The shape of online search has changed dramatically, with current engines increasingly powered by generative AI models whose training data was crawled and scraped from the web for free—and without any compensation going to the people who created it. As a result, news organizations, media companies, and even individual authors have sued AI startups for using copyrighted content without permission or payment.
Now, Bill Gross (BS ’81) is trying to render such lawsuits unnecessary by making online search fair again. His new startup, ProRata, has developed an answer engine that credits and compensates creators when generative AI draws on their content. The company has signed revenue-sharing deals with more than 500 publishers, including Time and The Atlantic, and plans to move into the realms of video and music as well.
Gross is well-positioned to change the search landscape: In the 1990s, he and his startup incubator, Idealab, launched Goto.com, a search engine that served up sponsored links but only required advertisers to pay when users clicked on them. That pay-per-click model soon became ubiquitous, and Gross is betting the same could happen with ProRata’s attribution-and-monetization model.
“The right way to solve this isn’t with lawsuits,” Gross says. “The right way to solve this is with a fair business model, because a fair business model would enable everybody to participate properly and you wouldn’t need to sue anybody.”
“The right way to solve this isn’t with lawsuits,” Gross says. “The right way to solve this is with a fair business model, because a fair business model would enable everybody to participate properly and you wouldn’t need to sue anybody.”
Other startups are focusing on arranging licensing deals for training data. Why isn’t that good enough?
Bill Gross: There are two forms of payment that creators should earn for their content. One is for training, when the content is used to teach a system how to give answers. That’s pay-per-crawl, or getting paid on input. Then there’s payment on output, when your content shows up in an answer.
Payment on input is a one-time fee. Payment on output is like a royalty that’s paid every time the content is used. I think content creators should get both.
How does ProRata identify the sources that inform an AI search answer?
BG: It turns out that you can use a large language model to break up the output from generative AI into a series of claims and insights, then look for those claims and insights in the original documents and calculate the relative attributions.
We only track content for people who give us explicit permission to do so, and we offer them a 50/50 revenue share: We take all our revenue, whether it’s subscription revenue or advertising revenue, and we just split that 50/50. They’re thrilled because right now they’re not getting anything.
Exactly what happens when someone performs a ProRata-powered search?
BG: You ask a question: “What’s the latest on tariffs?” It comes up with an answer in plain English that uses two articles from Time and two articles from The Atlantic.
It will give you those sources: It will say 50 percent of this answer came from The Atlantic, and 50 percent came from Time. And there will be an ad on the page—maybe from a shipping company that can help you get more products in your store before the tariffs hit. That shipping company might pay $5 for a click-through on their ad. Of that $5, $2.50 will be split between Time and The Atlantic.
In fact, we take all the advertising and subscription revenues over the course of a month and divide it pro rata, or proportional, to each of their relative contributions. Each partner will get a statement at the end of the month that says something like, “We had 42,150,000 queries. You were used in 7,422 of them. On average, you were 23 percent of the answer. Here’s your check.”
Where can people perform ProRata-powered searches?
BG: We have our own answer engine called Gist.ai. We also serve up answers on other people’s websites. Most of our partners already have internal search at their sites, but it’s not AI search and it’s pretty lousy. We ask all our partners, “How would you like to upgrade your search to AI search? We’ll power it and we’ll split the revenue with you 50/50.”
Are you hoping to challenge Google, or maybe get acquired by them?
BG: I am not trying to take on Google. I am trying to partner with all the others who are trying to take on Google. I am also not looking to be acquired. I want to provide the business model that powers everyone else who provides AI answers besides Google—and that group is growing fast.
You claim that your approach will also increase the accuracy of AI search. Why?
BG: Because we’re only answering questions based on articles from 500 publications, as opposed to the whole internet.
What’s the benefit of that? Most of these are vetted sources that have editors who try as much as possible to be truthful. We’re not including Twitter [now X] and Reddit and random people who are just shouting. By only sourcing quality content, you have fewer hallucinations.
Will this eventually trickle down to smaller content producers, like individual writers, artists, and musicians?
BG: I would like it to trickle down to everybody. Anybody who is a creator should benefit from this. Our tagline is, “We want to protect and advance human creativity in the AI era.” That applies to everyone, big or small.