Why AI-written cold emails get ignored (and what to send instead)

The cold email problem in 2026 isn't volume — it's sameness. When every rep has access to the same AI writing tools, and those tools produce similar output given similar prompts, the result is an inbox full of emails that feel like they were written by the same person. They weren't. But they read that way. And buyers have noticed.

What an AI cold email looks like to a buyer

Most AI-written cold emails follow a recognisable structure. An opener that compliments the company or the role. A name drop to show it's "personalised". A vague value proposition about saving time, increasing efficiency, or driving growth. A low-stakes ask — "would you be open to a quick call?" or "would love to connect."

There's nothing factually wrong with any of these elements. The problem is the pattern. A buyer receiving 30 or 40 cold emails a day has learned to recognise it in the first sentence, and their hand is already moving to the delete key by the second. The email isn't bad. It's just invisible.

This is what happens when a tool lowers the cost of production: the floor rises but the ceiling stays the same. AI made it easy to send a cold email. It didn't make it easier to send a good one.

The 52% problem

Industry research consistently shows that 52% of buyers say sellers don't personalise enough — and that was before AI made generic, high-volume outreach the default. As more outreach gets generated by tools rather than thinking, the gap between what buyers expect and what they receive keeps widening.

Source: industry research (Hunter.io / Salesforce).

Personalisation doesn't mean using someone's first name or mentioning their company. It means showing you know something specific about what they're working on — and that your message is relevant to that thing, right now. That's a much harder bar to clear with a prompt template.

AI as researcher, not writer

The better model is to flip how you're using AI. Instead of asking it to write your email, use it to surface the signals that make an email worth writing.

What is this company doing right now? What are they hiring for? What did they publish last week? What has their CEO been talking about on LinkedIn? These are the raw materials for a cold email that doesn't feel like a cold email. AI is actually quite good at finding and surfacing this information quickly. It's not good at deciding what to say about it, or how to say it in a way that sounds like you.

When the signal is specific, the hook writes itself. You're not fabricating a connection — you're responding to something real.

What a signal-driven email looks like

Here's the difference in practice. Both examples below are for the same fictional company — Grovestack — and the same product being sold. The first is AI-generated from a basic prompt. The second is written from a signal.

The signal-driven version has the same word count. It took about five seconds to write the first line once the signal was surfaced. It reads nothing like an AI email — because it's grounded in a specific, observable fact about what the company is doing right now.

The hook didn't come from imagination. It came from noticing that Grovestack had three active job postings for growth marketing roles. That's public information. It took AI about three seconds to find it. But it took a human — the rep — to decide what that signal meant, and what to say about it.

The one rule for cold email in 2026

Sound like a human who did their homework. Not a robot who skipped it.

That's the whole thing. When AI has democratised volume and made generic output the default, the only remaining differentiator is specificity. A cold email that references something real — something the buyer recognises as true about their own situation — will always outperform one that doesn't. Not because it's longer, or more cleverly written, or better structured. Because it proves you were paying attention.

Triage approaches this by surfacing the signals first — the homepage positioning, recent news, job postings, tech stack, and other public indicators of what a company is focused on — and giving reps hook options to choose from. The rep picks the signal that's relevant to their pitch, and writes from there. The output sounds like the rep, because it is. The research just happened faster.