AI Cold Email Templates That Don't Read Like AI

Why most AI cold emails fail now

The average cold email gets a reply about 3% of the time, and the gap between that and the 10%+ the best teams reach has almost nothing to do with having a fancier AI. It comes down to relevance and deliverability. When everyone uses the same model with the same lazy prompt, inboxes fill with near-identical emails, and spam filters fingerprint the patterns. The more people lean on "quick question" and "I hope this email finds you well," the faster those exact phrases get flagged.

There's a deeper point hiding in the data. Five minutes of genuine research before sending lifts reply rates several times over compared with template blasts. So the highest-leverage move isn't automating the writing. It's automating the research and the context, then letting a trained AI write something that sounds like you did your homework, because it did.

Train your AI before you write a single email

Most people open ChatGPT, type "write a cold email to a marketing director," and paste whatever comes out. That's why it sounds like everyone else. Instead, spend twenty minutes once to build a context base your AI keeps, using a Custom GPT, a ChatGPT Project, or a Claude Project. Feed it five things, and every email afterwards gets sharper.

• Your business, products, and services. What you sell, who you sell it to, the specific problem you solve, how you're genuinely different, and the proof to back it up.
• Your voice and writing style. Paste two or three of your best emails so the AI mirrors how you actually sound, not how an AI thinks a salesperson sounds.
• Your client's industry. The language insiders use, the pressures they face, and the jargon that marks you as someone who belongs in their world rather than an outsider selling in.
• Research on the specific prospect. Their LinkedIn "About" and recent posts, company news, and relevant social handles, so the email references something real about them.
• How your typical client writes. If your buyers are blunt and plain, you write blunt and plain. If they're warm and formal, you match that. Mirroring their tone earns trust faster than any clever line.

Here are the prompts to build that context base. Copy, fill in the brackets, and paste them into your Custom GPT or Project instructions.

You are my cold email writing partner. Learn my business below and follow these rules in every draft.

MY COMPANY
- We are: [company + one line on what you do]
- We sell: [product or service]
- We help: [ICP: industry, company size, the role we sell to]
- The problem we solve: [the painful, specific problem]
- Why we're different: [your real edge, not "quality and service"]
- Proof: [1-2 specific results, e.g. "cut new-rep ramp time 40% for a 200-person sales team"]

MY VOICE
- Tone: [e.g. direct, warm, plain-spoken]
- I always: [e.g. write short sentences, lead with the prospect]
- I never use: [your banned words, e.g. "synergy", "solutions", "I hope this finds you well", "circle back"]
- Two emails that sound like me: [paste 2 of your best]

RULES FOR EVERY EMAIL
- Under 120 words. One idea. One ask.
- Open with something about THEM, never about us.
- No jargon, no hype, no "I hope this email finds you well."
- A low-friction CTA: ask if it's worth a conversation, not "15 minutes Tuesday?"
- Write like a human who did five minutes of homework.

Confirm by summarising my business back in three lines, then wait for a prospect.

Here are [3-5] emails I've written that landed well: [paste them].

Analyse my writing and give me a short "voice guide" I can reuse: my typical sentence length, how I open, how I close, the words and phrases I lean on, and my level of formality. Then list five words or phrases I should keep using, and five AI clichés to keep out. Keep the guide under 200 words.

I'm about to email [name], [title] at [company].

Here is their LinkedIn "About" and three recent posts: [paste].
Here is recent company news or their website "About": [paste].
Here are their social handles if useful: [paste].

Give me:
1. Three specific things about them or their company I could genuinely reference (no flattery).
2. The one business problem they're most likely feeling right now, in their words.
3. How they write and talk: formal or casual, and the terms their industry actually uses.
4. One angle connecting what we do to what they care about.

Do not write the email yet. Just the research.

Using the research above, tell me how I should sound to THIS prospect.

- Match the vocabulary of [their industry]: the words an insider uses, not outside jargon.
- Mirror the tone of their own writing from the posts I pasted. If they're plain and blunt, be plain and blunt. If they're warm, be warm.
- Flag any terms I should avoid because they'd mark me as an outsider.

