So, the next generation of search, as Google has framed it thus far, leans towards giving people an answer on the spot rather than a list of links to go and read one. For media companies, the floor disappearing.
This is Google Zero. It describes the scenario where referral traffic simply tanks, it hits zero (nada, zilch). For sites built on search discovery, rather than via a direct readership, that's terminal.
So, who feeds the machine?
Here's the rub, search needs publishers, just as much as publishers need search.
Google's answers exist because someone, somewhere, reported the story, ran the investigation, or wrote the review. If the direct-answer model kills the economics that pay for that work, then the well runs dry. Bottom line: no new reporting, no new material for the next answer.
An inbreeding problem
Does this then create a recursive loop? This is a scenario where large language models generate text, that text ends up back in the training data for the next generation of models, and the cycle repeats. Slop squared.
One commentator referred to this as 'inbreeding', which, I think, is self-evident. Each pass, the output drifts further from anything a person actually observed or verified, computers increasingly learning from what other computers made up.
What next?
For anyone still building a content strategy around search visibility, that's the uncomfortable bit. The traffic problem might be survivable with a different distribution plan. The provenance problem, models trained on models trained on models, is harder to route around.