into an email:
{first name}, if we could get your brand recommended first by ChatGPT when buyers look for {solution}, all you'd have to do is share your differentiators.
Would that be worth sharing more?
PS - zero upfront payment until you're actually featured first.
Dream outcome first, effort reduced in the same breath, low friction ask, risk reversal in the PS. Under 50 words. The line breaks matter more than they look: a cold prospect scans before they read, so chunking lets them catch the offer in one pass and decide whether to keep going.
For years the body was solved with personalization: reference something specific, make it feel one to one, collect the reply. That edge is gone. AI personalizes everything at scale now, everyone's doing it, and prospects know it, so the line that used to read as a human touch reads as a tell. What's working instead is risk reversal, and not the money-back guarantee, which is washed out. Casual risk elimination, the way a reasonable person actually talks: "we'll do this for free and only move forward if we see something worth implementing", or "if it's not a fit, no hard feelings, I just wanted to get in front of you." That doesn't feel like a pitch. It feels like someone starting a conversation.
Two more things decide this section. First, the CTA. "Would that be worth sharing more?" outperforms "worth a chat?" because it's rare that a stranger hands over 30 minutes on the first message, and when they do, they're often not the buyer you want. The soft ask lets them steer: raise the objection, ask the price, ask for examples, or just say sure, when are you available.
Second, layout. The blocks that get read are the opening and the PS. If you lose anything, you lose the middle, so put the most important part of the offer at the top and the second most important in the PS, then test swapping them. The PS is where social proof goes, phrased around a company like theirs plus the pain they want to skip:
PS - just partnered with {similar company, nearby market} and helped them land 12 new accounting clients last month, without {the part they hate doing}.
Test that with and without. Social proof qualifies you in some industries and reads as invented in others, where the plainer email wins. And if your service is genuinely new or hard to make tangible, there's a whole other route that carries no offer in message one at all, which is the three no-pitch scripts.
This is the boring version of cold email and it keeps working. An M&A firm booked 31 calls in their first 30 days on it, off 91 positive replies. An eCommerce exit consultant, coming off an outbound agency that produced nothing, landed 42 sales calls in 30 days. A growth consulting firm booked 24 calls and closed 3 deals in their first month. No workflow, no research agent, just a short email with a real offer in it.
Step 4: follow up twice, three times if you must
You can't send one email and wait. Inboxes are full, people are in meetings, and a good share of your list never sees message one. Which means email two might be the first one they read, and that changes how you write it: the follow-up has to build on the first email and stand on its own.
Then cap it. Two, maybe three, and three only if your market is small enough that every contact has to count. One sender hit my inbox five times, with the second landing one business day after the first, Friday to Monday. Every message re-sold the same thing in new packaging: an overview, then a walkthrough, then a video, then "are you the right person", then "looks like this isn't a focus." Past three, I don't care how good your offer is, nobody is replying. You've just made them annoyed.
There are only two follow-ups worth writing:
- Switch the mechanism. Same benefit, easier to consume. You went for the call in message one, so now hand them something they can look at alone: a short video, a lead magnet, a one-pager.
- Resurface what got buried. Take the single strongest part of the offer, usually the risk reversal sitting in the PS, and make it the whole message.
{first name}, if it's easier, I can just send a 3-minute video showing exactly how we rank {industry} first on ChatGPT when buyers look for {solution}. Mind if I send it?
Know the trade you're making: a lower friction ask pulls replies with less intent than the original offer did. The alternative is silence. A yes or a no is data you can point at the next campaign, and video follow-ups only work if you can actually explain your craft on camera. If you can't, send the written asset instead. The full sequencing and what to do with the warm replies is in the follow-up system.
One note on AI. I ran the bad offers above through Claudeand had the rewrites in minutes, because the value equation is exactly the kind of judgment a model applies well once you hand it the real offer. What it can't decide is what you're willing to guarantee. That part is yours.
tl;dr: fix the offer until a stranger's only reaction is "why not", write a subject line that doesn't smell like outreach, chunk the offer into under 50 words with the risk reversal in the PS, then follow up twice with the same benefit delivered a different way.
FAQ
How do you write a cold email that gets a reply?
Four steps, in this order. Build an offer specific enough to filter out the people who were never buying. Write a subject line that does not read like outreach. Put the offer in the body, chunked so the first and last lines carry it. Then follow up twice with the same benefit delivered a different way. Skip step one and the other three cannot save you.
How long should a cold email be?
Under 50 words is the target. You are competing for attention against an inbox that is already full, so the faster you land the offer, the better you do. The two blocks that get read are the opening line and the PS. Anything you bury in the middle is the part you should assume nobody sees.
What is the best call to action for a cold email?
'Would that be worth sharing more?' beats 'worth a chat?' in most campaigns. Almost nobody hands a stranger 30 minutes on the first message, and the ones who do are rarely the right buyer. The softer ask lets the prospect pick their own next step: raise an objection, ask for pricing, request an example, or book. All four are conversations.
Does personalization still work in cold email?
The edge is gone. AI personalizes at scale for free now, everyone is doing it, and prospects can tell, so the line that used to feel human is now a reason to ignore you. Risk reversal is doing the job personalization used to do. One says you did research. The other says you are confident enough to take the pressure off them.
How many follow-up emails should you send?
Two, three at the absolute most, and three only when your market is small enough that you need every contact to count. Past that you are not persistent, you are annoying, and no offer is good enough to survive it. One sender hit my inbox five times, the second landing one business day after the first, and every message sold the same thing.
Should a cold email include social proof?
Test it. Social proof belongs in the PS line, phrased around a company like theirs and paired with the thing they hate doing: helped a similar firm get the result without the part they dread. In some industries it qualifies you. In others it reads as something you could have made up, and the plainer email wins.
If you'd rather have the offer built, the copy written, and the sending handled for you, that's exactly what we do: over $45M in pipeline and 3,000+ sales calls booked for clients in the past year. If you'd rather build it yourself, every training we've published is grouped by topic on the trainings hub, and the full version of this one, with the teardowns on screen, is on YouTube.
PS - one client closed a $250k deal off a 50-word email. The email was plain. The offer wasn't.