We Analysed 10,000 Cold Emails. Here's What We Found.

Over the past three months, Captchainbox's sender verification system has intercepted thousands of cold emails sent to users' Gmail inboxes. With user consent, we analysed a sample of 10,000 of these emails — messages from unknown senders who were challenged to verify and did not complete verification — to understand patterns in the AI cold email industry.

Here's what the data shows.

Key Findings

97.2% of unknown sender emails were never verified — indicating the sender had no genuine intent to reach the specific recipient
83% showed signs of AI generation — based on linguistic analysis patterns (consistent fluency, formulaic structure, research-heavy openers followed by templated pitches)
The average user received 31 cold emails per week from unknown senders
68% of cold emails came from domains less than 6 months old — indicating purpose-built sending infrastructure
Only 2.8% of challenged senders completed verification — and nearly all of those were legitimate human contacts

Sending Patterns

When cold emails are sent

Cold email sending follows clear patterns optimised for inbox visibility:

Peak days: Tuesday (22%), Wednesday (21%), Thursday (19%)
Peak hours: 8-10 AM recipient's local time (34% of all sends)
Lowest volume: Saturday and Sunday (combined 4%)

These patterns match the "best time to send cold email" advice published by every cold email platform — confirming that most campaigns follow the same playbook.

Follow-up cadence

Of the 10,000 cold emails analysed:

62% were initial outreach (first email)
24% were first follow-ups (typically 3-4 days after the initial email)
10% were second follow-ups (7-10 days after initial)
4% were third or later follow-ups

The standard cold email sequence is 3-4 emails over 2-3 weeks, with each follow-up typically adding a new angle or increased urgency.

Content Analysis

Most common opening patterns

AI-generated cold emails follow recognisable patterns in their opening lines:

"I noticed your [LinkedIn post/article/talk]..." — 28% of emails opened with a reference to the recipient's public content
"Congrats on [company milestone]..." — 14% referenced a recent company achievement or announcement
"As a fellow [role/industry] professional..." — 12% established perceived commonality
"[Mutual connection] suggested I reach out..." — 8% claimed a referral (often vague or unverifiable)
"I've been following [company name]..." — 7% claimed ongoing interest in the recipient's company

Most common call-to-action

"Would you be open to a quick 15-minute call?" — 41%
"Can I send you more information?" — 22%
"Would it make sense to connect?" — 18%
"Are you the right person to talk to about this?" — 11%

Subject line patterns

Short and personal: "Quick question" (9%), "[First name]?" (7%), "Idea for [company]" (6%)
Re: or Fwd: tricks: 4% of cold emails used "Re:" in the subject to simulate an ongoing conversation
Average subject line length: 4.2 words

Sender Infrastructure

Domain age

68% of sending domains were registered less than 6 months before the email was sent
23% were 6-12 months old
Only 9% were from established domains older than 1 year

This confirms the domain rotation strategy: cold email operations register new domains, warm them up over weeks, use them for campaigns, and rotate to new domains when reputation degrades.

Email provider distribution

Google Workspace: 52%
Microsoft 365: 31%
Other (custom SMTP, dedicated email infrastructure): 17%

Verification Behaviour

Of the 10,000 emails that triggered verification challenges:

2.8% completed verification (280 senders) — nearly all legitimate human contacts
4.1% clicked the verification link but didn't complete the CAPTCHA — suggesting curiosity or abandoned attempts
93.1% never clicked the verification link — consistent with automated sending where no human monitors the replies

The 2.8% verification rate is the most important number in this analysis. It means that for every 100 emails from unknown senders, approximately 3 are from real humans who genuinely want to reach you. The other 97 are noise.

What This Means

The data confirms what many professionals intuitively know: the vast majority of email from unknown senders in 2026 is AI-generated cold outreach with no genuine intent to engage. Sender verification correctly identifies and separates the signal (2.8% of genuine contacts) from the noise (97.2% of mass outreach) — without evaluating a single word of content.

Methodology

This analysis was conducted on 10,000 anonymised emails intercepted by Captchainbox's verification system over three months (December 2025 — February 2026). All emails were from senders who were not on any user's whitelist and who received a verification challenge. Content analysis was performed using automated linguistic pattern matching. Sending infrastructure data was collected from email headers. No personally identifiable information about senders or recipients was retained.

Frequently Asked Questions

How did you determine emails were AI-generated?

We used linguistic pattern analysis looking for: consistent high fluency, formulaic structure (research opener → value proposition → call-to-action), absence of personal voice markers, and correlation with known AI cold email templates. This is an estimate — definitive AI detection is not possible with current technology.

Could some of the unverified emails have been legitimate?

Possible but unlikely in significant numbers. Legitimate senders who don't verify may have missed the auto-reply, been confused by the process, or decided the recipient wasn't worth the effort. User reports of "important missed emails" from unverified senders are extremely rare — less than 0.1% of all unverified emails are later identified as important when users review their archives.

Will you publish this data regularly?

We plan to publish updated analyses quarterly as our dataset grows. Tracking trends in cold email patterns, verification rates, and sending infrastructure over time will provide valuable data for the email security community.