Reply rates are a fight, and the public numbers don't agree with each other. Some of the largest datasets say the channel got worse through 2025 and 2026. One sequencer says it didn't. Both can be reported without picking a side the data doesn't earn.
The useful question is the one in the title. If fewer people reply, is the copy worse, or have buyers gone numb? The honest draft answer is that both are happening, and they feed each other. A harder channel punishes a message that sounds like every other message. And a message that sounds like every other message is part of why the channel got harder.
What the Datasets Actually Say
These publishers sell into outbound, or run it. That doesn't make the counts fake. It means they are not a neutral bureau, and they don't measure the same way. Read the denominator before the headline.
- Belkins, a lead-gen agency, reported on 7.53 million cold emails in 2025. Average reply rate 0.45% of sends. 0.50% in the first half of the year, 0.40% in the second, with a low of 0.35% in December. (Belkins)
- Sales.co, an outbound agency, tracks human replies per contact first emailed. Their series runs from 2.36% in April 2025 to 0.45% in September 2026, a partial month, and the positive reply rate in that series falls from 0.72% to 0.18%. (Sales.co, monthly CSV)
- Carroway Research, one outbound operator, 42 programs, 736,078 emails. Human reply rate fell every month from 0.55% of sends in May 2026 to 0.40% in August, while monthly sends rose from 102,525 to 255,675. More mail, fewer replies per mail. (Carroway)
- Mailshake, a sequencer, surveyed senders in 2025. 69% said performance declined year over year. They partly blame decision-makers avoiding AI-written content. Only 5% of senders personalize every email. (Mailshake report PDF)
- Instantly, also a sequencer, is the counter. Their 2026 benchmark puts the average at 3.43% and says reply rates "remain stable despite growing volume," on a different method. They also credit problem-focused messaging, and testing new messaging weekly, for the top tier. (Instantly)
The defensible sentence is: several large datasets show declines through 2025–26, and one major sequencer says the rate is stable. It is not "replies fell to X." The spread is an order of magnitude because the denominators differ. Belkins and Carroway divide by sends. Sales.co divides by the contact's first email. Instantly's method is their own.
Bad Copy, Numb Market
The numb-market half. Buyers see more outbound than they used to, a lot of it personalized from the same public facts. Gartner reported that 73% of B2B buyers actively avoid suppliers who send irrelevant outreach. That figure is from their June 2025 sales survey, the same release as the finding that 61% prefer a rep-free buying experience. A signal in the first line ("congrats on the round") used to mean a person worked. Once every email does it, the line means a tool worked. Sales.co, on 1.28 million contacts, found AI-written first lines at 0.53% replies against 1.00% for a plain template. One agency, one comparison. A smaller test has pointed the other way. Treat it as a warning that the personalized line is not automatically the better one.
The bad-copy half. Volume went up while replies went down in the Carroway set. That is what a numb channel looks like if the answer is "send more." It is also what a channel looks like when the extra sends are the same email. Instantly's own note, on the side that disputes the decline, still says the top performers test new messaging every week and write to the problem. That is a copy claim from a company that would be happy to tell you the channel is fine.
Mailbox rules cap the volume answer on their own. Since February 2024, Gmail and Yahoo have expected bulk senders to stay under a 0.3% spam-complaint rate and to offer one-click unsubscribe. You cannot out-send a complaint problem.
What This Doesn’t Mean
It doesn't mean stop writing cold email. A cold sequence is one stage, and it can be written with the same care as any other letter: one reader, one offer, one ask. It does mean the email is doing less of the persuading than the team hopes, so the page it opens has to do more. If the email says "3 days, without hiring" and the page says "all-in-one platform," the channel gets the blame for a break that happened after the click.
It also doesn't mean a cleverer first line will restore 2022. List and offer still dominate. The old direct-mail rule of thumb, 40% list, 40% offer, 20% creative, is lore, not a measurement. After enrichment, the list is already fixed. What's left to work on is the offer and the words, including the words on the page.
I sell that work. The Message Match Audit is $400 at the founding rate: one live sequence, the page it links to, and the places the promise slips. Paid, scoped, not a free read.