Since its launch in 2022, Perplexity has staked its claim as the generative AI disruptor of Google search. Its core product uses conversational AI to synthesize and summarize sites and news articles from the web in real time, rather than simply offering up a list of links to those sites or brief article previews. With recent investors Jeff Bezos and Nvidia pushing Perplexity’s valuation to $3 billion in April, it certainly has established itself quickly in an already oversaturated generative AI field.
Now, Perplexity is setting out on a path not taken by Google search — paying news publishers directly to index their content.
On Tuesday, the startup announced the launch of its Perplexity Publishers’ Program. The first batch of publishers includes six partners: Time, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune, and Automattic.
The most notable prong of the program is a revenue-sharing agreement. Before the end of September, Perplexity plans to roll out a new advertising feature, through which advertisers can pay to display brand-sponsored suggested follow-up questions at the bottom of each Perplexity search answer.
If, say, Time is used as a source in a Perplexity response and its reporting appears above that sponsored follow-up question, Time will get a cut of the advertising revenue. Perplexity hasn’t made the specific revenue split public, but said the rates are currently standardized across all the publishers in the program.
“This is not a reactive PR thing for us. There needs to be a vibrant business model for the production of new facts about the world and journalism is that vehicle,” Dmitry Shevelenko, Perplexity’s chief business officer, told me during a recent call. “If people truly understand what the implications of revenue sharing are, that actually should address a lot of the skepticism, right? Because we’re truly tying our success to the success of the publishers creating content that’s useful to answering people’s questions.”
Over the past month, Perplexity has become a lightning rod for debates about the internet scraping practices of generative AI developers, and their use of original reporting from news publishers without consent or compensation.
First, Forbes accused the company of plagiarism for republishing its original reporting on former Google CEO Eric Schmidt without citing the story directly in the copy of its post.
The story appeared in Perplexity’s recently launched “Pages” feature, not Perplexity’s core search product. But the story was also sent as a push notification to Perplexity mobile users and used to create an AI-generated podcast, neither of which cited Forbes at all. “That became a YouTube video that outranks all Forbes content on this topic within Google search,” Randall Lane, the chief content officer and editor of Forbes, wrote in an opinion piece published in June. “Perplexity had taken our work, without our permission, and republished it across multiple platforms — web, video, mobile — as though it were itself a media outlet.”
The opinion piece was paired with a letter from Forbes to Perplexity threatening legal action for copyright infringement.
Perplexity’s knock off of our reporting feels like a crystallization of the "Can journalism survive AI?" convo. The company took our paywalled work, without our permission, and competitively broadcast it across web, video, mobile — as though it were itself a media outlet. https://t.co/kEi15MYOQI
— John Paczkowski (@JohnPaczkowski) June 11, 2024
The following week, Wired published an investigation that showed Perplexity was evading the Robots Exclusion Protocol, an honor code system that tells bots which parts of a website not to scrape. Condé Nast reportedly sent its own cease-and-desist letter to Perplexity after Wired’s reporting documented the startup’s unlabeled crawler hitting several Condé-owned domains.
Following a month of headlines effectively accusing Perplexity of theft, the partnership announcement may seem conveniently timed. But Shevelenko stressed the program was in development long before Forbes came out with its allegations.
He also admitted Perplexity has had a serious messaging problem. “We need to be more proactive in communicating how we’re different,” he said, pointing to the fact that Perplexity is not scraping the internet en masse to train foundation models the way OpenAI does. (Perplexity’s search product is, however, built on top of large language models, including GPT.) “We will treat this as a priority communication point for us, so people understand the nuances of these issues.”
Shevelenko also pointed to specific changes Perplexity made to “Pages” in the days after the Forbes story, including launching in-copy citations, as evidence of the company’s responsiveness. He stopped short of taking full responsibility for the recent criticisms.
“Some of this climate has nothing to do with Perplexity. When OpenAI announced their deals with Atlantic and Vox, Atlantic and Vox journalists were some of the most vocal critics of those deals,” he said. “I really empathize with the management of a lot of publishers who were like: What makes working with us hard isn’t necessarily that they don’t think this is a great deal. It’s just that any affiliation with an AI company is challenging in the current environment.”
