The short version: cold emails get replies in 2026 when they are short, specific to one person, and ask for something small, not when they are clever, flattering, or stuffed with AI-generated personalisation that everyone can now spot from a mile off. Reply rates come down to list quality and offer, not subject line tricks. If you send 40 tightly targeted emails with a real, checkable reason for writing, you will beat 4,000 generic ones every single time.
Why most cold emails still fail, even now
Inboxes are worse than they were three years ago. Everyone with a laptop and a ChatGPT subscription is now sending "personalised" cold emails at scale, and most of them read exactly the same. "I noticed you're the Head of Growth at [Company], and I was really impressed by your recent post about..." Stop. Every senior person I know deletes that sentence structure on sight now. It became a giveaway the moment it became easy to produce.
Here's the bit nobody selling cold email software wants to say out loud: personalisation that mentions your prospect's LinkedIn post or company blog is not real personalisation any more. It's a template with a variable slotted in. People can tell, and it now works against you rather than for you, because it signals "I ran you through a tool" rather than "I looked at your business."
Real personalisation in 2026 means referencing something specific and slightly awkward to fake: a pricing page that's confusing, a hiring post that reveals a gap, a competitor move that would worry them. It takes longer to write. That's exactly why it still works.
What I did when my own cold emails were dying
When I was rebuilding my consultancy after a rough stretch, I went back to basics and did the thing I tell clients to do: I tested my own cold email like a proper campaign instead of just firing it off and hoping. First batch, 40 emails to marketing directors at mid-sized UK firms, a fairly standard "here's what I do, here's a case study, want a call" message. Three replies. Two were "not right now," one was a hard no. That's a 7.5% reply rate, and honestly most of that was people being polite.
Second batch, 35 emails, same list quality, completely different approach. I cut the email to four sentences, removed every mention of my own credentials, and asked one specific yes/no question tied to something I'd found by looking at their website (in one case, a broken automated sequence that was clearly sending the wrong follow-up email to leads, which I could see because I'd signed up for their newsletter myself to check). Eleven replies. That's 31%. Same me, same offer underneath, completely different framing.
The difference wasn't cleverness. It was specificity and length. I gave them one thing to react to instead of a pitch to process.
The structure that gets a reply
Cold emails that work in 2026 tend to follow the same shape, whether they're selling consulting, software, or freelance services. Here's the framework I use and teach:
- Subject line, 3 to 5 words, no punctuation tricks. "Quick question about your onboarding" beats "🚀 Boost Your Conversions Today!" every time. Anything that reads like marketing gets filed as marketing.
- Opening line proves you looked, in one sentence. Not "I loved your recent article," which anyone can fake, but something they'd have to check to be sure you're right, like a specific number, page, or gap.
- One sentence on why you're writing. Not your company history. What you noticed and why it might matter to them.
- One sentence of proof, if you have it. A real result, with a number, from a comparable business. "We took a similar SaaS company from 2% to 11% trial-to-paid in eight weeks" does more work than three paragraphs of credentials.
- One small ask. Not "let's hop on a 30-minute call." Try "worth a two-line reply?" or "should I send the two-minute breakdown?" Small asks get answered. Big asks get ignored.
That's it. Five to seven sentences, total. If your cold email needs scrolling to finish, it's a proposal, not a cold email, and it will be treated as one, meaning ignored until someone has twenty free minutes, which is never.
Step by step: building a cold email that gets a reply
- Pick 30 to 50 people, not 500. A tight, well-researched list beats a huge scraped one. If you can't say why each person specifically is a fit, cut them.
- Find one real, checkable detail per person. Their pricing page, a recent hire, a job listing revealing a problem, a review complaint pattern, a gap in their content. Give yourself 90 seconds per prospect, no more, or you'll never send anything.
- Write the email in five sentences. Subject, opening observation, reason for writing, proof point, small ask.
- Read it out loud. If it sounds like something a human would type to another human, keep it. If it sounds like marketing copy, cut every adjective.
- Send Tuesday to Thursday, 7am to 9am local to them. Not because of some magic algorithm, but because that's when people clear inboxes before the day gets busy and before it fills up with the day's noise.
- Follow up once, four days later, in two sentences. "Wanted to bump this up in case it got buried, still worth a quick reply?" Second emails often outperform the first because they read as human rather than templated.
- Stop after two follow-ups. Three-plus touches on a cold prospect who hasn't replied reads as desperate, and it damages your name for the next time you email them.
