Media Briefing: The pros and cons of different AI revenue models for publishers

Pros and cons of AI revenue models

About this time last year, content licensing deals between publishers and AI tech companies like OpenAI and Microsoft were kicking off. Since then, newer entrants like ProRata.AI and Tollbits have started to catch publishers’ attention and signed new deals, in the hopes that they will bring in revenue from ads, subscriptions and content usage fees.

That’s especially important, given the growing contentious relationship between publishers and AI companies. This week, the trade association News/Media Alliance launched an ad campaign called “Support Responsible AI,” urging the U.S. government to protect publishers’ content from AI. Publishers like The New York Times, The Washington Post and The Guardian are running those ads in print and online. 

Publishers are fed up with AI companies scraping their content for free. They’re vying for a foot in the door as AI technology threatens to upend the digital ecosystem as they know it — so at least they can try to help shape the next iteration of the internet.

Digiday has compiled a list of some of the different AI revenue models that publishers are signing and the pros and cons of each.

Content licensing deals

Publishers get compensated for the content used to train large language models (LLMs), such as the ones run by Microsoft and OpenAI.

This year, OpenAI has signed The Guardian and Schibsted, while Google inked a deal with the Associated Press.

Pros:

These are the big players with the “deepest pockets,” said one publishing exec, who asked to speak anonymously. This gives publishers the opportunity to monetize their archives and content by having AI companies pay to use that content to train their models, rather than scraping and using that content for free.

Publishers are reportedly getting anywhere from $1 million to $5 million a year to license their content. News Corp’s five-year deal with OpenAI is reportedly valued at as much as $250 million (in cash and OpenAI credits).

AI tech can also boost publishers’ own tools, like Dotdash Meredith’s contextual targeting tool D/Cipher. It can also help develop internal and external products, such as the AI chat bar on Time’s site and the ability to integrate OpenAI’s chatbot ChatGPT into its AI enterprise tool Glean.

These kinds of partnerships also give publishers a seat at the table when it comes to AI platforms figuring out how to attribute content in their search engines, like OpenAI’s SearchGPT, and ChatGPT. The hope is that appropriate attribution — such as links in generated answers — can help drive traffic back to publishers’ sites when their content is referenced in answers to prompts. 

“The reason our OpenAI deal got done was not because of the economic relationship that we got out of that… It was actually because it also included… meaningful attribution, [which] is by far the bigger, longer-term economic value for a publisher,” said a second publishing exec who requested anonymity.

Cons:

Whether that attribution is actually helping publisher referral traffic is questionable.

Another issue is how much AI companies are paying publishers for their content. A couple million dollars is a nice revenue boost but how much is a publisher’s content really worth when it’s used to train an LLM?

A lack of transparency around how these deals are structured means execs are often left in the dark wondering if they are getting good deal terms.

Plus, companies like OpenAI that were once doling out million-dollar content licensing deals have also become more wishy-washy on pursuing more deals this year, according to three publishing execs that spoke with Digiday.

“We’re having a conversation [with OpenAI, and it sounds] awesome. Twelve weeks later, what the hell happened? And they’re going in a different direction. Twelve weeks after that, oh… it looks good. And then a deal never gets done and it just fizzles out,” said a publishing exec during the closed-door town hall at the Digiday Publishing Summit last month. “That’s where we’re at today. It’s like, what’s going to make us money and what’s not? Let’s go with what makes us money.”

Advertising revenue share deals

AI search engine Perplexity introduced an advertising revenue share model last July. When a user asks a question in Perplexity’s search engine and a publisher’s site is cited as a source in the response, that publisher can receive up to 25% of the revenue Perplexity makes if an ad is served.

Perplexity has signed revenue share deals with publishers like Time, Der Spiegel and the Los Angeles Times, among others.

Generative AI search startup ProRata is also touting an ad revenue share model. This month, it signed the News/Media Alliance to share revenue with its members if they opt into a content license with ProRata. 

Pros:

Publishers can monetize their content when it’s surfaced in AI engines.

