Cold Email Deliverability Is Killing Your Reply Rates

The conversation shifted this week

Most weeks on Twitter, cold email talk splits into two camps: people sharing wins and people sharing templates. This week felt different. There was genuine debate about whether cold email as a channel is deteriorating, a few really creative outreach approaches worth studying, and the usual mix of generic advice dressed up as insight.

I went through everything and picked out what's actually worth your time. Let's get into it.

The deliverability warning people are sleeping on

The numbers on emails-to-reply are directionally real. I've seen it in campaigns across dozens of industries. The 0.1% spam threshold stat is also accurate and it's the one most people aren't paying attention to. Two complaints per thousand sends and Google starts throttling you. That's nothing. One angry person having a bad day can quietly destroy a domain you've been warming for months.

Where I'd push back: "cold email in a silo is over" is both true and misleading at the same time. It's not that cold email stopped working. It's that the floor got raised. The people who get away with worse practices got filtered out, which means the people doing it right are seeing less inbox competition. The answer isn't to abandon the channel. It's to get the infrastructure right first.

If you haven't watched this yet, the warm-up piece is non-negotiable:

A new domain that you start sending from on day one is almost certain to land in spam or promotions. Tools like Instantly and Smartlead both have warm-up built in. Use them. This is not optional in the current environment.

And if your tech stack needs an audit, start here: /coldemailtechstack2025.

The most creative outreach approach I saw all week

The underlying principle here is what I call "do the work first." You remove all the friction from the prospect's imagination. Instead of asking them to picture what you could do for them, you show them. You've already done it. The deliverable exists. That shift changes everything about the dynamic.

I've coached people into this approach with audits, competitive analyses, and spec creative. One client I worked with was a video production company in New York. Instead of pitching with a deck, we had them build a short concept treatment tailored to each prospect's existing content before sending the first email. Open rates stayed consistent. Reply rates went up noticeably because the email had something real attached to it.

The math in this tweet is also worth examining. 5% conversion to a $200/month retainer is actually conservative if the product-fit is real. The tool cost is marginal. The issue is fulfillment quality at scale, which isn't addressed here. But the outreach strategy itself is sound. Do the work first, give it away, then ask for the relationship. That's a principle that ages well regardless of what tool generates the asset.

For list building at that scale, ScraperCity's Apollo scraper is worth looking at for pulling CMO contacts efficiently.

The numbers on job-seeker cold emails are rough

Over 900 saves on this, which tells me a lot of people are job hunting right now and grabbing anything that looks like a template. I get it. But this email has a core problem that templates can't fix: it's entirely about the sender.

Every line is "I have," "I am," "I can." There's nothing in there that shows any understanding of what the company actually needs right now. "I really think I could contribute to your team's success" is the softest possible way to say nothing.

The fix is simple. Do fifteen minutes of research. Find one specific thing the company is working on, hiring for, or struggling with publicly. Open with that. Then connect your skills to that specific thing. Five sentences total. That's it.

The LinkedIn message at the end is actually the strongest piece in the whole thread, only because it's short. Short wins in hiring outreach the same way it wins in sales outreach. If you want templates that work across job-seeking and B2B contexts, /top5scripts has frameworks that apply to both.

The "AI bots selling to AI bots" observation

Funny because it's close to true. The irony is real. But the conclusion people draw from this is wrong.

The answer isn't to stop using automation. The answer is to make your email something an AI would flag as worth surfacing to a human. That means a subject line that looks like it came from a real person, a first line that references something specific and recent, and a message short enough that a summary doesn't make it shorter. If your email reads like a bot wrote it, an AI filter will catch it. If it reads like a real person with a specific point, it gets through.

The channel isn't broken. The content running through it is. As one person in the replies this week put it more directly:

Accurate. AI can do a lot in outreach, but the prompting required to get genuinely good cold email copy is still more work than most people want to do. If you're using AI to generate emails and you're not happy with the output, the problem is almost always the prompt, not the model. Check out /coldemailgpt for prompts that actually produce usable emails.

The person who nailed the mindset in one tweet

This is the one. Nearly 4,000 likes and it deserves every one of them. The whole point of cold email isn't the technique. It's the decision to reach up instead of reaching sideways.

The 99% of the class who interviewed the retail manager didn't fail because they were bad at outreach. They failed before they sent a single message because they pre-rejected themselves. They decided what was realistic before they even tried. That decision - made silently, before any action - costs more than any bad subject line ever will.

I've sent millions of cold emails. The ones that opened the most doors weren't always the most polished. They were the ones I actually sent to the right person instead of the safe person.

