The Invisible Debt: AI Is Built on Media. Who Gets Paid?

Every time an AI platform answers a question from someone asking about where to stay, what to book, or how to buy, it pulls the answer from some form of online content.

This could be a review site like Tripadvisor, a blog post, or a carefully written article – each representing journalism that took time to research and edit. The editorial authority that makes the answer credible, aside from brand sites, reviews, and OTAs, is the content pulled from media publications. But that publication almost certainly did not get what it created the content for: traffic to its site.It is an AI media attribution problem that nobody has solved as yet.

This is not the training data debate, which has been in legal discussions between media companies and AI businesses for years. This is something more immediate. What happens every single day, across millions of queries, when AI cites a publication’s work to back up its answer, and the publication gets no traffic, no revenue, and no signal it was even used?

Maria at Make Lemonade recently spoke to editors and content directors at some of the UK’s most respected luxury travel titles. The same question was put to each of them: do you know whether AI platforms are citing your work?

Most didn’t know. One described it as finding out someone had been using photographs without credit. A larger-scale taking of content seen as licensed and owned, which needs a much bigger discussion.

What AI Citation Actually Does to a Publication’s Value

If you ask ChatGPT which hotel to book in Venice, the response is not a list of articles to read. It is a combined generated answer with a recommendation, shaped around traveller type and experience, sometimes with a source listed at the end. That source is generally a publication and its editorial credibility is what makes the recommendation trustworthy. This shift is already collapsing the traditional consumer research phase for high-value travel decisions.

AI bot activity increased 225% across the web during 2025. TollBit research found scrapers retrieving full versions of paywalled articles without permission. The plagiarism fear among publishers is real, and the lack of clear attribution equally so. ChatGPT drives less than 0.2% of referral traffic back to the publishers it draws from, according to Raptive data across 6,000 independent titles. The more AI platforms take, the less they return.

One publishing executive put it plainly in a recent Digiday piece: “The irony is all these AI companies are paying for the content and data. They’re just not paying the publishers.”

This is not just about a lost publishing partnership fee. It is about someone profiting from work they did not do, invisibly, at a scale the original creator cannot track.

The Attribution Signal Your Brand Might Be Missing

There is a related problem that affects brands directly. And it is about what AI can and cannot see.

Brands invest in editorial coverage partly because association with trusted publications builds credibility over time. That logic used to be straightforward. But AI does not read mastheads. It reads structural signals, topical signals, and authority signals that have nothing to do with editorial stature.

The research Make Lemonade and Spotlight Communications are working on for the upcoming whitepaper found a striking example of this. A supplement produced by one of Britain’s most trusted newspapers, same editorial team, same standards, scored significantly lower in AI citation authority than the parent title. Not because the journalism was different. Because the attribution signals that AI reads were not there in the same way. The content existed and the credibility existed, but the AI simply could not connect them.

For brands, this has a direct commercial implication. Coverage in a publication that AI does not recognise as authoritative for a specific query does not contribute to AI discoverability, regardless of whether the title has a strong reputation or high circulation. The brand invested in media exposure, but that did not translate into AI visibility.

AI does not read mastheads. It reads signals. And right now, most brands and their PR agencies have no visibility into what those signals are.

Should AI Platforms Pay Publications? It Depends.

The instinct is to say yes, and in some cases that is clearly right.

If a platform reproduces a publication’s content so thoroughly that the user never needs to visit the source, that is not a citation, it is substituting the content from the article. This is exactly the plagiarism fear coming from publishers, and compensation is a reasonable expectation. The New York Times and Ziff Davis, among others, have taken this argument to court specifically because their content was being reproduced in a way that displaced the original.

But citation is not always substitution. A journalist referencing three sources in an article does not pay them. The reference builds credibility, sometimes sends traffic, and is part of how knowledge circulates. At its most benign, AI citation works the same way. The question is whether AI platforms are building commercial products on the back of editorial authority without returning any value, or functioning more like a very efficient index.

Right now it is somewhere between the two, and the industry has not drawn the line yet.

The question is not whether AI uses media. It clearly does. The question is whether that use is citation or substitution, and who gets to decide.

The AI Media Attribution Gap Nobody Has Priced Yet

It gets more interesting when advertising enters the picture.

