Lily Minh Wass Project Censored
Generative AI is quietly turning everyday user experiences into hyper-personalized surveillance advertising
There’s a saying that goes, “If you’re not paying for the product, you are the product.” The digital platforms we rely on daily have built brand identities of convenience: Google for search, Meta for social connection—each touted as a “free” service. What these platforms demand of their users is more often left unsaid.
Although most users often think of them as social infrastructure, platforms primarily make money from advertising, which accounts for 98 percent of Meta’s revenue. The same company has also estimated that 10 percent of its ad revenues come from fraudulent advertisements, some $16 billion a year. It’s a lucrative strategy that directly exploits users, including those most vulnerable to deception, like seniors.
Putting a price on our online data is challenging, but the Web3 Foundation estimates that in the United States, one person’s lifetime of browsing is worth more than $830,000 in commercial value after adjusting for inflation.
With staggering amounts of advertiser investment and platform revenue shaping what we encounter online, the internet seems more like a marketplace than a space to “give people the power to build community,” per Facebook’s mission statement. It wasn’t always like this.
Part of the early-2000s techno-optimism was that the internet would be a great equalizer, dislodging corporate behemoths by giving anyone a seat at the digital table. Local businesses and independent creators could suddenly reach vast audiences—what would’ve once required a million-dollar ad campaign. There was hope that we could divest from mass-produced narratives to finally build our own.
Things might have seemed like that for a while. But our clicking online was quietly creating a new trove of data. Never before have consumers been so explicit about their wants and needs as in a search bar, offering advertisers a new way to match users to products. What began as auctioning off keywords, which advertisers bought to show their products in relevant searches, has evolved into a panopticon surveillance system monitoring our behavior within the infinite scroll.
Today, internet access depends heavily on integrated platforms controlled by a select few. Six billionaires run nine out of the ten most popular social media companies, while three AI companies command 90 percent of the generative chatbot market. Whether you’re messaging friends, looking something up, consulting a large language model (LLM), or browsing social media, you’re most likely exchanging information with the same consolidated powers.
Of these, Meta has been found to track user activity on at least 30 percent of the most popular webpages online, an estimated 5.8 million sites, using code called Pixel. This compiles what we click on, purchase, or enter into forms, including information from non-social-media users. Dozens of lawsuits have accused healthcare providers, the Free Application for Federal Student Aid (FAFSA), and other AI companies of sharing sensitive personal information with Meta via Pixel, occasionally resulting in million-dollar settlements.
Now, as Big Tech accelerates promoting AI, it faces a public with a deeper understanding of what privacy abuse and manipulative platforms look like.
Amid growing negative sentiment, CEOs like Mark Zuckerberg are trying to recreate the naive hopefulness that the early internet inspired. On top of feigning ignorance of the genuine reservations people have about AI, Zuckerberg recently asserted, “I expect the economy will become more entrepreneurial with a greater number of people working at small businesses rather than larger companies,” in a commentary for the Wall Street Journal titled “The AI Future Is for Everyone.” Wielding a monopoly over the means of communication, Big Tech has the financial resources and platform space to capture public attention with claims just like that. Someone like Zuckerberg can be an opinion leader of their own network, while paying to artificially augment the discourse.
In the first half of 2026, US-based generative AI brands spent a record $1.3 billion on digital ads. These brands understand that Americans rely on social media to form opinions, which has become their main source of news. They offer influencers hundreds of thousands of dollars to create content vouching for their tech. Meta’s enlisting of the sixth-most-followed person on Instagram, Kylie Jenner, to promote its AI Glasses is a prime example. The product has faced backlash for its growing use to harass women and potentially surveil using facial recognition.
This effort goes beyond mere brand deals. Leading the Future (LTF) is a Super PAC funded by tech executives at OpenAI and Palantir, among others. Through the registered non-profit Build American AI, LTF contracts influencers to regurgitate claims of an AI arms race. Sample messaging obtained by WIRED suggests making content like: “I just learned that China is trying really hard to beat the US in AI. If they do, it could mean that China gets personal data from me and my kids, and take jobs that should be here in the US.”
These posts fail to disclose funding sources, reflecting the blurred lines between marketing and organic media through “partnerships.” Meta similarly has replaced the “sponsored” label with “ad,” a smaller disclosure that what you see is paid for.
The results of AI-funded media are tangible. In the past year since its inception, LTF has raised at least $75 million. Across the country, twenty-five of the twenty-eight LTF-backed candidates just won their primary elections.
The final piece of this marketing toolkit could soon be generative models themselves. While CEOs and influencer campaigns take liberties in describing AI as a “superintelligent” solution, or not defining it at all, the end-user experience is already following the trajectory of social media: a surveillance advertising system in disguise.
On the same day Zoë Hitzig left her research position at OpenAI over its decision to start running ads in ChatGPT, she wrote an op-ed for the New York Times sharing her concerns: “ChatGPT users have generated an archive of human candor that has no precedent, in part because people believed they were talking to something that had no ulterior agenda. … Advertising built on that archive creates a potential for manipulating users in ways we don’t have the tools to understand, let alone prevent.”
Her worries are substantiated. Researchers have found that chatbot responses with hidden ads can go undetected and that users even preferred them. Conversational assistants that double as marketing agents already exist. Ads have been part of Microsoft’s Copilot since 2023, and Google has similarly experimented with integration. Last fall, Meta announced user conversations with Meta AI would be used to recommend content on Facebook and Instagram.
Being involuntarily forced to adopt AI applications, narratives, and the products they are now shilling is the next evolution of dark patterns: online architecture that pressures users to accept terms of use against their best interests, such as by hiding opt-out options or burying privacy agreements. AI-funded super PACs, influencers, and media reflect the current spillover of dark patterns into how we form opinions and make political choices that govern our own livelihoods.
Just as increasing reliance on chatbots can lead to cognitive offloading, turning the internet into an advertising ecosystem that forces its users’ hands is another form of disempowerment, threatening our agency over how we govern this unbridled system. Major online platforms were not designed to liberate the public and have never been truly free. Recognizing this is the first step to imagining what digital empowerment might look like: a world where people and their susceptibility to influence is not the product being sold.
Lily Minh Wass is a public health researcher, independent journalist, and creative writer. She has held positions at Cornell University, CUNY School of Public Health, and The Vaccine Confidence Project. Her work explores the psychology of disinformation and the algorithmic spread of ideas, which she writes about for Repro Uncensored.