TL;DR
- Founder Warning: Y Combinator founder Paul Graham warns that AI-written founder emails can damage trust before a pitch has time to persuade.
- Research Context: September 2025 BetterUp research linked AI-generated text to weaker recipient reactions and annoyance.
- Workplace Cost: BetterUp estimated shallow AI-generated writing can cost 186 USD per employee each month in cleanup time.
Y Combinator founder Paul Graham has warned that AI-written founder emails can feel like being lied to, framing generative-AI outreach as a credibility problem before the pitch itself has much chance to work. He reduced the reaction to six words: “It feels like being lied to.”
What gives Graham’s remarks an ironic twist is the fact he was the main mentor of OpenAI CEO Sam Altman, who indirectly is somewhat responsible for the surge in AI generated content since the launch of ChatGPT.
Founder emails often reach investors, recruits, or partners before there is a product demo, customer reference, or existing relationship to lean on. For Graham, emails written by AI are easy to dismiss once the note reads as machine-made, which makes the issue less about drafting speed than about whether the sender still sounds accountable for the message.
A lot of the emails I get from founders are now written in a hard-hitting journalistic style. I know they’re written by AI, because no founder ever wrote this way before. And once you realize something is written by AI, it’s hard not to ignore it.
— Paul Graham (@paulg) May 25, 2026
Why AI-Written Outreach Can Backfire
Cold outreach is unusually exposed to that test because the first note has to carry judgment, voice, and intent on its own. Graham links obvious AI use to weak writing or deception, and his short line “Any teenager can do that.” marks the distinction he sees between editing help and outsourcing the core voice of a pitch.
Investor inboxes make the penalty immediate. A founder can lose the persuasive window before the underlying idea is considered if the note sounds generic, detached, or machine-made. Graham’s broader point is that recipients do not have to prove who typed every sentence to react against language that no longer feels owned by the sender.
Many founder notes now arrive in a hard-hitting journalistic style that feels separate from the person asking for attention. Polished language alone does not create trust when the message is supposed to introduce a founder, not just summarize a company.
AI-written founder emails feel like being lied to.
— Paul Graham (@paulg) May 26, 2026
September 2025 workplace research from BetterUp and the Stanford Social Media Lab pushed that concern into a broader setting. One survey of 1,150 desk workers tied shallow AI-generated writing to cleanup work, and 40 percent believed they had received AI-generated workslop in the previous month. Another figure from the same September 2025 package found 53 percent of respondents felt annoyed by that kind of content, suggesting the friction starts before any direct correction or rewrite is counted.
BetterUp’s own estimate then moved from reaction to operational cost. The company said such incidents cost 186 USD per employee each month, based on time spent reading, clarifying, and fixing shallow AI-generated output.
Those figures do not prove how every investor will respond to an AI-polished founder pitch, and they should stay in the background rather than the article’s proof spine. They do help explain why Graham’s complaint can resonate beyond venture capital: once writing feels generic or unowned, convenience for the sender can become extra labor and lower confidence for the reader.
