For him, once it becomes clear that a text was generated by AI, it is difficult not to ignore it. Graham went even further, saying that he has never finished reading an email that was signed by a human but written by artificial intelligence. To him, it feels dishonest.

Graham is not against using AI in general. His point is that it should be used in the right way. In his view, AI-written messages make him think less of the sender. They suggest that the person either cannot write well without help or believes they cannot, and that they are trying to pass machine-generated text off as their own. Letting AI write for you, he argues, is not impressive — any teenager can do that.

This position is notable because Graham is not hostile to artificial intelligence. Y Combinator was among the early investors in OpenAI and remains deeply involved in AI-related startups. His criticism is not really about AI itself, but about trust, effort, and authenticity in communication.

Poor use of AI can quickly damage trust. Research from Ohio State University found that recipients tend to evaluate AI-generated messages more negatively because they feel the sender delegated the effort to a machine instead of taking the time to communicate personally.

The study’s lead author, Bingjie Liu, explained that AI use can be perceived as laziness or a lack of sincerity. Recipients may feel less secure and less satisfied in their relationship with the sender. Liu also suggested that people may now be running a kind of unconscious “Turing test” in their heads, scanning messages for AI-like patterns.

That is exactly what Graham describes. He recognizes the AI style immediately because, in his view, founders never used to write like that. Once the use of AI becomes obvious, attention shifts away from the content of the message and toward the question of how little effort the sender put into the interaction.

A separate study by BetterUp Labs, conducted with the Stanford Social Media Lab, surveyed 1,150 U.S. employees and found that 40% regularly receive shallow but polished AI-generated content from colleagues.

The social consequences are significant: 53% of respondents said such content annoys them. Around half viewed senders of so-called “workslop” as less creative, less capable, and less reliable, while 42% considered them less trustworthy. One-third said they would prefer to work less with those people in the future.

The BetterUp researchers distinguish between “pilots,” who use AI deliberately and responsibly, and “passengers,” who mainly use it to avoid doing the work themselves. Graham appears to place founders who send AI-written emails firmly in the second group.