Why micropayments can't save news

In an interview with Nicholas Thompson last month, OpenAI CEO Sam Altman backed a micropayments strategy for news:

“What really makes sense in a world of agents is we try a sort of micropayment-based approach. […] My agent can read it, pay $0.17, and give me a summary of that. If I want to go read the whole article, pay $1, or however that works. If my agent wants to calculate something for me that’s really difficult to do, it can go rent some cloud compute somewhere and pay for that, but I think there will be need to be a new economic model for these agents doing lots of small transactions and exchanges of value with each other on behalf of their human controllers or whatever, all of the time.”

What he describes is a combination of pay-per-view and a utility model. In pay-per-view, you’re paying a custom price set by the publisher in order to access a resource; you as the user might get a prompt asking whether you approve the transaction, or you might just give it an approved budget. When he talks about an agent renting cloud compute, it would likely be paying based on usage like a utility, with financial transactions similarly carried out behind the scenes. It’s not really different from the Spotify model in the sense that artists are paid per stream; there it’s a percentage of your total listens, whereas here it’s probably a total budget that you preload into a wallet.

This keeps coming up in software circles: I’ve heard, again and again, that an approach like Altman describes would help provide revenue for publishers in an AI-intermediated information ecosystem. I disagree; while I think there’s certainly value for utility-style pricing on the web, proposals to use it for news are based on a misunderstanding of how journalism is valued and paid for today, set up the wrong incentives for publishers, and conflate every kind of publisher into one very flat model.

So, let’s talk about it.

First, it’s important to understand what a micropayment actually is. Then I’ll discuss the incentives micropayments create, how they might apply to different kinds of publishers, where micropayments might be useful, and how platform owners might embrace the needs of publishers more directly in an agentic ecosystem.

What is a micropayment?

Micropayments are small payments — sometimes a fraction of a cent — that are charged to access digital goods or services. Usually, this happens more or less automatically: you load a wallet connected to your web browser or AI agent, and then when you or your agent visit a resource, a small payment is made. This could be a flat fee or it could be charged as a percentage of your total browsing for the month.

There have been a number of attempts to make micropayments work on the web. Flattr was an early example: a browser extension that paid for content on your behalf. It struggled with getting enough people to fund their wallets and was faced with the high underlying transaction fees associated with credit card payments. It ultimately shut down in 2023. The Brave browser attempts something similar through cryptocurrency.

Today, the Interledger Foundation is working on creating an open protocol that can be used for micropayments, among other kinds of compensation. The protocol is intended to overcome the kinds of financial friction that Flattr experienced. The Foundation is working hard on the problems I’ll raise here in the rest of the piece.

We’ll make it up in volume

Clickbait is enabled at scale on the web by display advertising. Because the revenue received by a publisher is directly connected to the number of pageviews they receive, publishers that are entirely tethered to this model have two incentives:

1. To drive as much traffic as possible to their articles
2. To lower the cost of each individual article as much as possible

That’s led to an information ecosystem where many publishers produce low-quality content with misleading headlines in an effort to get as many people as possible to look at them. For many, it doesn’t even matter if the article is true; the reader doesn’t build a long-term trust relationship with the publisher and likely won’t come back unless they’ve been hoodwinked into looking at another one. These publishers are the tourist traps of the internet. There’s no ongoing relationship, so there’s no duty of care.

Micropayments are effectively this model, with the difference that nobody has to look at an ad. Revenue is directly tethered to traffic, like a display ad model; the difference is that the money for a publisher comes from the reader’s pocket instead of the advertiser’s.

Not only does this continue to incentivize clickbait, but these publishers are now competing for the checkbooks of millions of people rather than a few well-funded advertisers. In an advertising-based world, every visitor is highly likely (ad blockers excepted) to produce a small amount of revenue; in a distributed wallet world, we’re relying on individuals to be well-funded.

It’s also worth considering the full user journey. For a reader to arrive at a particular article, one of two things is happening:

1. The reader has discovered the article off-platform, perhaps through a social media post or a search engine result
2. The reader already trusts the publication and is seeking out their information specifically

In the first case, the publisher is incentivized to find ways to surface their work. That will be through a combination of tried and tested methods like social media audience work, paid acquisition, and SEO. But they have to do this work for every single article, for every single reader. The most efficient thing they can do is try to build an ongoing relationship with that reader so they don’t have to work as hard for that reader’s second pageview. In other words, the most desirable end state for a cold interaction with a reader is some kind of trusted relationship: they want the reader to sign up for a membership.

In the second case, a trust relationship has already been established. Here, too, the most desirable end state is for the reader to have a membership. Because the reader already trusts the publication, it’s the most desirable end state for them, too: they likely want to prioritize this publication over other sources that are less trusted by them.

Okay, but what about agents?