"Don't self-reject" is the most useful phrase in this whole week's conversation. Almost everyone who tells me cold email doesn't work for them self-rejected before they ever built a real list, tested a real subject line, or sent enough volume to know anything. Send the email. The worst that happens is silence, and silence is free.

The "emails-to-reply" stat deserves more scrutiny

Already covered this above, but I want to add one more thing here because the comment section on this tweet had a lot of people using it as permission to give up.

The benchmark I use is different. In The Cold Email Manifesto, I lay out what a working system should produce: 200 leads generated per day, 200 emails sent per day, and a meeting book rate that keeps your calendar full. That system still works. What changed is the cost of getting there if your infrastructure is sloppy.

The people seeing 430 emails per reply are probably dealing with one or more of these: bad domain reputation, no warm-up, no email validation, sending to a cold list with no segmentation, or copy that triggers spam filters. Fix any one of those and the number drops. Fix all of them and you're back to something respectable.

Use an email validator before every send. If you're sending to bounced addresses, every bounce hurts your sender reputation and inflates your emails-per-reply number in the worst possible way.

The "be human" take that aged well

"In the age of AI, dare to be human" is a genuinely good line and it's correct. But I'd add precision to it: being human doesn't mean being casual or funny for its own sake. It means demonstrating awareness of the specific person you're emailing. It means the email could only have been written to them, not to a segment.

The Stanford gardening email this person mentions worked because it was absurdly specific and genuine. That's harder to fake than humor. If you're going to lean into the "be human" angle, ask yourself before you send: could this exact email have gone to 500 other people with just a name swap? If yes, it's not human. It's personalized automation wearing a human costume.

The "like and comment for the playbook" game

244 comments, 364 likes. The engagement bait works, which is why people keep doing it. I'm not going to pretend I haven't seen this format generate real reach.

But what's worth noting: the actual substance claim here is big. $4M in 60 days from cold email into investment banks. That's possible. Enterprise financial services is one of the harder verticals to crack because compliance teams are aggressive about filtering, and decision makers at that level get a lot of outreach. If the playbook is real, the interesting part isn't the number. It's the sequence structure and the offer framing that allowed it to work in a compliance-sensitive environment.

If you want scripts that have actually closed deals at that level without the engagement-bait mechanism, the /enterprise-outreach system is the place to start.

College students are asking the wrong questions in cold emails

This is a targeting problem dressed up as a content problem. Students are sending unanswerable questions because they haven't thought clearly about what they actually want from the interaction. They want access. But they're asking for opinions on abstract futures, which gives the recipient nothing to engage with and the student nothing actionable in return.

The fix: ask a question only that specific person can answer. Something they've done, built, or decided. "You moved from investment banking to running a portfolio company in [year] - what was the first six months like operationally?" That's answerable. It shows you did research. It signals you're serious. Abstract trend questions signal you did a Google search.

The advice to "learn to sell" at the end of the tweet is worth more than the whole thread. Every version of the economy rewards people who can generate revenue. That doesn't change regardless of what the models can do.

The simple template that outperforms everything fancy

This is the conditional offer structure and it works because it's not a pitch. It's a question. You're not telling them what you do. You're asking if they want a specific outcome. Psychologically, that's a much easier message to respond to than a paragraph of credentials followed by a calendar link.

The P.S. is where most people underuse social proof. Put the most specific proof you have there. Not "we've worked with hundreds of companies." Something like "we did this for a $40M e-commerce brand last quarter and their CPL dropped 34%." That specificity in the P.S. is often what converts a "maybe" into a reply.

If you want to see more frameworks in this style, /killercoldemails has a full set.

The fake iPhone footer joke that's funnier than it should be

I've seen this. More than once. The sad part is it actually used to work as a deliverability and credibility hack. Now it's a signal that someone read a cold email blog from four years ago and stopped there. Recipients notice. The goal isn't to trick someone into thinking you typed this on a phone while standing in line for coffee. The goal is to write something that would be worth reading regardless of where it was sent from.

What actually matters this week

Three things I'd take from this week's conversation and act on immediately:

One: Your domain health is more important than your copy. If you're seeing declining reply rates and you haven't looked at your spam complaint rate, your sending volume curve, or your warm-up status, start there. A brilliant email in a burned domain does nothing. Check the full stack at /coldemailtechstack2025.

Two: Do the work before you ask for the meeting. The free video ad approach in this week's tweets is a specific application of a principle that works everywhere. Audit something. Build something. Show your work before you ask for 15 minutes. The emails that get replies have already delivered value before the ask arrives.

Three: Stop self-rejecting before you send. The Versace story and the researcher who got into the McGill lab are the same story. The email worked because it got sent. Your dream clients aren't in a different universe. They have inboxes. Use them.

If you want follow-up sequences that keep deals alive after that first reply, /followup is where to go next.