Inside OpenAI’s and Perplexity’s shady schemes
You pay for AI, but you’re still the product. Here's what they're not telling you.
The Financial Times reported today that OpenAI plans to take a cut of any future online shopping done through ChatGPT.
This change is part of a larger push at OpenAI to find new revenue streams beyond subscriptions, as the company sharpens its focus on commerce inside its AI products. The move reverses prior commitments to not make money this way.
OpenAI rolled out shopping features in ChatGPT for all users — free, Pro, and Plus accounts — in April. This tool lets people search for products, compare prices, see photos, read reviews, and click to buy, all within the chat window. The checkout happens on the retailer’s website, not inside ChatGPT, but OpenAI’s system spots when someone wants to shop and presents carousels of product picks with links to buy.
These affiliate-revenue micropayments represent a spigot for OpenAI. In the future, if they need money (and they will), they have a strong incentive to tailor ChatGPT results — links and content — to lead people toward purchases.
OpenAI is working directly with companies and media brands to shape how content — like shopping recommendations and product offers — appears in ChatGPT. Over the past year, OpenAI has signed over 30 partnerships with major tech and publishing companies for access to data, distribution with favorable placement, and revenue sharing from linked sales.
In other words, despite what ChatGPT users likely believe, results are biased by paid, behind-the-scene deals.
Savvy users know that if they pay for a software product or service where subscription is the sole monetization, they are the customer and the product is the product. But if the product is “free,” the user is the product and the customer is the advertiser, affiliate, or partner.
In that latter scenario, the user is the product, but at least users don’t have to pay for it.
OpenAI charges subscriptions and now also wants affiliate marketing revenue.
Perplexity AI is even worse.
AI chatbot users know Perplexity sells “Pro” subscriptions and shows ads to “free” users. That means paid users are the “customer” and “free” users are the “product,” right?
Actually, no — they’re both the product.
Few Perplexity users know a third type of Perplexity monetization is selling prioritized links and content to “partners” in a revenue-sharing arrangement, similar to what OpenAI does.
Perplexity maintains formal agreements with publishers and other companies through their Select Channel Partner and Publishers’ Programs. Media companies in these programs let Perplexity use and cite their content for a share of revenue, access to analytics, and free access to Perplexity’s paid Enterprise Pro service for a year. These programs have been running since early 2024, and the partners include The Los Angeles Times, Adweek, The Independent, Time, Fortune, and Der Spiegel.
When Perplexity answers a user’s question using content from a partner, the contract requires Perplexity to prominently display the partner’s name and sometimes their logo. The paid branding is presented as objective search results.
Partner publishers are often placed at the top or in a visible spot in the answer, making users more likely to notice them and click through to the partner’s original story, driving more direct traffic to them.
Perplexity is selling priority placement and user traffic. There’s built-in bias. Users assume every result and citation in a Perplexity answer is due to relevance or quality, but in partner cases, money and contracts shape which brands and content appear, how, and where. Both sides benefit when users trust the answer and click on branded links.
Few users know the search or prompt results are from behind-the-scenes monetization deals.
OpenAI is triple-dipping with monetizing through paid subscriptions, paid prioritization and now affiliate marketing. Perplexity is also triple-dipping, using paid subscriptions, advertising, and paid prioritization.
If Google did this, everyone would freak out. If nobody freaks out about the AI chatbot companies doing it, maybe Google will join in.
The AI companies dominating the world right now are deep in debt and have been spending money like drunken sailors. Now they want you to pay their bills by tricking you into believing that the information service you’re paying for is unbiased.
If AI chatbots clearly labeled which parts of their results were unbiased, and which were biased through monetization schemes and back-room deals, that would be fine.
What’s actually happening is that they’re presenting paid prioritization as unbiased results. And that’s shady.
Exclusive Machine Society Interview: Dr. Andrew Rogoyski
Dr. Andrew Rogoyski is a prominent UK-based business leader, strategist, and technologist with over 30 years of experience across industry, government, and academia. He currently serves as the Director of Innovation and Partnerships at the Surrey Institute for People-Centred Artificial Intelligence at the University of Surrey.
We talked about AI innovation, partnerships, and the societal impact of artificial intelligence, with a particular focus on how people-centered approaches are shaping the future of AI.
(This interview is available exclusively in this issue of Machine Society.)
Superintelligent: ‘AI Uses Your Browser (So You Don’t Have To!)
Become a Superintelligent listener by subscribing to our newly renamed Superintelligent podcast! Search your podcast app for "Superintelligent" or find us on YouTube, Apple, or Spotify.
More From Elgan Media, Inc.
Send in the clones
The one secret to using genAI to boost your brain
Is Microsoft’s new Mu for you?
Where’s Mike? Mexico City!
(Why I’m always traveling.)
This has got to be the fastest move to enshi**ification yet. Question, though: couldn't users simply train or tell the model to give them and "unbiased" version?
Create your own "Shopping" project in Perplexity, and whenever you go shopping, simply tell Perplexity how you want your results shown. Solved?