The AI Domino Effect: Medium’s Policy Impact on Top Writers.
Unravelling the AI Paradox: Why 100% AI-Generated Content Isn’t the Core Issue.
A recent post on The Medium Blog raised concerns over fully AI-generated articles bombarding various publications on Medium and how they are trying to react to it. Now, it isn’t too difficult for humans to spot this content or at least have strong suspicions about the sources. But, with the volume of content published daily on Medium, 47,000 according to Medium, it’s easy to see how more than just a few will slip through the filtering process.
There are tools to help with this process, but they are not 100% reliable. The most worrying problem with AI detection tools would be the return of false positives. According to Medium: Medium is for human storytelling, not AI-generated writing.
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.
But the methods Medium is, apparently, using to counter this is having other, much more severe effects. For the established top writers, the impacts of recent changes are significant. And not in a positive sense,
See more articles, posts, and discussions about business, project management, Generative AI and Creative Writing on Medium here. If you have not already, subscribe with my Medium Referral Link. Or follow me here on Substack.
Disclosure. I use Generative AI tools to help me when writing. From outline suggestions to topics or subtleties I had yet to think of.
The quotes reproduced in this article are taken verbatim from the sources. As such, there are a few grammatical errors which I’ve left untouched.
Decoding ‘Responsible Use’: What Does it Mean in Medium’s Context?
The blog post reproduces an extract from the Medium Distribution Guidelines. This is the total of any references to AI-assistive Technology in the Guidelines as far as I can see:
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.
This is ok, as far as it goes, but what, then, is the Medium view on responsible use? In my recent articles, I’ve detailed some aspects of my creative process, which does include the use of several AI tools.
I Was in Trouble: How GPT-4 Rescued Me from Writer’s Block.
I even included the prompts I used with GPT-4 to help me overcome Writer’s Block. Is this responsible use or not? I can’t tell from the Medium statement, can you?
I use Grammarly to go through my almost-finished work. It always picks up some faults I can’t get right the first time. It is enormously helpful and has improved the grammar, structure, and tone of my writing over the months. But Grammarly is another AI tool, so is this frowned upon or not?
Incidentally, try running the extract from the Medium Distribution Guidelines through Grammarly or similar tools. Medium, apparently, doesn’t even use simple grammar-checking software. But I suppose it could just be my settings, but I repeatedly have to tell Grammarly to ignore the fault in the text.
Amidst the Controversy: A Glimmer of Hope in Medium’s Approach
To be fair to Medium, they finish the blog post with the honest admission that they are working on new ways to clarify the use of AI assistance and are planning to publish a Q&A about AI, hopefully soon. I will be participating as soon as it is available.
It can’t be easy to see the way forward as the goalpost constantly moves. But, we need some generally accepted, perhaps high-level, guidelines on what is acceptable. So far, this article has not benefited from AI assistance (apart from Grammarly, which is currently monitoring my work). I will, however, be looking to improve the title, subtitle and section headers. My working text may or may not survive this process.
Hopefully, this will be seen as acceptable. My starting point for the last few articles has mostly been with a synopsis of what I want to write about, asking for suggestions on titles and outlines etc. Last week I asked you, the readers, if they considered my process cheating:
From Drained to Inspired: Rekindling Creativity in High-Pressure Environments.
So far, I’ve not seen a single comment.
The Ostrich Syndrome: Can Ignoring AI-Generated Content Solve the Problem?
Some other platforms seem to be ignoring the AI-generated content problem completely. Substack, for example, doesn’t mention the term “AI” anywhere in their Content Guidelines or related documents. This discrepancy can be explained by the different business models implemented by Medium and Substack.
Members pay Medium a monthly subscription for unlimited access to any content hosted by Medium. This is their primary revenue stream, so anything discouraging paying members directly impacts income. Unsurprisingly, Medium suffers when the platform is peppered with low-quality, mass-produced, and soulless content. What better way to cause an otherwise loyal member to cancel? If they are constantly trying to sort out genuine articles of interest from the chaff of 100% AI-generated content.
The Business Model for Substack is similar, but the difference is significant. In Substack, all content is open, except content placed explicitly behind a paywall by the authors and publications. Users wishing to access this content must take out a subscription with the author or publication, and Substack takes a 10% cut. Say a publication pumps out one, or worse, more, obviously 100% AI-generated post; their subscribers would leave. More likely, such a publication on Substack would never gain any subscribers in the first place. Thus, zero impact on the revenue stream for Substack.
It doesn’t help that Substack is struggling, according to The Verge, so it’s no surprise that they currently have no incentive to address the issue of AI-generated content.
Heightened Worries: The Deep-Rooted Concerns for Medium’s Future
As far as Medium goes, other issues are much more concerning, none as significant as the recent changes to the Partner Program. I, and likely many others, have been working our asses off to improve the quality of our content enough to earn the 100+ followers required before being considered for the Partner Program. The same week, I finally hit the required 100+ followers, Medium announced their changes.
Frustrating, sure, financially impactful, not at all for me; after all, I’ve only just started. But the impact has been devastating and immediate for the established earners on Medium. Coincidently, while writing this week’s article, I noticed in my Inbox a weekly newsletter detailing the impact of the recent changes: Why I’m concerned about the future of Medium. This is an extract from the email:
Over the last three days, I’ve earned $27.48 – despite having one story boosted during that time. You don’t need to be Einstein to work out that I’m highly unlike to hit anywhere near my $2,000 per month earnings target if that continues. In fact, it’s a cataclysmic drop in revenue.
Final Thoughts
Medium, you need to pay attention to the negative impacts of these recent changes; the changes only seem to have been implemented due to the fear of AI-generated content. If this is the motivating force behind this all, then you are fooling yourselves. You are not helping either your members or Medium as a platform.
AI-generated content is an issue, but combatting it by killing off your loyal subscriber base is the most short-sighted and dumbest way forward. Take a breath, stand back and re-evaluate. The alternative is, to put it bluntly, terminal for the platform.
I’m now wondering what’s next and if it makes any sense to continue my subscription. It isn’t a lot, but perhaps I can use that €5,49 for something else.
KodifyIT B.V. is an advisory bureau targeting businesses that have either been on the receiving end of a failed project or are aware of the potential pitfalls and wish to mitigate as much risk as possible while developing a project’s client requirements. We aim to side-step any issues before they cost time and money.
I apologise to my readers for some of the spellings you may feel are incorrect. I was born and brought up in the United Kingdom, and this is the spelling I am comfortable with (Grammarly is happy with it anyway).