Companies like Perplexity are also engaging with smaller publishers that haven’t had much luck with OpenAI. “We reached out to OpenAI, Perplexity and Anthropic,” a third anonymous publishing exec told Digiday. “Perplexity was the only one that responded affirmatively.”

Mid-sized and smaller publishers don’t necessarily have the time or resources to cut licensing deals with AI companies, and vice versa. 

“A lot of our members who are smaller, and even some of the larger ones, might not have the time to spend six to nine months negotiating with as many AI companies as there are out there, and they’re all different,” said News/Media Alliance CEO Danielle Coffey.

That exec also appreciated the flexibility of a deal with Perplexity, which allowed their company to exit if they found it to no longer be “mutually beneficial,” they said. “So a year from now if there’s tech that comes out that we want to explore, we aren’t locked in.” (Although, Perplexity can end the deal at any time too.)

One of the perks of ProRata’s offered deal is a 50/50 revenue share split, three execs told Digiday. Bill Gross, ProRata founder, said that means 50% of the company’s overall revenue (which comes from ads served on its search engine Gist and subscriptions for an ad-free search experience) is going to publishers that have joined its program.

Cons:

These players can’t yet match the commercial opportunity provided by the big dogs like OpenAI, the first exec said.

And while the revenue model may be transparent, the amount of money publishers stand to make isn’t so clear. Because the ad infrastructure is still nascent, with not much revenue coming in (or referral traffic, for that matter).

“Looking at our traffic from Perplexity, I’m not seeing enough of an opportunity that it will yet be a huge revenue driver,” said a fourth anonymous publishing exec. “But [I’m looking at it] more as the publisher program grows and more users use Perplexity, I’m hoping that will grow.”

An exec at DPS last month shared the same sentiment during the town hall session. 

“If you ask [Perplexity] about meaningful revenue there, they really can’t answer on it right now. They’re very, very quiet on analytics,” they said. “We ultimately ended conversations with Perplexity because it just felt like there wasn’t enough there yet.”

Pay-per-use model

ProRata and Tollbit revenue share models are still in their early stages, and a mix of models are being developed.

For example, ProRata — which has signed over 400 publishers to its platform — has content licensing deals with publishers like Arena Group, BuzzFeed and Recurrent Ventures, among others. It’s also developing a per-use compensation structure, similar to platforms like Spotify, Gross said. 

ProRata’s algorithm analyzes generated content outputs on its search engine Gist (which went live a few weeks ago, according to Gross) to see where that content came from, and then shares revenue with the publishers. It’s built on Meta’s Llama LLM. 

Tollbit, a data marketplace for publishers and AI companies, gives AI scrapers the option to pay a “toll” to access a publisher’s content. A web scraper or AI agent tries to go to a publisher’s webpage, gets redirected to Tollbit’s platform, and then is offered a “flat rate” transaction fee to access that page, TollBit’s co-founder and CEO Toshit Panigrahi explained to Digiday. Publishers can set their own rates, Tollbit co-founder and COO Olivia Joslin said. Panigrahi declined to share what those rates are. Tollbit has struck deals with publishers like Penske Media Corporation, Time and Trusted Media Brands, among others.

Pros:

Payments from ProRata are divvied out to publishers based on how often they’re cited in response to queries in Gist, Gross said. When a publisher’s content is cited as a source, that publisher can make up to 50% of the revenue paid out to publishers, with a maximum of 5% per citation, the second exec said. (For example, if a publisher’s site is cited seven times, they’ll make 35% of the revenue share). Perplexity’s payout to publishers maxes out at 25%, in comparison.

Tollbit, meanwhile, can serve as a “middle man” to monetize content for publishers, especially at a time when it seems like OpenAI has slowed down its content licensing deals with publishers, the fourth exec said.

Tollbit’s tool, called Content Cache, lets publishers control which AI bots get access to content and can reduce content delivery (CDN) costs, which are rising due to an increase in AI bot traffic, Panigrahi said. Tollbit doesn’t charge publishers for this tool (for now).

Cons:

Panigrahi said publishers are saving hundreds to a couple thousand dollars a month by using Content Cache to reduce their bot-related traffic costs. But in terms of Tollbit’s larger ambitions to serve as a quasi-metered paywall tool against AI scraping, this model remains mostly hypothetical.

