The tagline on Jeffrey Archer’s debut novel Not a Penny More, Not a Penny Less, reads “Don’t get mean. Get even.”
It’s a good phrase.
It implies a preference for ruthless pragmatism over outrage. So in the spirit of not getting mean, and remaining on task, I am penning this short article imploring people to write their own README files by hand, without an LLM.
AI-generated READMEs are of course just one small part of the bigger problem of AI-generated writing, but here I am focusing specifically on READMEs, and why I believe it is better to write them yourself.
“Ight, imma head out”
Our 2026 brains are still at the point where AI-generated writing is off-putting, and the tells are very obvious. It remains to be seen how this plays out for mankind, but the fact is that people are going to exit the repo much faster if they encounter an AI-generated README. From a “business” perspective, if you want people to look at your work and consider using it, you’re losing a big chunk of your audience straight away by having an LLM write the README.
An AI-generated README signals that you don't care that much about your potential users. It’s like a business having a robot at reception. It might succeed in getting the visitor to the right place, but emotionally, the whole interaction is off, and it doesn’t give the feeling that the business cares about the visitor’s experience. Having the very first interaction with a software project be “The creator didn’t personally take the time to tell me anything about what they made” isn’t very appealing.
Continuing the robot at reception illustration, the visitor might wonder “are all the employees here also robots?” There’s nothing like an AI-generated README to tell readers that the code itself is AI-written as well. That’s not an inherently bad thing — good luck getting to the bottom of that discussion — but it will make people wonder if you put any critical thinking into the software. It’s possible that an AI-generated software project could be useful, but not everyone is going to take the time to try it out if they get that feeling.
Write it yourself
Anyone who’s ever written anything knows that the act of writing forces a degree of thinking that greatly aids the output. The writer goes through many loops of thinking and writing, thinking and writing, and then comes out with a good quality piece of work. For a software project, the act of writing a README forces you to ensure that there is a simple way for people to come to grips with the program and start using it. If through the README writing process, you get stuck trying to explain to users how to get started, it’s a sign that something basic about the program isn’t right and should be fixed.
This doesn’t need to take very long. If you have a good mental model of the software, writing a high-quality README can easily take under an hour. If you don't have a good mental model of what you've made, then the repo is probably not destined for greatness anyway. And if you feel that an hour of “extra” work without an LLM is too much, how can you expect users to give their time to get accustomed to the program and start using it?
So to all creators of software, if you want to vibe-code the repo itself, we’re probably past the point of no return on that front, but please, please, at least write your own README. Your future users will thank you.