The year AI companies pay for the value of publishing

Twenty-twenty-six will be the year that one of the AI search companies builds into their model a genuinely fair exchange of value with the publishing industry.

Right now, AI search is mostly extractive. The bots scrape the sites and answer search queries inside a box that users never leave. In rare, but positive, circumstances, the publishers might get paid for the initial scraping — and they might be given assurance that their content will be surfaced accurately, ethically, and with fair attribution. But the old model of the Internet, in which nodes like Google and Facebook send readers back to publishers and creators, doesn’t exist yet in AI. Very few publishers are actually getting readers from AI search.

There are many ways to resolve this problem. External links from search boxes could become more prominent. The AI companies could discern user intent, or even particularly high-value queries, and guide certain users back to the sites that provided the initial information. A large company could adopt a model, like the one build by ProRata, where publishers are paid based on their contributions to answers to specific queries. Advertising inventory on the AI search sites could be offered based on the value of the information that’s been scraped. And these marketplace features could roll out with stronger guardrails for publisher attribution, factual accuracy, and journalism ethics.

For now, the AI companies haven’t adopted ideals like these because they don’t have incentives to. They’re in what they see as a race to the infinite wealth and power of AGI, and they don’t want to slow for a second. But a healthy publishing industry is good for the AI industry because it means better, more accurate, information flowing into the models. And, next year, one of the big AI companies is going to figure that out.