Where AI helps, and where it wrecks your reply rate
AI is useful for the unglamorous parts: pulling company details fast, drafting a rough first version you then rewrite, checking tone, building a follow-up cadence. It is not useful for the sentence that has to sound like you noticed something. The moment you let AI write your "personalised" opening line wholesale, it reverts to the template pattern every recipient has already learned to spot. I've seen this pattern so often in inboxes I manage for clients that I can now tell within one sentence whether an email was AI-drafted end to end or human-edited.
If you're sending cold email as part of a wider outreach system, it's worth setting up an AI-assisted sales email sequence for the follow-ups specifically, since those benefit from consistency, and saving the human effort for the first, hardest-to-fake message. Automating the boring bits of the process, list building, CRM logging, reminder scheduling, is a good use of workflow automation, and it frees you up to spend your actual thinking time on the fifty words that need a human brain behind them.
If research is the bottleneck, that's also a task worth handing off. A part-time virtual assistant can build you a list of 50 tightly qualified prospects with one specific detail each, in a fraction of the time it would take you, and it's a common early hire for consultants and agencies scaling outreach.
The uncomfortable bit about targeting
Here's what most cold email guides skip: no email template fixes a bad list. I've watched people spend three hours perfecting subject line wording on a list of 200 people who were never going to buy from them in the first place, wrong industry, wrong budget, wrong stage of business. You cannot write your way out of that. A 15% reply rate on the wrong 200 people is worthless. A 15% reply rate on the right 30 people is a full pipeline.
Before you touch the copy, ask whether the person you're emailing has the actual problem you solve, the budget to fix it, and the authority to say yes. If any of those three is missing, no amount of clever phrasing rescues it. This is also why agencies and consultants who build proper lead generation systems tend to outperform lone cold emailers, because the targeting work happens before a single email gets typed. If you're running outreach for an agency at any scale, it's worth looking at how lead generation tools built for agencies handle list building and enrichment, since that's where most of the reply rate improvement lives, not in the email itself.
What "getting a reply" is worth
Not every reply is a good reply, and this matters more than people admit. A "no thanks" reply is still useful, it tells you the list or offer needs work. A "not right now, check back in Q3" reply is a placed bet, worth a calendar note. The reply you want is a question, because a question means they're thinking about it rather than dismissing it. If your emails are only generating flat "no" replies, the problem usually sits in your targeting or your proof point, not your grammar.
Track this for at least 100 sends before you decide a cold email approach doesn't work. Most people give up after 15 to 20 emails with no reply and conclude cold email is dead. It isn't dead. It's just being done badly at scale, which makes the well-done ones stand out more than they used to, not less.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
A quick example of a working email
Here's a stripped-down version close to the one that got me the 31% batch:
Subject: Question about your trial flow
"Hi [Name], noticed your free trial signup doesn't send a follow-up email until day 5, which is a long gap for new users. I helped [comparable company] cut that gap and lift trial-to-paid from 2% to 11% in eight weeks. Worth a two-line reply if you want the three changes we made?"
That's 48 words. No flattery, no company history, one specific detail, one proof point, one small ask. It works because it's easy to answer and hard to fake.
This same logic applies whether you're a consultant, a freelancer, or an agency owner. If you're new to outreach and worried about sounding too salesy, the same principle that gets a freelance writer their first client without sending 200 emails into the void applies here too: fewer, sharper emails to the right people beat volume every time.
Free resource: The Review Generation Email Swipe File.
Frequently asked questions
What's a good cold email reply rate in 2026?
Anywhere from 8% to 15% is solid for a well-targeted B2B list. Above 20% usually means your list is small and extremely well qualified, which is exactly what you want. Under 3% almost always points to a targeting problem, not a writing problem.
How long should a cold email be?
Five to seven sentences, roughly 50 to 90 words. If someone has to scroll on their phone to finish reading it, it's too long and will get skimmed, not answered.
Does AI-written cold email still work?
Fully AI-written cold emails, front to back, tend to underperform now because recipients have learned to spot the patterns. AI is useful for research, drafting a rough first pass, and follow-up sequences, but the opening line that proves you looked at their business specifically works better written by an actual human.
How many follow-ups should I send after a cold email?
Two, spaced four to five days apart, is enough. A third and fourth follow-up on someone who hasn't replied usually reads as pushy and can damage your reputation for future outreach to the same company.
Want this done for you? See how to get your business cited by ChatGPT and Perplexity.
Related reading: I Got AI to Chase My Late Invoices. Here's What Happened and How I Use AI to Chase Late-Paying Clients Without Sounding Like a Debt Collector.