Tollbit would not say how much revenue publishers had made from Content Cache or its bot paywall. Joslin said there is money being exchanged between AI companies and publishers on the TollBit platform, with over 5 million transactions in Q4 on TollBit.

The second publishing exec told Digiday their main concern with Tollbit is that there aren’t guardrails in place to ensure that its data marketplace doesn’t get spammed with fake content, undercutting publishers that have high quality content. Premium publishers should be able to charge more for their content, they argued. After this story was published, Tollbit’s Joslin noted that publisher rates for content can be dynamic — for example, they can change based on how old the content is.

ProRata is working on a subscription model, though none of the publishers interviewed for this story would comment on what kind of revenue they could expect from that model.

Two publishing execs told Digiday it’s simply too early to say how much money will come from their deals with ProRata. Gross said checks won’t go out to publishers until Q2.

“There’s a lot to be worked out,” the second exec said. “Nobody knows how this makes money yet. [So far], the revenue shares are totally fictional.”

This story has been updated to reflect how the Content Cache tool works, and a comment from Tollbit’s Joslin on publisher fee rates. A correction was made regarding the revenue publishers are making from Tollbit’s platform.

What we’ve heard

“It’s been a decent amount of… ancillary revenue. But it’s not something we count on.”

– An anonymous publishing exec on the uptick in Facebook revenue they’ve seen in the past year.

Numbers to know

70%: The decline in traffic some small and mid-sized publishers are seeing since Google’s AI Overviews rollout.

Over 300: The number of tech workers that are part of the newly-formed union at The Washington Post.

8: The number of years Radhika Jones served as editor of Vanity Fair before announcing she would step down this year.

$4.99: The cost of Gannett’s standalone true crime subscription.

What we’ve covered

Basketball team Atlanta Hawks launches NBA’s first creator collective

• The Atlanta Hawks Creator Collective is a group of 25 content creators that the NBA team has granted special access to home games and events
• Members of the collective are paid a fee to create content about the games they attend, in addition to a mixture of other paid and unpaid opportunities.

Read more about the new collective here.

Some news publishers see resurgence of Facebook referral traffic

• An uptick in Facebook referral traffic at The Hill, Newsweek and Salon may signal a “Trump bump” in political news coverage, or a renewed interest in news following the presidential election.
• However, publishers aren’t sure what to make of the increases they’re seeing.

Read more about why publishers are in the dark here.

Here are the challenges publishers are confronting, from AI to Google

• During the spring 2025 edition of the Digiday Publishing Summit, executives from Dotdash Meredith, Punchbowl News, TelevisaUnivision and more spoke out on the most urgent challenges facing publishers today.
• Those issues range from AI to Google

Watch Digiday’s video on what challenges publishers are confronting here.

Subscriptions and events gain steam among publishers’ most significant sources of revenue

• Direct-sold ads continue to be the dominant source of publishers’ revenue in Q1 2025.
• Other revenue sources like subscriptions and events are also gaining importance.

Read more about where publishers are getting their revenue from the Digiday+ Research report here.

What we’re reading

BBC upset about minimizing source attribution on Apple and Google news services

The BBC filed a complaint to a U.K. antitrust regulator alleging that Apple and Google news services don’t make it clear enough which sources are providing the news, Apple Insider reported. Minimizing the source attribution devalues BBC’s content, the organization argued. 

The Washington Post is looking into micropayments

The Washington Post is internally discussing using micropayments to give more people access to its paywalled journalism, Axios reported. The Washington Post has tested a flex payment system that lets people pay to access The Post for seven days.

G/O Media sells Quartz

G/O Media has sold the business news site Quartz, as well as commerce site The Inventory, to Canadian software company Redbrick, Axios reported. G/O Media’s remaining publications include Kotaku and The Root.

Will AI enhance or hollow out the media business?

The news industry has been in decline for a while, and AI technology can make it easier to access information — which could upend media companies and make human writers more valuable. This New Yorker piece looks at the threats and opportunities, and how people can integrate AI into their news-reading processes.