A Simple Plan to Save Medium

Before it's too late

Photo by Mick Haupt on Unsplash

This newsletter will be a bit unusual.

I’d prefer to write about making good money as a writer on Medium. However, that hasn't seemed possible in the last few months. My views (and pay from the partner program) have fallen off of a cliff.

Yet my writing has stayed mostly the same.

My last newsletter addressed the growing distrust between Medium and the writers who use its platform. This distrust was made worse by comments from Medium's leadership stating that earnings aren't down, despite most of us experiencing it firsthand. This newsletter is a plan to help make Medium the place we all want it to be.

But it's incomplete.

It needs your input. Some of these ideas may not be any good, or feasible to implement. If you have input, leave a comment. If it makes sense, I'll adjust the list. I'll post it as a Medium article and ping Medium staff when it's complete.

I'm not going to revisit the trust issue, but I am going to try to generate some ideas around how to fix the problem of AI-generated articles being passed off as human-written articles. I suspect this problem is why Medium says pay isn't down. It's a technicality: the money going in is being paid out to an increasing number of fake accounts, reducing pay for real writers.

Writers using AI are getting more savvy and earning more funds from Medium's partner program.

I may be out of my league here trying to fix Medium's challenges. When I was a cop, we knew we couldn't eliminate crime, but we could take a multifaceted approach to greatly reduce it. It was about deterring unwanted behavior while incentivizing the desired behavior. Knowing we won't eliminate AI-generate writing, I've taken that same approach here.

Here’s what I have so far:

Require a Writing Verification Process

Let's call this the human verification phase. New writers submit a short writing sample during account setup, which human moderators review to establish a writing style baseline. Require new writers to complete a brief written response in a monitored environment (browser-based, time-limited, and without external tools).

Randomized Human Reviews

Randomly select a percentage of articles for human review to assess writing style consistency and originality. Medium could then compare flagged content against the writer's previous work to identify drastic inconsistencies that indicate AI use.

Introduce a Strikes System for AI Abuse

A first violation results in a warning and a request for a rewrite. A second Violation results in a temporary account suspension and loss of earnings for flagged articles. A third and final violation results in a permanent ban from the platform. To ensure fairness, writers can appeal flagged content and speak with a Medium employee if requested.

Digital Fingerprints for Human Content

Develop a proprietary system that tracks human-generated drafts over time, capturing keystrokes, edits, and revisions—no more copy-and-paste. Medium can incorporate Grammarly-like capabilities within the platform. Simply removing the ability to paste AI content directly in Medium will deter a large percentage. Allow AI tools for brainstorming or outlining, but this software will automatically note when AI is used and in what capacity.

Create Incentives for Human Content

Offer engagement bonuses for writers who consistently pass human authorship checks. Medium can highlight top-performing human writers in a "Verified Writer Spotlight" to encourage authentic writing. To increase transparency with readers, create a "Human Certified" badge for verified human-written content.

Community & Peer-Review System

Allow experienced and trusted writers to help moderate content by flagging suspicious submissions. Medium could encourage peer verification by offering small incentives for writers who assist in maintaining content authenticity. This could function like "Community Notes" on Twitter.

Behavioral Analytics to Identify AI Users

Medium has (or should have) the ability to track sudden spikes in publishing frequency or changes in writing style. They can use software to detect suspicious activity, such as no pauses in typing or instant submissions without editing.

Final Thought

Eliminating AI content is nearly impossible, but making it unprofitable and inconvenient isn't. Providing a list to Medium isn't much, but brainstorming feels better than complaining. Like most Medium writer’s my pay and article distribution is the lowest its ever been.

I want to keep Medium as a platform we all enjoy, and I hope you'll help me.