AI Prompts for Cold Email & Sales Outreach
8 fill-in-the-blank prompts for cold email copy, follow-ups, objection handling, and call prep — plus the deliverability and legal facts that actually decide whether cold email works.
Fill in the blanks below — each prompt updates live. These work the same in ChatGPT, Claude, and Gemini; nothing here is tied to one specific model.
How it works
Every prompt below is plain-language and works the same in ChatGPT, Claude, or Gemini — the idea that this is a Claude-specific trick just describes claude.ai having a free tier, not something unique to how Claude writes.
The bigger thing the original framing skips: cold email reply rates are driven mostly by deliverability and targeting, not copywriting. A properly warmed sending domain gets roughly 87% inbox placement and around a 4% reply rate; the identical email sent from a cold, unwarmed domain lands in the inbox only about 12% of the time. If your reply rate sits under 3%, the more likely cause is your domain setup or your list — not the words, however well a prompt writes them.
Cold email is also regulated, and the rules differ by where your prospect is. In the US, the CAN-SPAM Act permits unsolicited B2B cold email without prior consent, but requires a truthful subject line and sender identity, a real physical postal address, and a working opt-out you honor within 10 business days — penalties run up to $53,088 per non-compliant email under 2026 FTC adjustments. The EU and UK are stricter under GDPR and PECR, and generally expect consent or a documented legitimate-interest basis, particularly for messages to an individual rather than a general business inbox — check the rules for wherever your prospects are before sending at volume.
Prompt #8 ("AI Outbound System") is framed as a design exercise on purpose: a single chat reply can lay out a workflow and flag where a human needs to review it, but it can't wire up a CRM or send anything on its own — turning the design into something that actually runs still means connecting it to real tools.
FAQ
Can AI actually write 'high-converting' cold emails?
AI can write the words, but reply rate is driven mostly by deliverability and targeting, not copy — a well-warmed sending domain gets roughly 87% inbox placement versus about 12% for a cold, unwarmed one. If your reply rate is under 3%, that's usually a domain or list problem, not a writing problem, and no prompt fixes a damaged sender reputation.
Is cold emailing legal?
In the US, yes for B2B — the CAN-SPAM Act permits unsolicited commercial email without prior consent, but requires a truthful sender and subject line, a real physical postal address, and a working opt-out honored within 10 business days. Violations can run up to $53,088 per email under 2026 FTC penalty adjustments. The EU and UK (GDPR/PECR) are stricter and generally expect consent or a documented legitimate-interest basis, especially for messages to an individual rather than a general business address.
Can prompt #8 actually build me a working AI sales automation system?
No. A chat reply can design the workflow — what AI should handle at each stage, where a human needs to check it, what tools it would need to connect to — but it can't wire up your CRM or send real automated emails by itself. Turning that design into something that actually runs still means connecting it to real tools: your CRM, a sending platform, or an automation tool like Zapier or n8n.
Do these prompts only work with Claude?
No. None of them are model-specific — they read the same way in ChatGPT, Claude, or Gemini. The idea that this is a Claude-specific trick just reflects that claude.ai has a free tier, not a capability unique to Claude.
What actually predicts whether a cold email gets a reply?
Personalization still matters — genuinely tailored openers get roughly 2-3x the response rate of generic ones — but only once deliverability is solid. If opens look normal but replies are near zero, the more common cause is a spam-filtered or under-warmed domain, not the wording.