AI: should publishers sign up or sue?

Some publishers are signing deals with the AI companies. Some are suing them. Which is the right response? A lively, post-lunch panel discussed the future…

Panel

• Charlotte Tobitt, UK Editor, Press Gazette (moderator)
• Barry Adams, SEO consultant, Polemic Digital
• Madhav Chinnappa, VP, Partnerships Human Native AI
• Denis Haman, CEO, Glide Publishing Platform
• Carly Steven, Global Head of SEO, Mail Online

Up until now, the Future of Media Technology Conference has been pretty positive about AI. That ended abruptly with this panel. “It’s abusive,” snarled Denis. Nobody on the panel thought it was right that the AI companies had trained their products on journalism without recompense. There was a distinct sense of scepticism about what we might get out of the relationship. And a certainty that we should all block the AI crawlers from accessing our sites —unless they’re paying.

Surprisingly, the panelists seemed less worried about AI overviews in search than one might have expected. Barry, one of the two SEO specialists on the panel, has seen less impact in traffic than he expected, and despite having researched the percentage of queries likely to be impacted, Carly, wasn’t seeing what they expected either.

Each side in this power struggle thinks about value in a very different way, suggested Madhav. The tech companies think about value in terms of dollars in, while journalists think about the value to society of what they do.

Barry recalled a study that found that people trust search results more when they have a news brand at the top of the results — although Madhav thought the methodology of that study was “bollocks”.

“We’re not friends. We’re not even frenemies. The media plays a more important role in society than some random number cruncher at Meta assigns to it,” said Denis. “There’s an opportunity, with the emerging crop of new technologies, for us to band together and approach the government. Yes, the big publishers can do deals, but who will compensate smaller publishers?”

Publishers need to organise against AI

Google’s spam update that killed all the affiliate and voucher content slashed a revenue stream almost overnight. But the publishers all came together to insist that Google talked to them about it. Carly isn’t suggesting that the publishers got what they wanted — but the ability of publishers to co-operate was a really positive step forward.

“Collaboration is really important,” said Madhav. “In this era of AI, publishers should be collaborating to work out what the sustainable model is here.”

He doesn’t believe that the sorts of one-off deals some publishers are cutting with the AI companies are sustainable. They’ll last until there’s a court case, and then they’re done, he predicted. But if we can work out a model that benefits both sides, that’s sustainable.

Denis suggests that the value of content is greater to the AI companies than it was to Google for search. If the AI companies are to avoid model collapse, they need fresh human-generated content to train the models on.

Does the AI industry need us?

“You either care about this industry, or you leave,” said Denis. “Something needs to happen that is different from what we’ve done before, otherwise the outcomes will be the same.”

“The more I learn about LLMs, the less impressed I am,” said Barry. He read a paper that suggested that to approach anywhere near human intelligence, it needs to consume more contents that humanity is capable of creating.

Denis reminded us that the AI head at Microsoft describes all content on the web as “freeware”. Journalism has more inherent value than a Fortnight streamer (whatever nine-year-olds might think).

Easing the tension

“There’s always a tension between publishers or platforms,” said Carly. “You rely on them for revenue, but you have no control over what they do. But there have been some green shoots recently. Our relationship with Google, while it can be frustrating, there have been some small victories.”

The publishers are collectively speaking to Google at scale with data — which is what Madhav suggested you needed to do to have an impact.

Most of the signals Google use are those around content, rather than the content itself. Google is still not good at doing that. They know about how people engage — or don’t — but are still not good at understanding the signals that humans use to understand content quality.

“They’re just human-coded algorithms, making editorial decision on what we see. It’s OK to hold them to account,” said Barry.