Publishers cannot wait for fairness from companies that have no structural reason to provide it. We are here to be the technical partner on the publisher’s side: to help publishers understand how AI systems see their work, how authority is being captured or lost, how attribution can be preserved, and how the economics of quality information can survive in an AI-mediated world.
AI is quickly becoming the place where people decide what to buy, where to eat, what to book, which expert to trust, and which product or service is “best.” That shift is much bigger than a new search interface. It changes the structure of decision-making itself.
Search gave us a list. Social media gave us a feed. Marketplaces gave us filters, badges, and buy buttons. AI gives us something more powerful: a recommendation that feels like advice.
That is why neutrality matters.
When an AI system tells me which flight to book, which hotel to stay in, which doctor to call, which product to buy, or which restaurant has a table tonight, I am not experiencing that answer the same way I experience a banner ad. I am not even experiencing it the same way I experience a Google result. I am experiencing it as an intelligent assistant making sense of a complex world on my behalf.
That creates an enormous obligation.
The next great battle in AI will not only be about who has the best model. It will be about who can be trusted to maintain a neutral layer between user intent and commercial incentive.
Convenience Is Not Neutrality
The problem is not commerce. AI should help people transact. It should help us compare products, book tables, buy tickets, schedule appointments, and complete annoying tasks faster.
The problem begins when convenience becomes invisible bias.
We already know what this looks like. Amazon Prime trained millions of people to shop inside the universe of what could arrive quickly, easily, and with less friction. For many purchases, this works beautifully. The selection is broad, the delivery is fast, the interface is familiar, and the consumer feels served.
But over time, a convenience filter becomes a quality filter in the user’s mind. “Prime eligible” starts to feel like “best.” “Available here” starts to feel like “available.” “Easy to buy” starts to feel like “right to buy.”
That is the danger for AI.
If AI systems start recommending what is easiest to transact through their own commercial pipes, rather than what is actually best for the user, we will have recreated the worst parts of platform commerce inside a tool that people trust much more deeply than a shopping site.
The Difference Between Discovery and Transaction
There is a meaningful difference between an AI system saying, “Here are the best options,” and an AI system saying, “Here are the options I can book.”
Both are useful. But they are not the same.
A neutral AI layer should be able to say: “This is the best option based on your stated needs.”
And separately: “This is the option I can complete for you inside this interface.”
Those two answers must not be collapsed into one.
If the best restaurant is not integrated with the AI’s reservation partner, the AI should still be able to recommend it. If the cheapest flight is not available through the AI’s preferred booking rail, the AI should still show it. If the highest-quality product is sold by a merchant that does not support instant checkout, the AI should not bury it beneath a worse product that does.
The user needs to know the difference between the best answer and the most convenient transaction.
Product Quality Depends on Trust
Compromising neutrality may look attractive in the short term. There is obvious money in affiliate fees, sponsored recommendations, merchant commissions, booking fees, and closed-loop attribution. Every platform wants to move closer to the transaction because that is where the money is.
But for AI, this is more dangerous than it was for search or social.
AI products depend on trust at the interface level. If users start to suspect that the assistant is steering them toward the partner, the sponsor, the integrated merchant, or the highest-paying booking rail, the product gets worse. Not just morally worse. Functionally worse.
The core product promise of AI is that it helps the user think, choose, and act better. If commercial incentives distort the answer, then the intelligence layer itself becomes polluted. The assistant becomes less useful precisely at the moment it becomes more powerful. That is a bad future.
It is bad for consumers, because they receive narrower and more compromised choices. It is bad for independent businesses, because visibility becomes dependent on integration deals and platform economics rather than merit. It is bad for AI companies, because trust is much harder to rebuild than revenue is to capture. Publishers know this more than anyone. And it is bad for the internet, because it pushes us toward a world where the answer to every question is shaped by who has access to the transaction layer.
But the companies worth trusting are the ones that understand the bigger game: in AI, trust is not a brand value layered on top of the product. Trust is the product
Of course, companies will do what they believe is in their economic interest. AI companies, marketplaces, booking platforms, payment providers, and merchants will all be tempted to capture more of the transaction. That is not surprising, and it is not even inherently wrong. But the companies worth trusting are the ones that understand the bigger game: in AI, trust is not a brand value layered on top of the product. Trust is the product. A system that trades neutrality for short-term transaction revenue may win a few more conversions, but it will damage the very reason users came to it in the first place. The most valuable AI systems will be the ones disciplined enough to evaluate true information and make suggestions based on what is best for the user, not based on what is “easiest” for the AI.
This is why we built Intelligent Attribution. Not because we are optimistic that the platforms will eventually do the right thing, but because we are not. We do not assume that AI companies, search engines, marketplaces, or technology platforms will take the long view. We do not assume they will choose quality information over convenience, revenue, or control. We do not assume they will build a future in which journalism wins simply because journalism is valuable to democracy, culture, and public life.
Not because we are optimistic that the platforms will eventually do the right thing, but because we are not.
That may sound pessimistic. It is. But it is also earned. Every generation of technology companies has promised to organize, democratize, connect, or empower the world. And again and again, the incentives have bent toward capture: capture of attention, capture of distribution, capture of commerce, capture of trust. The people building these systems often do not think like the writers, editors, reporters, and publishers whose work they ingest, summarize, rank, and monetize. They do not carry the same obligations. They do not pay the same costs. They do not understand, or do not sufficiently value, what it takes to produce information that is actually worth trusting.
People often ask us whether the AI companies will eventually solve this problem themselves. If they do, we have no reason to assume they will solve it in the publisher’s best interest. They will solve it when it becomes strategically or economically useful to them, on terms shaped by their own incentives. By that point, publishers may have lost value and power. We are not interested in waiting until quality journalism has been extracted, summarized, and devalued before publishers are offered a fraction of what remains.
Investigative reporting, long-form journalism, local coverage, original criticism, and serious editorial work are expensive. They are slow. They are inconvenient. They require independence, institutional memory, fact-checking, legal risk, human judgment, and moral courage. These are not the values most platform systems naturally reward. Platforms reward what can be scaled, extracted, compressed, ranked, and monetized.
Intelligent Attribution exists because publishers cannot wait for fairness from companies that have no structural reason to provide it. We are here to be the technical partner on the publisher’s side: to help publishers understand how AI systems see their work, how authority is being captured or lost, how attribution can be preserved, and how the economics of quality information can survive in an AI-mediated world.
To Learn More about AI Neutrality and Trust See The Mozilla Foundation