Perplexity is already selling sponsored placements inside AI responses. OpenAI has quietly launched its ads manager. Their pitch to brands is simple: pay, and your name appears at or near the top of the answer. Which sounds straightforward until you look at where the answer gets its credibility.

A sponsored hotel recommendation in an AI response is not trusted because the AI said so. It is trusted because the AI has absorbed years of editorial opinion from publications that travellers trust. The brand paid for access to that trust. The platform collected the revenue. The publications whose journalism built the trust are outside the transaction entirely.

Who does what	Who gets paid
Publication builds editorial authority over years	The publication is not paid for any AI citation it is used in
AI platform trains on and cites that content	The AI platform benefits by collecting ad revenue from brands
Brand pays for sponsored placement in AI responses	The sponsored content is a recommendation backed by editorial credibility
User reads AI recommendation and books	Journey ends in the AI interface

The entity that created the underlying value is the only one not in the transaction. This is the attribution gap publishers need to be talking about more, and looking at new commercial models to benefit from.

What Attribution Could Look Like

There is no settled model. But the questions are worth stating clearly, because the window for shaping them is shorter than most people in media or PR realise.

Citation-based licensing. AI platforms pay publications a fee tied to how often their content is cited in live responses. Different from training data deals. Closer to syndication, triggered by use rather than access.

Revenue share on sponsored answers. When a brand pays for a recommendation that draws on a specific publication’s authority, a portion of that spend returns to the source. The publication’s credibility is part of what the brand paid for.

Citation-based media investment. Brands move upstream from individual placements to content strategies that build citation authority within publications AI trusts. The investment shifts from single articles to the structural signals that make AI recommendation more likely over time. The PR agency that knows which titles AI actually cites, not just which ones have the biggest reach, becomes significantly more valuable in this model.

The PR agency that knows which publications AI trusts, and why, is working with information the rest of the market does not have yet.

Where This Sits Right Now

The media industry is still largely having the training data conversation. The citation question is newer, less defined legally, and AI platforms have no particular incentive to move quickly on it. Most publications are still figuring out how to measure their own citation authority, let alone negotiate around it.

Some have resorted to blocking AI bots entirely, however the data suggests it makes little difference. Around 75% of sites that block AI crawlers still appear in AI citations, which tells its own story about how much control publishers actually have right now.

The programmatic advertising market took a decade to consolidate into structures that were largely unfavourable to publishers. The conditions for AI attribution are forming now. The publications, agencies, and brands that understand what is actually happening at a data level are the ones best placed to shape what those structures look like. An AI Visibility Snapshot is a practical starting point for any brand that wants to understand where it currently stands.

That is what the whitepaper research with Spotlight has been building towards. Stay tuned for the full findings, releasing in June.

Make Lemonade is a boutique AI search optimisation consultancy specialising in luxury hospitality and travel brands. This article is the first in a series exploring AI citation, media attribution, and the emerging citation economy. The Signal Noir Whitepaper 2, Invisible or Influential: How Leading Travel and Luxury Publications Are Adapting to AI, produced in partnership with Spotlight Communications, publishes June 2026.

Frequently Asked Questions

What is the AI media attribution problem?

When AI platforms cite media content in their responses, the publication that created that content receives no traffic, no revenue, and no measurable credit. The value of their editorial authority is being used commercially, without any of that value returning to the source.

Should AI platforms pay publications for citing them?

It depends on the nature of the citation. Where AI substitutes for the original content so thoroughly that the user never needs to visit the source, there is a strong case for compensation. Where AI is functioning more like a reference, the model is closer to how media has always cited other sources. The line between the two is what the industry needs to work out.

What is citation-based media investment?

An emerging approach where brands invest in content strategies designed to increase their AI citation authority within trusted publications, rather than buying individual placements. The investment moves upstream from single articles to the relationships and structural signals that make AI recommendation more likely over time.

How does AI advertising affect media attribution?

When brands pay AI platforms for sponsored recommendations, those recommendations draw credibility from the media sources the AI cites. The publication whose content makes the recommendation trustworthy receives no share of that advertising spend. That is the core of the attribution gap in the emerging AI advertising economy.

What is AI citation authority?

AI citation authority is the measurable likelihood that a publication will be referenced by an AI platform such as ChatGPT, Claude, Perplexity, or Gemini when answering queries relevant to a specific sector or topic. It is distinct from domain authority, circulation, and organic search rankings, and does not appear to be predicted by any of them.