Give me a four-line "how to sound to this person" note before we draft.

The craft of a cold email that earns a reply

A cold email has one job, and it isn't to close. It's to start a conversation. The moment you try to sell the whole thing in 120 words, you sound like every other email they delete. Your goal is for the prospect to think, "this person understands my world and might actually help," and then reply. Hit that and you clear the 5% average easily.

The anatomy

The rules that actually move replies

• Keep it 50 to 125 words. One idea, one ask. Long emails read as work, and work gets ignored.
• Lead with them, not you. The first line should never start with "I" or your company name.
• Be specific. "We help companies grow" says nothing. A real number, a named-style result, or a precise observation earns attention.
• Make it easy to say no. A low-pressure ask ("worth a chat, or not right now?") gets more yeses than a demanded calendar slot.
• Write for a phone screen. Most emails are read on mobile. Short paragraphs, no walls of text.
• Follow up, properly. Roughly 4 in 10 replies come from follow-ups, not the first email. Plan a short sequence where each message adds something new.

Phrases that flag you as a template

These are so overused that filters and prospects both pattern-match them instantly. Cut them from every draft, and tell your AI to ban them.

The prompts that write it

With your context base trained and the craft in hand, these are the prompts that turn research into a finished, human-sounding email. Run them in the same Custom GPT or Project so the AI keeps your business and voice in mind.

Write the cold email to [name] using my business context, the research, and the voice note above.

- Under 120 words. One idea. One ask.
- Line 1: a specific observation about them or their company from the research, not about us.
- Then one sentence connecting that to a problem we solve.
- Then one specific proof point (a result or number).
- End with a soft, low-friction CTA: ask if it's worth a conversation, not "do you have 15 minutes Tuesday?"
- No "I hope this finds you well", no jargon, no hype.

Give me two versions: one slightly warmer, one more direct. No subject line yet.

Write 8 subject lines for that email.

- Under 7 words each. Sentence case, not Title Case.
- Curiosity or relevance, never clickbait. No "quick question", no fake "re:".
- At least 3 should reference something specific about them or their company.
- No emojis. They should read like a colleague wrote them, not a marketer.

Build a 4-email follow-up sequence for [name] if they don't reply, spread over about two weeks.

- Each email adds something new: a relevant idea, a useful resource, a different angle, or a short proof point.
- Never "just bumping this" or "circling back". Each one earns its own open.
- Keep each under 80 words.
- The last one is a graceful break-up that leaves the door open.

Here's a draft email: [paste].

Rewrite it so it reads like I wrote it, not an AI:
- Vary the sentence length; break the smooth, even rhythm.
- Cut every AI cliché: "I hope this finds you well", "in today's landscape", "seamless", "leverage", "circle back", "I wanted to reach out".
- Make it sound like a specific person typed it quickly, with a point of view.
- Keep it under 120 words, with my one idea and one ask intact.

Show me only the rewritten email.

Generic AI vs. trained AI

What good looks like

Mistakes that send you to spam

• Using the template everyone uses. Overused patterns get fingerprinted by filters. The "quick question" subject line dies a little more every time it's sent.
• Sending from your main domain. One bad campaign can burn the reputation your whole company emails from. Use a separate sending domain.
• Skipping authentication and warmup. Deliverability is the biggest variable now. No SPF, DKIM, DMARC, or warmup means spam, regardless of copy.
• Writing about you, not them. If the first line is about your company, you've already lost the reader.
• Sending once and giving up. Most replies arrive on follow-ups. A single email throws away the majority of your pipeline.
• Automating the writing instead of the research. Let AI do the homework and the heavy lifting; you keep the voice and the judgement.

Where cold email fits

A great email sent to the wrong person at the wrong time still fails. Cold email is the first touch in the Nurture stage of the AI-ENABLE framework, and it only works when the targeting before it is right. Start by finding the people already showing intent in our guide to AI lead generation, and sharpen the prompting craft itself in how to use ChatGPT for sales. Train the AI, get the timing right, and the email almost writes itself.

AI cold email: FAQ