AI Cold Email Prompts: 20 YC-Inspired Rules for Founders

Using generative artificial intelligence to draft cold emails often backfires when models produce verbose paragraphs, hollow flattery, and ungrounded claims. To generate outbound messages that prospects actually answer, lean founders and growth marketers should combine verifiable prospect data with the concise communication frameworks popularized by Y Combinator. This guide provides a research-based workflow, four copyable prompt templates, and strict quality assurance checks to turn AI models into disciplined drafting assistants rather than automated spam generators.

The YC Origin: 3 Sentences vs. 20 Social Rules

Cold outreach fails most often because senders demand too much reading time from busy executives. In 2016, Y Combinator Managing Director Michael Seibel published a seminal framework titled How to Email Early Stage Investors. Seibel established an exacting standard: the cold email must contain exactly three sentences.

Sentence one describes what your company does in plain language without buzzwords. Sentence two delivers the single most impressive, verifiable metric or traction point. Sentence three presents an explicit, low-friction ask. As highlighted in the YC Startup Library guidance on talking to investors, clarity and brevity signal operational maturity. Investors and business executives receive hundreds of pitches weekly; an overlong message signals an undisciplined communicator.

In recent years, creators and founders on social channels (such as StartupbyDOC) expanded these principles into widely shared carousels listing 20 rules for cold communication. While these 20 points offer valuable practical advice, it is vital to understand that Y Combinator did not author a formal 20-rule doctrine. Instead, those 20 rules represent a community synthesis combining Seibel’s 3-sentence brevity, classic sales engineering, and modern deliverability standards. Applying them successfully with language models requires feeding the AI precise constraints rather than vague requests for persuasive copy.

Before launching any outreach sequence, teams should evaluate their source copy against our AI marketing copy QA checklist to eliminate synthetic filler and unsupported claims.

Investor Outreach vs. B2B Sales Outreach

A frequent error made by founders using generative AI is treating commercial customer acquisition identically to investor fundraising. The core psychological dynamics differ significantly.

Seibel explicitly warned founders pitching investors: "Do not ask for a meeting or phone call in the first email… if they are interested, they will ask." In commercial B2B sales, however, offering a frictionless conversational next step is standard practice, provided you do not demand an inconvenient 30-minute calendar commitment on touch one.

The 20 YC-Inspired Rules Mapped to AI Workflows

To implement the 20 viral outreach principles without generating robotic prose, growth operators must clearly divide responsibilities: language models draft variations against strict structures, while human operators curate verified inputs and verify claims.

1. Read in Under 60 Seconds: Restrict LLM output to under 90 words total.
2. Three-Sentence Structure: Enforce strict three-sentence prompts for initial cold touches.
3. Describe the Offer First: State the exact utility of your product in sentence one; eliminate lengthy preambles.
4. Lead with Verifiable Proof: Ground claims in real numbers supplied directly in the prompt.
5. Explicit Ask: Instruct the model to end with a single, binary or low-effort question.
6. No Premature Meeting Demands: For investors, offer updates or deck links; for sales, ask if solving the target problem is currently a priority.
7. Skip Backstory: Ban sentences discussing how long you have worked on the problem or personal histories.
8. Avoid Generic Flattery: Disallow AI-generated compliments like "I love your recent post" unless grounded in a real, user-supplied quote.
9. Identifiable Company Email: Send from a authenticated Google Workspace or enterprise domain matching the brand website.
10. Conversational Language: Forbid corporate buzzwords such as "synergize," "streamline," or "revolutionary."
11. Genuine Relevance: Supply one real account trigger (such as a recent job posting, tech stack change, or expansion announcement).
12. Recipient Problem in Their Words: Use industry-standard terms sourced from customer research transcripts.
13. One Single Goal: Prompt the model to aim for a single response type rather than multi-option pathways.
14. Verifiable Credibility: Include only references, customer names, or accreditations that are demonstrably true.
15. Short, Honest Subject Lines: Use 2 to 4 words describing the specific topic without deceptive prefixes like "Re:" or fake meeting notices.
16. Clear CTA Near the Sign-Off: Keep the concluding ask unencumbered by extra attachments or links.
17. Target Fit Over Volume: Prioritize highly qualified accounts over mass outreach lists.
18. Manual Pilot Before Automation: Write and send the first 25 messages individually before scaling templates.
19. Thoughtful, Limited Follow-Up: Limit outreach sequences to 1 or 2 high-value touches rather than automated nag sequences.
20. Respect Non-Response and Opt-Outs: Suppress unresponsive contacts immediately to protect domain reputation and respect recipient privacy.

The 5-Step AI Outbound Workflow

Achieving consistent outbound results requires a disciplined sequence of human research, constrained AI drafting, and thorough verification.

