Getting Paid in the Age of Answer Engines: exploring People Inc.’s AI Model
Media companies urgently need to take control of who can access their information and on what terms to cash in on the ever-expanding AI information market. That is the advice of Jonathan Roberts, Chief Innovation Officer, People Inc. – one of the first media companies to strike up AI partnerships, including with Microsoft and Meta.
“There is an explosion of need for content in this new AI world, and just like everybody understands magazine distribution, everyone understands search,” he said during a conversation with Stig Ørskov, Chief Executive Officer, WAN-IFRA, Germany, on the first day of the FIPP World Media Congress.
“There’s a 100 times bigger search market getting built today, and nobody owns it. It’s a wide-open field for a huge new search market. We know how valuable those are, and that’s only if you ask new, different, more interesting, more nuanced, more specific, more niche questions than you ever had before, which means not only is there more demand, but we need new supply.
“This isn’t for us a question about how we make ourselves whole for the damage that we’ve lost. This is a question of there being a completely new market that is huge. How do we become part of it?”
Locking the doors
People Inc. has adopted a three-pronged strategy when it comes to AI partnerships: block unauthorised access, negotiate payment and build better services for legitimate commercial partners.
Roberts revealed the company moved to a comprehensive blocking strategy in January, preventing automated systems from accessing its content unless they had been specifically authorised.
Rather than relying on robots.txt instructions, People Inc. blocks traffic at the content delivery network (CDN) level.
“If you’re a parasite, we block you. If you’re a platform builder, we want to partner with you.”
“That’s the layer that now blocks anything non-human apart from 38 partners, and when we did this, we saw a few things. Firstly, we lost no revenue, and we lost no traffic,” Roberts said.
“Our ad value actually went up, not down. It has been a universal success from a traffic and revenue perspective, and we now block tens of thousands of unique bad actors per day and tens of millions of daily attempts to get our content. It is industrial what is happening to your site.
“If somebody is robbing your house, why would you not lock the door, right? Independent of whether or not you’re getting a business out of this, for the love of God, lock your door, or at least close it, which the majority of people have not yet done.”
Show us the money
Roberts explained that the AI partnerships they have with OpenAI, Meta and Microsoft serve different purposes. The OpenAI and Meta agreements involve licensed use of content within their respective AI products.
Meanwhile, the Microsoft arrangement offers a marketplace through which publishers’ information could potentially reach numerous businesses developing AI-powered services – something that, Roberts believes, offers a longer-term opportunity.
Roberts compared the current market for AI content to the rise of Napster in music industry when digital distribution created huge demand but offered little compensation to rights holders.
That, of course, led to the emergence of legitimate platforms such as Spotify which provided a commercial structure. Roberts believes the AI industry needs to follow down the same road.
“We didn’t get Spotify until we had Napster,” he said. “Now, what I would like is let’s get some platforms where this obviously large and vibrant market can actually start paying for the content it needs, because then we can invest into it.”
Being a great partner
Roberts also outlined how People Inc. is changing the way it supplies information to its AI partners – with good communication key to success.
“With our partners, it’s great when we can find out what’s needed, so we can build into it,” he said. “It’s also better when you’re able to provide information in a way that’s much more efficient, right?
“So, web crawling is a very bad way to get information, right? It’s mostly HTML junk created for humans on the internet. It is absolutely useless for agents. It’s only getting crawled because that’s kind of the USB connector of content.”
Instead of requiring systems to crawl conventional webpages, the company has developed an API that delivers structured content in Markdown format, together with relevant metadata and rights information.
“It’s maybe 5% of the world on an HTML page of all the stuff that a human would ever care about. It’s therefore 95% cheaper to reason over than the full page, it has no distractions because it’s just the content.
“It has clean metadata fields that are very easy for people to make decisions about like when was it written, when was it last updated, what are the bylines – things that every journalist knows is important, but no AI company has yet figured out is useful.”
Offering advice to smaller publishers, Roberts implored them protect their content and take part in collective licensing initiatives that could strengthen their negotiating position.
“At the moment, you’re putting yourself on a newsstand and everyone’s stealing it and not paying for it. Step one: stop them stealing it.”
