Paul Graham (폴 그레이엄), a Silicon Valley figure and co-founder of Y Combinator, publicly criticised the practice of using artificial intelligence (AI) to write promotional emails.

On May 26 (local time), blockchain outlet BeInCrypto reported that Graham said the approach was closer to deliberate deception than efficiency.

In a recent post on X, formerly Twitter, he said emails written by AI stand out immediately in their writing style. He said a journalism-style tone that founders did not previously use began appearing in emails after the spread of large language models (LLMs). He said the problem was not the subject line or the proposal, but the sentences themselves. He said an overly polished style does not look like something a founder wrote personally.

He said the focus of the criticism was not the quality of the writing but who wrote it. If a human signs an email but AI wrote the actual sentences, he argued, it is not convenience but an act that misleads the other party. Graham said, "I have never finished reading an email signed by a human but written by AI," and added, "It feels like being lied to, and who would stand for that?"

Graham also dismissed claims that AI ghostwriting improves a founder's efficiency. He said AI-written text can be seen as a sign of lacking the ability to communicate independently and as an attempt to make the reader misunderstand. He said handing writing to AI does not show an ability to use resources, adding, "Even a teenager can do this."

Graham did not reject AI use across the board. When some pointed to his past remarks that AI boosts the growth speed of Y Combinator startups, he said AI, like other technologies, must be used properly.

The remarks are drawing attention as AI-generated content quickly spreads across work emails and messages. The more automated phrases flood inboxes, the more investors and recipients may focus on substance over expressions, and who wrote it may emerge as a standard of trust. In this situation, sentences written by founders themselves are becoming a point of differentiation.

In the industry, there is also growing recognition that using AI for drafting or polishing differs from sending a finished message while hiding who wrote it. As generative AI becomes more common, authenticity and specificity of a message may become more important than how complete the sentences are.

Ultimately, the core of the debate is not whether AI is used, but how far it can be considered human communication. Graham said using AI as a work tool is different from using it to create the first point of contact to build relationships in place of a human voice. For founders who must persuade investors or customers, the ability to explain directly and take responsibility is also becoming an evaluation standard, as much as the ability to use technology.

I have never knowingly finished reading an email signed by a human but written by AI. It feels like being lied to, and who would stand for that?