By Anabelle Nicoud, who is based in San Francisco and will write a series of articles for WAN-IFRA, focusing on the intersection of AI, Silicon Valley and news media.

A number of significant deals between AI companies and publishers in the US and Europe were announced this summer, including AI startup Perplexity. It was one of the first such deals that actually included a revenue-sharing element with what Perplexity calls its Publisher Program.

Time, Fortune, Der Spiegel, Texas Tribune, and WordPress are among the first publishers onboarded to the program. Perplexity says it plans to introduce advertising through its related questions feature. Brands can pay to ask specific related follow-up questions in its answer engine interface and on Pages. When Perplexity earns revenue from an interaction where a publisher’s content is referenced, that publisher will also earn a share.

What publishers will get

The partnership involves multiple facets, including a revenue-sharing model, access to Perplexity’s tools and APIs, and analytics sharing.

These elements aim to create a mutually beneficial relationship between the AI company and publishers. Perplexity also announced a $250,000 gift to the Northwestern University School of Journalism to support research about AI and journalism.

“Our success is tied to the success of the production of new facts about the world,” says Perplexity’s Chief Business Officer, Dmitry Shevelenko.

Der Spiegel, one of the largest magazines in Germany, is experimenting with AI tools at various stages of content production, from versioning content to fact-checking. A partnership with Perplexity offers a new way to utilise AI, believes Christoph Zimmer, Chief Product Officer at Der Spiegel. We are convinced that AI will have a big impact on the way news is consumed,” he says. “For us, it’s not just about revenue sharing, but rather about experimenting and collaborating.”

An answer engine?

Perplexity uses large language models (LLMs) to summarise internet information. Unlike traditional search engines that provide links, Perplexity aims to directly answer people’s questions to summarise information when answering users’ questions while also featuring sources.

“The best way to think of Perplexity is as an answer engine. People come and ask questions, and we provide answers”, explains Shevelenko. 

This is an important distinction the company has strongly emphasised: contrary to other San Francisco and Silicon Valley giants and models, Perplexity doesn’t train any models or foundational model.

“To the extent to which Perplexity has its model, it’s taking the (Meta) Llama 3-based model, but fine-tuning it for conciseness and for being good at synthesising web content,” says. Shevelenko.

In its free version, users can pose a question resulting in an answer that shows sources then Perplexity’s answer, and possible follow-up questions.

Since its launch in 2022, Perplexity, which was funded by ex-Meta and OpenAI employees, has received solid financial backing by major tech players, such as Nvidia and Jeff Bezos – and received an additional funding round of $250 million this summer, reportedly increasing its valuation from $1 billion to $3 billion. 

The idea of working with publishers came before a recent wave of criticism and accusations of plagiarism from Forbes and Wired – which Shevelenko attributes to a misunderstanding of how the product works. 

“It’s unfortunate, but it is very fair for publishers to be skeptical of new tech platforms because the history of the last 30 years is that publishers haven’t gotten the best deal,” he adds.

The program encompasses several aspects. First: the revenue-sharing program. “The reason we’re doing this now is we wanted to put this in place before we even launch our advertising products,” says Shevelenko.

The company is careful about how it wants to integrate news, and insists on saying if users don’t come primarily for news, access to good information is going to create a better experience. That’s why it decided to build a revenue-sharing program, based primarily on advertising. Some mention a double-digit percentage for publishers as part of that equation, but Shevelenko didn’t confirm during our interview.

“What I’ll say is it needs to be significant, because this is our main business model. If it’s not significant, Perplexity as a company isn’t doing well, and we’re not in a good place,” says Shevelenko.

Also: the publishers will have access to Perplexity’s tools and APIs. That would help them run the search engine on their own content, for example – an idea that could be an incentive for publishers, as AI-powered chat bots have been used by several newsrooms this year. 

The publishers’ content won’t be used to train the models, and the publishers won’t necessarily be favoured by the algorithm when it will display an answer. “It’s like church and state: our search engine doesn’t know who our publishing partners are,” says Shevelenko.

Where ScalePost fits in

It is also interesting to note that Perplexity will work with another tech start-up, ScalePost, to share analytics with their publisher partners, allowing the publishers to see granular insights about how their content is monetised. 

Co-funded by Zachary Todd and Ahmed Malik, ScalePost, based in the San Francisco Bay Area, sees itself as a bridge between publishers and AI companies and says it’s currently working on different partnerships. 

“We’re the default third party that will provide analytics to publishers,” says Malik. “We want to be the one-stop shop for any content needs from different types of partnerships.” 

Perplexity aims to extend its program to more publishers, notably in Europe and Asia, two of the company’s fastest-growing markets. “If you’re producing high-quality content, we absolutely want to work with you,” says Shevelenko.