There’s an argument to be made that, in an agentic world, the relationship isn’t between the publication and the reader; instead, the AI vendor sits in the middle as an intermediary. That’s much worse. Rather than allowing every reader to build their own information landscape, the decisions made by companies like OpenAI about which sources are trustworthy will affect the information that everybody receives. If agents build their own trust relationships with publications, the underlying assumptions that dictate how those agents select sources and process their information govern how everybody learns about the world around them.

Micropayments probably are the financial backbone of that world, but I’d argue that it’s not a world we want to live in. On the other hand, if agents act based on human reader preferences, then the relationship returns to being between the publication and the reader, and we’re back to the incentives for both readers and publishers tending towards membership.

If an agent-first world really is coming — and, to be clear, the jury’s still out — then finding ways to encode memberships into the underlying protocols and mechanisms is important. Micropayments are ephemeral.

Publishers are not monolithic

When we talk about “publishers” we miss a ton of nuance. What works for a local for-profit paper is not the same as what works for a non-profit national news website is not the same as what works for a premium international newspaper. These businesses all try to serve their communities with information, but they have different financial dynamics, communities, and needs. There are multiple overlapping spectra of business models, information types, communities served, publishing surfaces, and more.

At some intersections, micropayments might make more sense than others. For example, if a local for-profit paper is serving event listings for the area, it might quite reasonably charge an agent to access it as a dataset on a utility basis. In general, micropayments do make sense as a way to access raw datasets or lists of facts; you don’t establish a trust relationship with a spreadsheet.

But even then, it’s heavily dependent on the nature of the publisher. If it’s a non-profit paper, those charges might not fit within its mission. If we’re instead talking about ongoing, in-depth qualitative reporting on the characters behind a neighborhood, rather than flat facts, or an ongoing series by a journalist with a specific point of view, a reader might want to establish a subscriber or member relationship. But that, in turn, might make more sense for a for-profit publisher, while a non-profit might be serving its articles for free but asking for a donation. (Or it might not! There’s nothing requiring that it does this.)

The medium matters, too: some models work better with the ways people interact with text, others with the ways people watch video. You also need to consider the characteristics of the intended audience. Can they pay for a micropayment or a membership, or does informing them require some other kind of underwriting like non-profit donations from an institutional backer or, yes, advertising? If we only inform audiences who can pay, we create news deserts in the communities that might need real reporting the most.

There’s no single solution. Every publisher needs to figure out the business model that’s right for it based on its specific context, mission, audience, and focus.

There’s a real distinction between commodity facts (a stock price, a sports score, an event listing) and the analytical, contextual, investigative work that makes journalism valuable. The latter is an ongoing process of building trust. It’s not ephemeral; it is relationship-based. And although, as I’ve discussed, there are also real differences across publisher type, medium, and community, it’s generally not a good fit for micropayments.

So what should platforms do?

I’ve discussed that publishers need to navigate all their contextual factors and choose the business model that’s right for them from first principles. Platform owners would likely prefer that publishers all use the same template — it’s much easier to build a scalable platform that way. Sam Altman isn’t just thinking about how to provide compensation to publishers; he’s also considering what would result in the least friction for his business.

Micropayments are useful for people who build agentic platforms because they reduce friction. The user asks for information (or the system contextually decides they need to see it); the system serves up that information. The micropayment transaction happens behind the scenes in a non-interruptive way.

Supporting trust relationships can be similarly non-interruptive, but they require a little more thought. Today, the thinking is that AI agents will use protocols like MCP to request information from sources. The protocol might change in the future, but the principle probably remains the same: it’s an API that happens to be designed for AI systems rather than traditional client applications.

Adding a mechanism for explicit calls to action would be a great start. This would allow the publisher to prompt the user to start a membership. The publisher would receive more information about the reader, as well as potentially some revenue; in turn, the reader would receive access to more articles or data that their agent could use.

One embodiment of this is that a user could subscribe to a publisher and receive its articles in a reader, perhaps even using an existing substrate like RSS; they could then use their accumulated corpus of subscribed articles with their AI agent. The agent is not intermediating a relationship with the publisher; the relationship is with the human subscriber, and the agent then operates on what they’ve subscribed to. This doesn’t just build trust in the publisher: it builds a loyalty relationship with the AI platform too. Both the publisher and the AI platform build a stronger relationship with the reader.

Membership-first platforms like Ghost are already approaching this as they start to add more affordances for AI. It wouldn’t surprise me to see existing publisher platforms move in this direction, building more foundational AI technology for memberships in the process.

Publishers don’t need to compromise on their relationships with their communities in order to adopt new technology. Platforms don’t need to flatten relationships in order to achieve scale. There’s a lot to be gained by working together and understanding each other — and remembering that human relationships build the kind of trust and loyalty that an ephemeral transaction never could.