Shevelenko also pushed journalists to use Perplexity themselves, implying many critics have not: “See for yourself what the fault lines are, rather than approach it from arguably misleading, sensationalist reporting.”
There are some standout names in the initial list of publishing partners, including The Texas Tribune, the award-winning Austin nonprofit. “It was definitely non-trivial to get them comfortable, because they obviously take their brand seriously,” said Shevelenko. “The revenue-sharing piece — really, nobody’s offered them this before. I think that got them kind of out of their comfort zone.”
OpenAI, to date, has failed to sign a similar major partnership with any nonprofit news publisher as part of its ongoing licensing deal agreements. The blind spot means some nonprofit outlets fear they won’t have a seat at the table as norms in the burgeoning generative search industry are cemented. The concern is one factor in the decision of outlets like The Intercept to sue OpenAI for copyright infringement, as I reported in March.
“Citations and referrals are important. Hopefully, they are driving people to our products, and hopefully, it makes people see the value in donating to us. But direct revenue is important, too. Our journalism may be free, but it’s not free to produce,” said Sonal Shah, the CEO of The Texas Tribune. Shah emphasized that she can only speak for the Tribune — which, unlike some news publishers, already permits free republication of its stories.
The dark cloud over Perplexity was discussed in negotiations over the partnership, according to Shah. “We have had these conversations with Perplexity,” she said. “We want to ensure that we can be at the forefront of these conversations and build a relationship with Perplexity and other AI developers to ensure that our content that is already shared for free is done in a way that is transparent and sustainable for everyone involved.”
Automattic, the parent company of WordPress, is another notable name on the roster. As part of the partnership, Perplexity will have access to publishers that host their work on WordPress.com — though not to Automattic-owned publications like The Atavist and Longreads.
WordPress.com is the company’s free tier. In a statement, Megan Fox, a spokesperson for Automattic, clarified the deal excludes publishers hosted on the premium WordPress VIP, including customers like NewsCorp. The deal also carves out an exception for smaller news outlets that use Newspack, a service for local news publishers hosted on WordPress.com, including CalMatters, Capital B, Reveal and Houston Landing.
According to Fox, WordPress.com publishers whose blogs and sites are surfaced through the Perplexity partnership will take a cut of the revenue earned and can opt out of the program if they choose. “We believe this is a great opportunity to explore new ways to bring more value to our customers in two main ways: Drive additional traffic to their websites and provide additional opportunities to participate in other revenue-sharing programs,” she said in a statement.
“This isn’t just for big publishers, right? This is for anybody on the internet that creates high quality, fact-rich content,” said Shevelenko of the move to fold the blogosphere into its program. The partnership allows Perplexity to sidestep individual outreach to these sites. “It’s not this kind of rich-get-richer type of program. We want to make this broadly available. WordPress is just an incredible partner for that scale.”
Currently, there is no plan to tweak Perplexity’s search to prioritize a partner publication’s story over a competitor’s, for fear of eroding user trust. Partners may receive other perks, though, including different design treatments when their work surfaces organically. “Where there’ll be more prominence is how we cite sources at the top of an answer,” Shevelenko added. “That ostensibly would lead to higher click-throughs to go back to their content.”
Shevelenko also notes that publishers will have free access to Perplexity Pro for their teams, and to Perplexity’s API. Some partners are already experimenting with embedding Perplexity’s search product into their sites, with the option to filter the searches to only surface articles in their own domain.
The platform will also allow partners to see what queries are surfacing their content and driving revenue through access to the analytics platform, Scalepost.ai. Shevelenko hopes this information can be used by publishers to inform editorial decisions. “It’s almost like a healthy version of SEO, where we’re incentivizing the production of facts, not the production of meta tags to show up in 10 blue links,” he said.
At minimum, the partnership marks a new chapter in the era of generative AI content licensing — and a turn away from the terms set by ChatGPT.
“One of the things that doesn’t feel right about OpenAI’s strategy is the very big names get the big lump sum deals, because OpenAI is worried that those are the folks that have the legal resources to sue them. And everybody else gets nothing,” Shevelenko claimed, pointing to recent OpenAI contracts with NewsCorp and Vox Media. “We’re truly going to open this up to everyone.”
