Medium is for human storytelling, not AI-generated writing

Updating our approach to AI

Scott Lamb
The Medium Blog

--

Photo by Gift Habeshaw on Unsplash

When we wrote about how we’re approaching AI-generated writing on Medium back in January, we were still in the early phases of understanding how generative AI was going to play out, both here on Medium and across the internet in general. Over the course of the past few months, things have certainly accelerated, in both expected and unforeseen ways.

Today, I want to give an important update and clarification to one key part of Medium’s approach: We are a home for human writing, full stop.

Readers tell us constantly that they want stories written by people about their experiences and insights. While we understand that AI technology can assist writing in many ways, stories that are generated entirely by AI don’t fit with Medium’s mission of being a place for high-quality human stories that deepen people’s understanding of the world.

Our distribution system already filters out a lot of fully AI-generated stories. Although AI detection tools are currently unreliable to the point of being unusable, the human editors and curators who contribute to our recommendation system often spot it instantly. Likewise, we are hearing from publication editors across Medium that they are seeing too many submissions that are obviously AI-generated. At its worst, AI-generated text is filled with hallucinations, which is the word the AI industry is using to describe egregious mistakes. But even at its best, AI-generated text is bland, lacking in human wisdom, and filled with subtle deficiencies.

To reflect all this, we’re amending our distribution guidelines to be more clear on this point. We’re also making a small change to clarify that we allow the responsible use of AI-assistive technology, rather than welcome it. This may seem like a small difference in wording, but it’s an important distinction. Medium is a place that is focused on substance. In the best case scenario, AI-assistance empowers a writer to level up—to make their ideas clearer, for example, or help them express themselves in a second language. However, in practice, AI-assist is often a way to insert small snippets of AI-generated substance that often suffers from the same problems as fully generated stories. We both recognize the value and potential of AI to assist writers, and are committed to our core mission of being a home for human writing.

To be clear: Your blog on Medium is your home to publish whatever you want and we imagine that some of you may want to experiment there. So if you’re wondering whether you can still publish fully AI-generated stories for your followers, the answer is yes. However, if you are on Medium looking to find readers from our network, then the answer is no — we will not promote fully AI-generated stories to a broader audience.

Here are the new relevant parts of the distribution guidelines:

Medium is not a place for fully AI-generated stories, and 100% AI-generated stories will not be eligible for distribution beyond the writer’s personal network.

We currently allow the responsible use of AI-assistive technology on Medium. To promote transparency, and help set reader expectations, we require that any story incorporating AI assistance be clearly labeled as such. AI-assisted text without a disclosure at the beginning of the story (within the first two paragraphs), or other AI-generated content not labeled as such (for example, AI-generated images should include captioning identifying them as such, along with proper sourcing) will similarly be restricted to distribution on the writer’s personal network.

You can read more about distribution to your personal network here. If you see a story on Medium that you think is fully AI-generated, the best way to avoid similar stories is to hit the “Show less like this” button at the top of the story page.

We’re working on new ways to reduce the reach of fully AI-generated content on Medium, and also on increasing transparency around our approach to AI. In that spirit, we’re planning to publish a Q&A with our Trust & Safety and Content teams all about AI. What questions do you have for our team? Leave them in the comments—we may not be able to respond to every question, but we’ll publish a follow-up post later with answers, examples, and thoughts from our team.

--

--

Scott Lamb
The Medium Blog

VP, Content @ Medium. I'm here to support people writing words on the internet. Priors: BuzzFeed, YouTube, Salon.com