1. Step 1: Define the Ideal Customer Profile (ICP) and Lawful Basis. Identify the specific titles, company sizes, and tech stacks experiencing the operational friction you solve. Verify your legitimate interest or local regulatory basis for contact before searching for email addresses.
2. Step 2: Collect Minimal, Verified Account Facts. Gather two or three indisputable details: the prospect’s primary responsibility, a recent public development at their firm, and their current workflow challenge. Never ask an AI model to research an individual’s private inbox or invent observations.
3. Step 3: Run Constrained AI Generation. Feed your verified facts into a structured prompt that mandates the 3-sentence limit and establishes hard boundaries on prohibited claims.
4. Step 4: Conduct Human Quality Assurance. Check every sentence against reality. Remove any residual conversational padding, awkward phrasing, or inflated promises.
5. Step 5: Execute Low-Volume Delivery and Measure. Send messages manually or via low-throughput staging. Evaluate positive sentiment responses, qualified meetings, and unsubscribe requests rather than opened email notifications. For post-reply handling, refer to our automated lead follow-up guide.

Cold Email Prompt Pack: 4 Production Prompts

The following four prompt templates can be copied directly into language models like Claude, ChatGPT, or local LLMs. Replace bracketed placeholders with verified operational details.

Prompt 1: ICP Fit and Evidence Extractor

Role: B2B Research Assistant.
Task: Extract 1 operational challenge and 1 verified company trigger from the supplied raw notes.
Constraints:
- Use ONLY facts explicitly present in the notes below.
- Do NOT invent initiatives, team sizes, or goals.
- Return output as two bullet points: Trigger, and Implied Friction.

Raw Notes:
[PASTE UNFORMATTED COMPANY / PROSPECT RESEARCH HERE]

Prompt 2: The Constrained 3-Sentence Commercial Pitch

Role: Direct B2B Outbound Specialist.
Task: Draft a 3-sentence cold outreach email using only the provided facts.
Structure:
- Sentence 1: State what our company does and connect it directly to the prospect's verified friction.
- Sentence 2: Provide our specific, verified customer metric or proof point.
- Sentence 3: Ask a single, low-friction question to see if exploring this friction makes sense.
Constraints:
- Total length must be under 75 words.
- No buzzwords (e.g., streamline, leverage, revolutionary, synergy).
- No generic compliments or flattery.
- Do not ask for a 30-minute call or send a calendar link.

Our Company: [COMPANY NAME] - [ONE SENTENCE PRODUCT SUMMARY]
Our Proof Point: [VERIFIED CASE STUDY METRIC, E.G., REDUCED QA TIME BY 42% FOR B2B SAAS]
Prospect Name: [NAME]
Prospect Company: [COMPANY]
Verified Trigger/Friction: [PASTE EXTRACTED EVIDENCE]

Prompt 3: Jargon and Hallucination Filter

Role: Skeptical Growth Copyeditor.
Task: Review the draft email below and eliminate unverified claims, corporate clichés, and synthetic filler.
Rules:
- Strip out polite preambles (e.g., "I hope this email finds you well").
- Verify that every claim reflects the Provided Evidence.
- Ensure the entire body contains exactly 3 clear sentences.

Provided Evidence: [PASTE VERIFIED FACTS]
Draft Email: [PASTE AI-GENERATED DRAFT]

Prompt 4: Low-Friction Single-Touch Follow-Up

Role: B2B Growth Marketer.
Task: Draft a polite, single-sentence follow-up message sent 5 business days after initial outreach.
Rules:
- Provide 1 additional piece of verifiable documentation (e.g., a relevant public case study link or teardown).
- State clearly that if the topic is not a priority right now, no further emails will follow.
- Word count under 40 words.

Original Subject: [SUBJECT]
Value Asset Link: [LINK TO VERIFIED RESOURCE]

Annotated Before and After Examples

Reviewing contrasting drafts demonstrates why strict constraints are essential when prompting AI tools for outreach.

Illustrative Example 1: B2B SaaS Prospecting

Unconstrained AI Output (Poor):

Subject: Synergizing Acme’s Revenue Engine with Cutting-Edge AI Solutions!

Dear Sarah, I hope this email finds you well! I have been following Acme Corp’s incredible journey and was so impressed by your recent expansion into enterprise fintech. In today’s fast-paced macroeconomic environment, revenue leaders struggle to manage disparate workflows. Our innovative, best-in-class AI platform revolutionizes pipeline analytics by streamlining team workflows. Would you be open to a 30-minute introductory call next Tuesday at 2 PM to explore a strategic partnership?

Why it fails: The subject line looks like promotional spam. The preamble contains empty flattery. The body relies on vague jargon ("synergizing," "best-in-class") without a concrete proof point. It ends with a high-friction request for a 30-minute calendar block.

Constrained 3-Sentence Output (Improved):

Subject: Acme billing reconciliation

Hi Sarah,

We built LedgerPulse to automatically match stripe invoices against enterprise bank deposits, cutting monthly close time from five days to four hours for scaling fintechs like PayFlow. Given Acme’s recent international expansion, are you handling multi-currency reconciliations manually right now?

