No one loves ads, but generally people will tolerate them if it means getting something free or at a lower price. What people won’t forgive? Being lied to.
A study looking at how consumers see native ads on news sites—which look similar to editorial content, just with labels—found that fewer than 10% of consumers could identify them as advertising. And when people find out after the fact, they don’t take it well. In a Contently survey, two-thirds of readers said they felt deceived when they realized an article or video was sponsored by a brand.
That’s why there’s a whole set of regulations and standards around advertising disclosures that are meant to protect consumers. Now that AI is slowly becoming a central part of our media ecosystem, it’s unclear to what extent those disclosures should carry over into AI answers. After all, if the public is increasingly using AI to shop (47% of U.S. shoppers who use AI started their last product search with AI, according to an impact.com survey, and Meta says shopping is the most common use of its new personal AI agent, Muse), it feels intuitive that the user should see plainly the commercial interests of the sites that informed any answer.
Truth in labeling?
A reliable label in AI sounds like a good idea, but it gets complicated quickly. For starters, the label might cause confusion that the AI platform itself was paid rather than the original publisher. In the case of ChatGPT that would potentially conflict with its own ads, which are separate from the answer and labeled as sponsored. OpenAI’s answer to the disclosure question, so far, is the citation link: since launching ChatGPT search, it has included links to sources in answers so users can check them for themselves.
All of this gets at the fundamental question of judgment, and which parts of judging a source users have handed over to AI, consciously or unconsciously, simply by using it. When you ask AI to come up with, say, recommendations for air fryers, it’s more like hiring a researcher, not looking at a billboard. In fact, some people turn to AI precisely to get away from the billboards: 24% of the shoppers in the impact.com survey said they use AI to avoid sponsored results and ads. You trust it to hoover up all the information it can find about the question, weigh all the factors—including whether or not any sources are sponsored—and give you the best possible answer.
How important is a label at that point? It’d be silly to ask a researcher to stamp every brochure they look at, and pass those brochures over to you in a pile. If AI can solve some of humanity’s most intractable math problems, it can probably reliably weigh the commercial intent of a brand’s FAQ page. In fact, you could argue a label in the answer might cause the reader to overscrutinize an answer the AI has already deemed the most useful possible.
AI’s affiliate addiction
If we’re considering disclosures, we need to talk about affiliate content, and AI’s outsize reliance on it. As you might imagine, for any query about buying decisions, AI seeks out product recommendations. Those articles are often supported by affiliate revenue, meaning the site gets a fee when the reader takes action, typically a purchase. Legitimate sites that monetize via affiliate disclose that fact on any page with affiliate links; indeed, FTC guidelines require that they do so.
However, almost none of those disclosures survive the journey into an AI answer. AI summaries rely heavily on affiliate content: one study found that about a quarter of the pages cited by ChatGPT, Google’s AI Overviews, and Perplexity carried a commercial disclosure, and 86% of answers cited at least one such page. If those disclosures carried over, they’d be in almost every shopping answer. Beyond the UI nightmare that would result, would the label have any meaning at that point?
To be clear, no one’s arguing that disclosures don’t matter generally. Journalists take them very seriously, and they should. The SPJ code says to “distinguish news from advertising and shun hybrids that blur the lines.” Recommendation sites like Wirecutter or The Strategist make their business model clear, on every page, to everyone who cares to know.
The number of people who care to know, however, is likely rather low. When Pew tracked how people actually use Google, they clicked a link inside an AI summary on just 1% of visits. According to impact.com’s survey, Gen Z is split between those who never click through (11%) and those who always do (21%). That doesn’t contradict the anger over native ads: People hate finding out they were misled, but that’s different from wanting to audit every source up front. Again, auditing the sources is exactly the job they’re handing to the AI. While it wouldn’t solve all issues, you can picture a world where AI users have the option of seeing labels or not, perhaps via something like a “flag paid sources” toggle.
The kickback question
The label debate, however, misses the real concern. Anyone familiar with today’s models understands that AI can tell the difference between paid content and editorial (being able to tell isn’t the same as being impossible to game, but that’s another column). What is less clear are the commercial interests of the platform itself. In the researcher analogy, you might trust the judgment of the person you hired—that’s why you hired them. But you also want to be assured they’re not getting kickbacks from anyone they’re recommending. That worry isn’t hypothetical. In a University of Michigan study, people rarely noticed when a chatbot wove ads into its responses, and many actually preferred those answers. Even when the ads were labeled, about half the participants didn’t spot them.
This concern is at the heart of OpenAI’s promise that “ads do not influence the answers ChatGPT gives you,” which is directly from its ad principles. Ads only appear in ChatGPT’s free and Go tiers, and they always appear separate from the answer itself. Perplexity made similar promises when advertising was part of its business model, though it walked away from ads earlier this year, with one executive saying, “A user needs to believe this is the best possible answer.”
OpenAI’s promise is important because of the trust factor, which is evident when you look at conversion rates. Conversion from OpenAI ads is considerably lower than from the answers themselves. Early research from the Frankfurt School of Finance & Management found shoppers who clicked ChatGPT ads bought at a rate of 0.37%, versus 2.21% for those who arrived through ChatGPT’s regular answers.
That makes sense—the answer gives what the model deems the “right” recommendation whereas an ad is just one of those options trying to sell you something. It also explains why OpenAI has every reason to keep that wall up. The moment answers look bought, they’ll convert like ads. Logically, this incentivizes brands to prioritize presence in AI answers over advertising. Not only are you more likely to drive business, you can also reach the people who pay for ChatGPT, who never see ads at all (and thus are more likely to pay for other things, like your product).
As I found out in my tests with Meta’s Muse, it will generally recommend what you can actually buy within Muse. When I asked it which robot vacuum was the best for me, it recommended a model that I could check out right there, but was not the top model on the review it cited. And Meta CEO Mark Zuckerberg has said directly that Muse would monetize via transactions. On Meta’s October earnings call, he said the company will “build commerce tools so we get paid a fee based on the transactions that we help facilitate.” That suggests Muse will have every reason to consider what’s best for Meta when it steers you toward product decisions.
An agent you can trust
Concerns over agent influence have sparked calls for engineering “loyalty by design” in agentic systems, a phrase from Consumer Reports’ work on AI agents. The idea: agents should always put the needs and desires of the user first, and make transparent any other considerations, or even conflicts, they might have when finding information or carrying out actions. Sen. Mark Warner has floated draft legislation that would write a similar duty of loyalty into law for some AI agents.
There’s a strong case to be made that the market will correct for this: If any particular agent doesn’t act in the user’s interest in a material way, there’s certainly plenty of competition. That assumes switching is easy, though. An agent that knows your email, your calendar and what you bought last month is a lot harder to fire than a search engine.
Ultimately the fight over commercial labels is a good instinct, but it aims the question of trust at the wrong target. By using AI, I’m implicitly trusting the agent’s judgment to find the best information to answer my query. The better question is whether the agent is loyal, and how I can check. Labels protect readers from sketchy sources. What protects us from sketchy agents hasn’t been invented yet.
