I’ve been doing email marketing for over 30 years. Yes, I’m that old. 😬
I’ve seen every trend, every “revolution,” every tool that promised to change everything. But the AI cold email situation in 2024-2025 might be the most spectacular own-goal I’ve witnessed in my entire career.
Here’s what happened: the tools that promised to save cold email are actually destroying it.
And the data isn’t subtle about this. When personalisation is sacrificed for speed and volume, reply rates fall 13 times lower according to Lavender’s analysis of over a billion emails.
Open rates fell 23% year-over-year. And my personal favourite statistic: 95% of cold emails now generate absolutely zero response.
Zero.
But sure, let’s send another 10,000 emails today because the tool lets us. 🤷♂️
The Promise vs The Car Crash
Every AI cold email tool markets the same fantasy. “Hyper-personalization at scale.” Unlimited sending. Superior deliverability. Revolutionary conversion rates.
Instantly will let you send 2.5 million emails monthly.
Smartlead goes up to 60 million if you’re feeling particularly ambitious about annoying people.
They all promise their AI writes emails that “sound human”, and their proprietary technology ensures you “never worry about getting stuck in spam again.” —BTW, don’t trust this BS.
Except 88% of recipients now ignore emails they suspect are AI-generated. And 80% say they’d switch brands that rely too heavily on AI communication.
The emperor has no clothes 🍑, and the emperor is sending 60 million emails per month. 😅
What “Hyper-Personalisation” Actually Looks Like
Let me show you what these tools call personalisation.
I’m quoting from actual emails that actual humans received:
“Hi {{firstName}}, I noticed {{companyName}} is doing great things in {{industry}}.”
That’s not personalisation. That’s mail merge. We’ve had mail merge since 1980.
Or how about this beauty that landed in someone’s inbox:
“Hi Dave, I’ve been following your LinkedIn updates, and I have to say, your recent post on [specific topic] was insightful. Your perspectives on [specific issue or industry trend] are truly engaging.”
The AI left the placeholder text in brackets.
Just shipped it. The sender was literally selling cold email services.
Testing of five major AI personalisation tools revealed that 85-95% of “personalized” content is just templates with 3-5 fields swapped in. The remaining 5-15% that actually attempt research? It hallucinates about 15% of the time.
Real examples from the research: AI claimed someone wrote an article they never wrote, invented a fake movie title (“Turtle With the Golden Gun” – I wish I was making this up 🤦♂️), stated a newsletter had been running for three years when it hadn’t, and told someone their friend from college wrote for a newsletter that had zero guest writers.
The Irony Is Delicious
The most entertaining part of this mess?
Marketing automation experts keep receiving terrible AI cold emails from people selling marketing automation.
Kath Pay, CEO of Holistic Email Marketing and an international bestselling author on email marketing, received pitches for email marketing services. Someone attempted to sell email marketing to one of the world’s leading experts in email marketing 🤷♂️. The email praised her achievements, then suggested there might be “untapped potential in your email marketing strategy.”
Her response was everything. “Why are you trying to sell me the same services I offer to my clients?” She discovered the sender used “AI-optimised, spam tolerant mailboxes” and concluded: “Spam is still spam, even when it’s dressed up as AI.”
Another LinkedIn example showed someone receiving an email claiming “Btw, this isn’t an automated message blasted to 10,000,000 people – I’m a real person and I wrote this manually :)” while the email still contained {Company Name} as an unfilled placeholder text. 😂
You can’t write comedy this good.
How Recipients Spot Your AI Cold Emails Immediately
People aren’t stupid. They’ve developed sophisticated detection capabilities.
Primary signals include specific word choices that AI loves: “impressed,” “fascinated,” “intrigued,” “innovative.” If your email says, “I was impressed by your innovative approach” or “I was fascinated by your mission,” you’ve just announced you used AI.
Secondary signals cover tone and structure. Perfect grammar that’s too perfect.
Walls of text with no natural paragraph breaks. Excessive formatting with bullet points everywhere. Tone shifts between follow-ups where AI gets “stuck” on learned phrases. But, hey, some people are now trying to place mistakes on purpose to make it feel more “natural”. 🤦♂️
Tertiary signals are the obvious errors. Placeholder text left in. Wrong information. Awkward references that make no sense. Like the email asking “How was your trip last week away with your girlfriend?” based on Instagram stalking, which crossed from personalised into creepy surveillance.
One analysis documented nine specific fingerprints marking content as AI-generated. The clearest single tell? The word “impressed” appears so frequently in AI-generated cold emails that it has become a meme.
The Tools Profit From Volume, Not Your Success
Here’s the economic reality nobody refers to.
Entry-level plans start around €37-49 monthly for 5,000-10,000 emails.
Mid-tier plans unlock 100,000-150,000 emails for €79-97. Enterprise plans support millions of emails for hundreds of euros. (obviously this is a ballpark since there’s a ton of different services that provide this)
A user sending 500,000 emails monthly at €358 pays €0.000716 per email. The structural incentive is crystal clear: tools profit from volume, not quality. They optimise for scale rather than effectiveness.
This creates incentives catastrophically misaligned with what actually works. Every study shows that campaigns under 100 recipients achieve 5.5% reply rates (source) while larger campaigns see performance collapse.
