Cold email is dying and there is nothing you can do about it

Human replies per contact fell from 2.36% to 0.45% in 17 months

Sales.co runs cold email campaigns for B2B clients through a sending tool we built. We counted every contact first-emailed from April 8, 2025 to September 25, 2026, 1,279,153 in all, and checked whether each one ever replied. Auto-replies, out-of-office notices, bounces and spam do not count.

In April 2025, 2.36% of contacts sent a human reply and 0.72% sent a positive one. By September 2026 those were 0.45% and 0.18%. September is not finished and its contacts have had less time to answer, but August, which has had more time, was 0.56% and 0.18%.

We grew a lot over the same period, from 5,291 contacts first-emailed in April 2025 to 144,317 in July 2026. More volume usually means new clients and less careful lists, and that could explain part of the fall. So we looked inside single clients. Nine clients sent both up to October 2025 and from June 2026 onward. For 8 of them the human reply rate fell, from 2.03% to 0.82%, from 3.54% to 1.03%, from 2.98% to 0.92%, and from 1.27% to 0.49%, for example. One client went up, from 0.97% to 1.26%.

Sales.co also keeps a yearly series that goes back much further, Cold Email Reply Rates by Year, 2010 to 2026. It puts the average cold outbound reply rate at 22% in 2010 and 2% in 2026. That series is counted differently from the per-contact numbers here, so compare the direction of the two and leave the levels alone.

Banner ads went from a 44% click rate in 1994 to 0.2% by May 2000

We have seen this shape before. The first banner ad ran in 1994, bought by AT&T on HotWired, and 44% of the people who saw it clicked on it. By 1995 the industry click rate was around 2%. By October 1998 it was 0.5%, and by May 2000 it was 0.2%.

In 1998 two researchers at Rice University, Benway and Lane, gave a name to the habit of web users skipping over anything that looks like an ad: banner blindness.

Three tools launched between 2018 and 2022 let every sender do the same thing at any volume

Cold email went through something similar. lemlist was founded in Paris in January 2018, and its best-known feature was the personalized image: a mass email could carry a picture with the recipient's name in it, so it looked made for one person. Instantly was founded in 2021 and offered unlimited email accounts on every plan. Smartlead, founded in 2021 and launched in 2022, offered unlimited mailboxes too. With enough mailboxes, a sender can spread a large campaign across many accounts so that no single account sends much.

Once most senders had these tools, buyers got many more emails that looked hand-written, and we think that is a large part of why fewer of them answer. Our own data starts in April 2025, so we cannot measure the effect of these tools directly. The public benchmarks further down cover the earlier years.

Personalization, follow-ups and randomized copy all fell with the rest

We sell cold email, so we tried to find a way around the decline. Three things we tested or measured:

Personalization. We built an AI first line for every contact. Contacts who got it replied at 0.53%, and contacts who got a plain template replied at 1.00%. The template matched or beat it in every month where both ran on 500 or more contacts. The details are in Personalization no longer matters for cold email.

Follow-ups. In an earlier study, The Cold Email Reality Check, 69% of replies came from the first message. Most of what a sequence gets back is decided by the first email.

Randomized components. Our tool can mix the value proposition, tone, length and call to action for each email from a set of options. Contacts who got these emails replied at 1.81% in August 2025 and 0.83% in July 2026, falling with everything else.

Public benchmarks, each measured its own way, show the same direction. Backlinko's email outreach study found an average response rate of 8.5% (Backlinko, 2019). Belkins reported 6.8% in 2023 and 5.8% in 2024, a 15% fall in one year (Belkins, 2024). Instantly's 2026 benchmark report puts the average reply rate across its customers' sends at 3.43% (Instantly, 2026). Their levels are higher than ours and each counts replies its own way, so compare direction, not level. Read in date order, they fall.

Positive replies are now about two per 1,000 contacts

If you buy cold email, from us or from anyone else, ask for positive replies per 1,000 contacts. Raw reply rates include auto-replies, which made up 53.7% of all replies in Cold Email Response Rates in 2026: The Real Numbers From 1,288,605 Sent Emails, and most human replies are a no.

In September 2026 our positive reply rate was 0.18%, about two positive replies for every 1,000 contacts. In April 2025 it was 0.72%, about seven. Nothing we changed in the copy moved that number, so what is left is how many of the right people you reach and what you offer them. A campaign at two per 1,000 needs a large list of people who could actually buy, and an offer some of them want now.

Methodology

Source. The Sales.co sending tool database: every email sent and every reply received between April 8, 2025 and September 25, 2026.

Unit. A contact is one recipient address within one client workspace, dated by the month of its first email (not a follow-up). 1,279,153 contacts were first-emailed in the window.

Replies. A contact counts as replied if the same address replied to the same client workspace at any time. Auto-replies, out-of-office notices, error notifications and spam are excluded from human replies. Positive means the reply was labeled possibly interested, interested, meeting booked or ongoing conversation.

Personalization classes. AI-personalized: the first email opened with a line written by a language model for that contact. Plain template: fixed copy with no line about the contact. Randomized components: value proposition, tone, length and call to action drawn from a component set, with no line about the contact.

Per-client check. Clients active both up to October 2025 and from June 2026, compared on their own human reply rate in each period. Clients are anonymized in the published files.

Limitations

• The client mix changes from month to month. The per-client check covers only 9 clients.
• Recent months have had less time to collect replies. September 2026 is partial, up to September 25.
• A contact is classed by its first email. Later follow-ups may have been written another way.
• Replies are matched by address within a client, so a reply to a follow-up is credited to the contact, not to a specific email.
• The AI-personalized class is concentrated in May to September 2026.
• This is our own sending tool and our own clients only.

Data

• cold-email-reply-rate-by-month-2025-2026.csv: contacts, human and positive replies by month of first email
• cold-email-reply-rate-by-client-month-2025-2026.csv: the same by anonymized client
• cold-email-personalisation-by-month-2026.csv: by personalization class and month

Sources

1. 1. Backlinko. Email outreach study. 2019.
2. 2. Belkins. Cold outreach benchmarks. 2023 and 2024.
3. 3. Instantly. Cold Email Benchmark Report 2026. 2026.
4. 4. Benway, J. P. and Lane, D. M. Banner Blindness: Web Searchers Often Miss Obvious Links. Rice University, 1998.
5. 5. Pagendarm, M. and Schaumburg, H. Why Are Users Banner-Blind? Journal of Digital Information, 2001 (click rates of 2% in 1995, 0.5% in October 1998 and 0.2% in May 2000; the 44% figure for the first HotWired banner in 1994 is the commonly reported industry figure).

How to cite

Sales.co. "Cold email is dying and there is nothing you can do about it." September 27, 2026. https://sales.co/research/cold-email-is-dying