How LLMs will kill Medium's business model

Medium is struggling to adapt its Partner Program to the deluge of LLM-written content. Here is what it means for centralized content platforms.

Medium recently notified its users that AI-generated content will not be eligible for their Partner Program, the company’s paywalled articles. The company will go as far as removing writers from its Partner Program if they put AI-generated content behind the paywall. It is unclear how the company will enforce this policy because AI detectors are inaccurate, which can result in human-written content being flagged as AI-generated. 

But what is clear is that Medium’s business model is struggling to adapt to the shifts caused by large language models (LLM). And Medium is not the only platform that LLMs are disrupting. LLMs will challenge any platform where a centralized authority controls the distribution and monetization of content. 

Why Medium’s business model doesn’t work will LLMs

Medium users can sign up for the Membership Program for $5-$15 per month. This gives them unlimited access to the paywalled Member-Only articles. Part of those proceeds will go to the writers of those articles, who are members of Medium’s Partner Program. 

Medium has set up a system to pay for high-quality content. For example, articles that are read for longer durations will earn more. Other factors such as engagements (e.g., replies and clapbacks) and follower counts will also improve compensation. At first glance, this seems like a solid policy that can withstand the deluge of low-quality LLM-written content that is flooding the internet.

But here is how it becomes problematic. Writers can generate a large number of articles with language models. Even if each article earns very little, the writers can compensate by generating more articles at a small cost. 

In the short term, this will degrade the user experience of paid members, who will be frustrated as they will be bombarded with a flood of low-quality articles. Authentic writers will also suffer as part of the limited supply of money coming from paid memberships will be siphoned by low-quality AI-generated articles that distract readers. 

The long-term effects will be more dramatic:

• Paid members will ignore Medium’s content curation algorithms and rally to a few writers they know and trust. 

• Part of the paid members will just leave the platform as the user experience will continue to degrade.

• As Medium mistakenly flags human-written articles as AI-generated, writers will start to lose their trust in the platform.

• The top-earning writers will not be affected much and will possibly benefit from the new rules.

• But many other writers will leave the platform or become less active as their earnings dwindle and their content gets wrongly censored by malfunctioning AI filters.

• New writers will have fewer incentives to join Medium because of low earnings and the difficulty of being discovered among the sea of AI-generated articles.

Measures to suppress AI-generated content will only provide a temporary reprieve and the long-term disruption of the business model cannot be avoided. At the end of the day, LLMs are here to stay and will change the digital content landscape. A lot of writing tasks will be done by language models, especially as they continue to get better. Just as banning the printing press did not prevent the adoption of printed material, bans on AI-generated content are doomed to fail.

The correct business model for content platforms in the age of AI

In my opinion, business models that will work in the age of AI-generated content are those that directly connect content creators to consumers. The question should not be whether the content is generated by AI or not—it should be quality. If you can use AI to create articles that readers enjoy and are willing to pay for, then you deserve to be compensated.

One example is Substack, where readers choose which blogs they want to pay for. Writers can also provide incentives such as trial access to their paid content so that readers can judge for themselves whether they will sign up for the paid version or not. Substack has an active community of writers who can interact and recommend each other. Writers also have more control over their audience. Every user who subscribes to your blog gives you their email address. You can easily reach out to them directly or decide to take your business elsewhere if you ever wish to do so.

In a fashion, Substack’s business model is the opposite of Medium. While Medium takes readers’ money and distributes a part of it among writers, Substack lets readers directly pay writers and takes a cut as the facilitator, like a bank that takes a fee off each transaction (though I honestly think 10% is a bit too much).

But my main advice is to the writers. Don’t invest in getting clicks, views, and a small share of the revenue of a huge content distribution platform. Write for the people, not the algorithms that will make your content visible. Try to reach out directly to the people who will read your articles. Find ways to create a tightly-knit community of readers who appreciate your content and are willing to pay for it. Robots might make the job of writing easier, but it is the humans who will push you to make your writing better.

You can project the same pattern to other content platforms where a central authority controls both distribution and monetization. For example, many website owners who are monetizing their websites through Google ads and search traffic have seen a severe drop in their revenue. This is partly due to the ripple effect caused by the growing masse of LLM content being distributed on the internet. 

Google controls both the distribution of content and ad revenue. It is trying to update its search algorithms in a way that prevents what it deems low-quality content (including a lot of AI-generated content) to increase its ad revenues, part of which goes to publishers. This has caused massive shifts in the way websites receive traffic from Google Search. Top-earning websites with high domain authority have not been hit hard but many medium and niche sites have suffered severely and several had to shut down.