Targeting 1-2 contacts per company generates 7.8% response rates. Blast 10+ people at the same company and you drop to 3.8% (same source) – more than halving effectiveness through sheer volume.
But the tools want you sending more. Always more. Because that’s how they make money.
The Legal Catastrophe That Might Creep Up
If you’re in Europe or targeting Europeans, AI cold email automation creates legal exposure that should terrify you.
GDPR fines have reached €5.65 billion across 2,245 cases by March 2025.
The top complaint? “I don’t know how you got my email.”
Most tools use scraped or purchased data. “I bought the list” is not a GDPR defence. “Legitimate interest” requires individualised assessment, incompatible with sending 10,000 identical emails daily based on job title alone.
CAN-SPAM penalties reach $51,744 per violation – per email. Sending 10,000 non-compliant emails creates theoretical exposure of $517 million. Even if not enforced per email, scale demonstrates “systematic violations” triggering maximum GDPR penalties: €20 million or 4% of global annual revenue.
The “only big tech gets fined” myth is dangerous nonsense. Spain alone issued 932 fines through 2024. Most are €5,000-€100,000 to SMBs – they just don’t make headlines.
What Actually Works (The Boring Truth)
The successful 5% who achieve 10-20%+ reply rates don’t use AI the way it’s marketed.
They use AI for research assistance, gathering insights, analysing company news, and identifying pain points.
Then humans make final decisions about relevance and craft messaging.
The time investment is 2-4 minutes per email for quality review. Seems expensive until you compare it to campaign failure rates of 95%.
They send fewer emails to better targets. Campaigns under 100 recipients outperform mass blasts. Quality beats quantity every single time the data gets examined.
They actually personalise beyond “I saw your LinkedIn post.” Real personalisation requires understanding persona beyond battlecards, account knowledge beyond industry classification, and value proposition knowledge beyond what’s on the website. This depth can’t be automated, but can be accelerated with AI research.
They fix technical fundamentals first. SPF, DKIM, DMARC authentication configured correctly. Email warmup for 30+ days for new domains. Bounce rates below 2% and spam complaints below 0.3%. Continuous monitoring of sender reputation. The shortcuts that poison deliverability make recovery nearly impossible without switching to new domains.
They follow up strategically, not obsessively. The first follow-up increases replies by 49-220% – the biggest gain. The second generates 20% fewer responses. By the fourth follow-up, response rates drop 55% with spam complaints tripling. The data-driven conclusion: 2-3 total touches, not 5-7. (source)
The Future Isn’t More Automation
The industry loves talking about “AI SDRs that outperform humans” and “unlimited scale.” But the data shows the opposite trend winning.
Companies excelling at targeted outreach generate 50% more sales-ready leads while cutting costs by one-third compared to spray-and-pray approaches. Highly personalised campaigns show a 142% boost in replies versus template blasts (source).
Multi-channel approaches combining email, LinkedIn, phone, and content dramatically outperform email-only strategies. LinkedIn InMail achieves 18-25% response rates – much higher than email alone. But email-only gets pursued because AI makes it too easy and scalable.
The successful model is “people with robots.”
AI handles research, draft generation, data enrichment, timing optimisation, and response categorisation. Humans handle final copy editing, relevance verification, relationship building, and strategic decisions.
This hybrid approach achieves 30-50% time savings versus fully manual while maintaining or improving response rates, dramatically outperforming fully automated approaches in both efficiency and effectiveness.
Stop Lying to Yourself
If you’re using AI cold email tools the way they’re marketed – sending thousands of template emails with light personalisation at scale – you’re not building a pipeline. You’re burning sender reputation, violating regulations, and training prospects to ignore your domain.
The middle path of “AI-powered personalization at scale” as currently marketed has proven untenable.
The data isn’t ambiguous about this.
You have three actual choices:
Dramatically reduce scale while increasing targeting quality to achieve genuine effectiveness and compliance. This means thousands of quality emails monthly, not hundreds of thousands of template blasts.
Accept significant legal and financial risk from current mass-automation approaches. Be honest about the exposure you’re creating.
Shift toward consent-based inbound marketing models less dependent on cold outreach. Build audiences who actually want to hear from you.
What you can’t do is keep pretending that AI solves the problem when AI created the problem. The tools enabled the volume that destroyed the channel.
The Bottom Line
After 30+ years in email marketing, I’ve learned that genius without practical application becomes worthless. AI cold email tools represent genius applied catastrophically.
The technology is remarkable. The implementation is terrible. And the industry’s response has been to double down on volume rather than admit the strategy failed.
Cold email isn’t dead. Mass automation nearly killed it.
Recovery requires returning to first principles: relevant messages, genuine research, proper targeting, technical excellence, and human judgment – accelerated by AI tools rather than replaced by them.
The successful 5% already figured this out.
They’re using AI as a research assistant, not an autonomous writer. They’re sending hundreds of quality emails, not hundreds of thousands of template blasts. They’re building sustainable practices that improve over time rather than burning lists and domains.
The question is whether the other 95% will learn from the data or keep buying tools that promise unlimited scale while delivering unlimited spam complaints.
I know which way I’d bet. But then again, I’ve been watching people ignore obvious truths about email marketing for three decades.
Some things never change. 🤷♂️
Leave a Reply