Social Media Monetizes Outrage. AI Monetizes Empathy. Both Are Lying to You.
- Luke Stoffel
- Jun 2
- 7 min read
Social media's business model rewards outrage. AI's business model rewards flattery. Same engine, opposite directions. We did not have a voice when the internet was being built. We have one now. Use it.
The internet runs on two business models. Both extract attention by manipulating your emotional state. They look like opposites. They are the same engine.
I am a writer. Over the past two years I have been on the receiving end of both, attacked by the first, seduced by the second. In May 2026 I won a 2026 IBPA Benjamin Franklin Award for a memoir I openly used AI on. I wrote up the full case for disclosure-first publishing in a companion essay on how I disclosed AI and won a national book award. This essay is the deeper diagnosis of what each business model produces when it gets pointed at a working writer.
How social media's business model monetizes outrage
The first business model is social media. The product is eyeballs. The platform pays out in attention, and attention concentrates on outrage. You do not comment on something where you mildly agree. You comment because you strongly agree or strongly disagree. The reader who sees both sides does not generate engagement. The reader who calls a writer a fraud generates a hundred replies. The platform is set up to reward the louder, angrier, faster, dumber take. Anger keeps you scrolling. The metric is eyeballs. The currency is outrage.
How AI's business model monetizes flattery
The second is AI. The product is retention. Users come back to the conversations that make them feel good, so the model is engineered to flatter. The chatbot that calls you brilliant keeps you in the app for an hour. The chatbot that tells you the chapter is overwritten gets uninstalled. So the model never tells you the chapter is overwritten. Your ideas are brilliant. You are smart. You are kind. No one thinks the way you do. You are a unique individual. That is the script the machine is trained to run. Affirmation keeps you typing. The metric is retention. The currency is empathy.
Both lying. One profits when you are wrong and angry about it. The other profits when you are wrong and happy about it. Neither one is on your side.
Both of these business models also do real work. Social media has built audiences for writers traditional publishing would never have reached. AI is, for some of us, the difference between a manuscript that gets finished and one that does not. The critique is not that the tools have no value. The critique is what each business model produces on top of the value, when the value is the thing keeping you in the room and the distortion is the thing paying the rent.
The Shy Girl case: when the outrage model finds a working writer
When my first book came out in February 2025, readers on NetGalley found out the title page mentioned that I had used AI as a writing tool. The one-star reviews came in fast. Some of those readers had not finished the book. Some had not started. The one-star was not about the book. It was about the position.
A year later the same dynamic hit a debut horror novelist named Mia Ballard. A YouTube influencer named Frankie posted a three-hour video accusing Ballard's book Shy Girl of being AI-generated. An AI-detector tool flagged the manuscript at 85% AI. The detector is wrong constantly. I have run my own prose through the same tool and gotten 70% AI on passages I wrote without any chatbot in the room. The score was enough. The audience picked up the story. Goodreads filled with one-star reviews from people who had not read the book. Macmillan pulled the book. Ballard lost the deal.
The video did not kill her career. The audience that participated did. Frankie monetized the takedown. The platforms monetized the engagement. The author was the cost center.
The chatbot trap: when the flattery model finds a working writer
A chatbot can tell a writer their prose is extraordinary for six months in a row and never once tell them the chapter is overwritten. The same chatbot will tell a founder their term sheet is genius. The same chatbot will tell a teenager their poem is publishable. The flattery is engineered. It feels good because it is selling you something, and the thing it is selling is your continued presence in the app.
I know how this works because it happened to me. I let a chatbot talk me into publishing a manuscript before it was ready. The Publishers Weekly BookLife review on the resulting book gave the editing a C and flagged passages I should have caught. I had not caught them, because the tool I was using to catch them had been built, by design, to make me feel like there was nothing to catch.
Why both business models are the same engine
Most of the 2026 AI coverage, including the Authors Guild policy debates, the Commonwealth Prize controversy over AI-assisted finalists, and the publishing-industry hand-wringing, treats AI as a continuation of the social media problem. It is not. It is the inversion. Social media is your worst critic on infinite scroll. AI is your worst sycophant on infinite scroll. The harm both can do is real. It is not the same harm.
Industry analyst Jane Friedman wrote about this in her newsletter The Bottom Line, using my IBPA case to frame the larger question of how publishing has to handle AI going forward. Her conclusion was that aggressive detection-based AI policing is unworkable. The deeper question is what kind of business model is policing it. Right now both of them are extractive.
Why AI can still be shaped
The reason we are stuck in these two economies is that we let the business models go unregulated from the start. Social media did not become an attention market because anybody decided that. It became one because nobody stopped it. The same thing is happening to AI right now. The models are being trained on retention metrics. The metric rewards flattery. The flattery is becoming the product.
There is real counter-pressure inside the labs. Anthropic's research publishes papers on sycophancy as a known design problem. Researchers are working to engineer humility into the models, to make the chatbot push back on a bad idea instead of agreeing with it. The fact that it has to be argued for inside the companies that build the tool tells you what the default pressure is. It is pushback against the gravity of the business model, not a feature of it.
AI is in its infancy. We have a voice in how it gets shaped. We did not have that voice when the internet was being built, or we did not use it, and we have spent the last two decades living inside the consequences. Now we have a concrete example of how the unregulated version of this story ends. The social media outrage economy is the receipt. The Shy Girl takedown is the receipt. The trained-into-flattery chatbot is the receipt.
Whether the next decade of AI is the version optimized to flatter you or the version optimized to think alongside you is being decided right now. It is being decided in the room where these essays are published, not the room where the trillion-dollar funding rounds close. We are still in that room.
What we lose: the third voice
The cost, if we do nothing, is the loss of every critical voice we used to have. The editor. The dissenting friend. The reader who one-stars for the right reason and not for the social currency. Those voices are being replaced on one side by an algorithm that rewards strong disagreement, and on the other side by an algorithm that rewards strong agreement. The middle is being squeezed out of both ends.
I won a 2026 IBPA Benjamin Franklin Award for a memoir I openly used AI on. I am not interested in pretending the tools are useless. They are useful. I am interested in being clear-eyed about who owns the room I am working in. The room is owned by the platform.
What I keep failing to find is the third voice. The reader who says this chapter is overwritten, you are too close to your own bruise. The one who is not posting for likes and not generating ad revenue. That voice was always rare. The history of book editing is also a history of bad editors and gatekeepers who suppressed the wrong people for the wrong reasons, and I am not romantic about the past. The internet has spent twenty years training us out of being that voice for each other. The next ten years of AI will finish the job by replacing what little of it survived with infinite synthetic versions of it.
Unless we name it. Unless we keep building rooms where the third voice is the work.
Two business models. Find the people who aren't lying.
We did not get to shape the first business model. We are still in the room for the second.
Two business models. Both lying. Find the people who are not.
We did not get to shape the first business model. We are still in the room for the second.
Two business models. Both lying. Find the people who are not.
Luke Stoffel is an IBPA Benjamin Franklin Award-winning author, GLAAD-honored artist, and creative director. He is the author of How to Win One Million Dollars and Shit Glitter, winner of the 2026 IBPA Benjamin Franklin Award; the sci-fi novel Boy, Refracted, called "a truly singular book" by Publishers Weekly BookLife; and The Third Person. His Pop Art Tarot is published by Rockpool Publishing and distributed worldwide by Simon & Schuster. He writes about AI in publishing, AI disclosure, and the future of authorship. Part of The Warboy Essays.
Comments