Best regards,
Alex Rivera | LedgerPulse.io

Why it works: Exactly two sentences in the body (well under the three-sentence maximum). The offer is clear immediately. The proof point is specific and verifiable. The call to action is a low-friction diagnostic question.

Illustrative Example 2: Early-Stage Founder Pitching an Angel Investor

Constrained Output (Following Michael Seibel’s Rules):

Subject: TraceHealth seed round ($45k MRR, +28% MoM)

Hi Marcus,

TraceHealth automates clinical trial patient compliance tracking for regional biotech labs using HIPAA-compliant device APIs. Over the last six months, we grew to $45,000 in monthly recurring revenue across 14 enterprise health systems with zero customer churn. We are raising a $1.2M seed round; please let me know if you would like me to share our confidential memo and summary deck.

Best,
Elena Vance | Founder, TraceHealth

Why it works: Sentence one clearly defines the product without technical jargon. Sentence two provides verifiable revenue, growth rate, customer count, and churn metrics. Sentence three makes an explicit ask for permission to send materials without prematurely requesting a phone call.

Quality Assurance, Deliverability, and Legal Compliance

Writing concise copy matters little if messages trigger spam filters or violate privacy regulations. Cold outreach must adhere to technical standards and international laws.

Deliverability Standards

According to the official Google Gmail Email Sender Guidelines, domain authentication is mandatory. Senders must configure SPF, DKIM, and DMARC records for their sending domains. Google requires senders to maintain spam complaint rates below 0.10% in Google Postmaster Tools, and strictly warns that spam complaint rates should never reach 0.30% or higher. Furthermore, bulk senders directing messages to personal Gmail accounts must provide clear, one-click unsubscribe headers. Violating these technical standards will route your outbound domain directly to junk folders.

Regulatory Compliance

Cold outreach is strictly governed by regional legal jurisdictions:

• United States: The Federal Trade Commission enforces the CAN-SPAM Act Compliance Guide for Business. Commercial B2B communications are fully covered. You must never use deceptive headers or misleading subject lines, you must display a valid physical postal address, and you must provide a clear opt-out mechanism that is honored within 10 business days.
• United Kingdom: Under guidance from the Information Commissioner’s Office on B2B Marketing, rules under PECR distinguish between corporate subscribers (such as registered limited companies and LLPs) and individual subscribers (sole traders and unincorporated partnerships). Unsolicited marketing emails can be sent to corporate contacts without prior consent, but must identify the sender and provide an opt-out. Sole traders require prior opt-in consent. Processing corporate employee names remains subject to the UK GDPR, requiring a defensible Legitimate Interests Assessment (LIA).
• European Union: Under the ePrivacy Directive and the GDPR, member state implementations differ significantly. Several European countries (such as Germany) enforce strict requirements regarding unsolicited electronic contact. Outreach must always reflect a defensible legal basis, rigorous data minimization, and immediate suppression of non-responders.

Measuring Outbound Success Beyond Vanity Metrics

Tracking email open rates has become unreliable due to privacy protections like Apple Mail Privacy Protection, which pre-fetches tracking pixels regardless of human attention. Relying on open rates creates a false impression of performance.

Instead, growth teams should evaluate campaigns using commercial business metrics:

• Positive Reply Rate: The percentage of contacts who respond with interest or request additional information. Benchmark target: 3% to 8% on highly targeted batches.
• Qualified Opportunity Conversion: The percentage of sent messages that lead to a verified discovery meeting or technical demo. Benchmark target: 1% to 3%.
• Spam Complaint Rate: Must remain under 0.10% across all mailbox providers. Any campaign exceeding 0.20% requires immediate cessation and copy revision.
• Opt-Out Velocity: Senders should monitor unsubscribe requests. An opt-out rate above 2% on cold outreach suggests poor list segmentation or irrelevant messaging.

Frequently Asked Questions

Did Y Combinator officially author 20 cold email rules?

No. Y Combinator did not publish an official 20-rule outbound manual. The 20-rule list gained popularity through startup social carousels (such as StartupbyDOC), which synthesized startup communication principles. However, the foundational 3-sentence email doctrine was directly authored and published in 2016 by YC partner Michael Seibel.

Why shouldn’t I ask for a 15-minute call in my cold sales email?

In early-stage investor outreach, requesting a meeting prematurely signals disrespect for the investor’s screening process. In commercial B2B sales, asking for a meeting in your very first sentence introduces high friction before you have established relevance. Asking a diagnostic question or offering a useful resource yields lower friction and higher response rates.

Can I automate prospect research using an LLM?

Language models cannot reliably research private prospect attributes without access to real-time search APIs, and even then, they frequently hallucinate details. Founders and marketers should supply verified facts directly in the prompt to prevent the model from inventing fictional company initiatives or mutual connections.

How many follow-up emails should I send?

While social carousels often suggest four or five follow-ups, modern inbox deliverability guidelines and recipient expectations favor restraint. Sending one thoughtful follow-up that offers additional value or documentation, followed by graceful suppression if there is no reply, preserves domain reputation and avoids spam flags.