Did AI Invent Content Slop?

In November 2017, The Shed at Dulwich was the top-rated restaurant in London on TripAdvisor. TV executives used their work email addresses to request bookings, and a PR agency pitched a Batman-themed launch. Suppliers sent product samples, and a local council wrote to offer a relocation site in a new development.

Yet, The Shed at Dulwich had no kitchen, no chef, and no food. Only the address was real: it was a garden shed in Dulwich, in south London. The entire enterprise turned out to be nothing more than a $10 burner phone and online reviews written by the friends of Oobah Butler, a journalist who created the prank.

Nearly a decade later, Skift’s “Welcome to the Travel Slop Era” makes a legitimate case: AI systems have recommended destinations that don’t exist, doing so with a confidence that reads as authority. Tourists traveled to the mountains of Peru in search of turquoise pools hallucinated by an LLM, raising concerns. Misinformation like this can cause real harm and have real consequences.

The era of unreliable travel information, though, did not begin with generative AI. Fake reviews, SEO content farms, and social media algorithms have promoted engagement over accuracy as a matter of structural design for decades now. AI has added a new dimension to the problem, but it didn’t cause it.

Instead, AI has introduced a new combination of scale and unearned authority. Butler’s hoax required eight months of coordinated effort. Fake review operations required an army of friends and some money. Generative AI produces confident misinformation instantly at volume, in a register that mimics deep expertise.

At the same time, the technology carries, embedded in its architecture, something no previous travel recommendation system has ever provided. The most capable AI systems cite authoritative third-party sources and name them. A well-built system, asked where to eat in Lisbon or where to stay in New York, reaches for trusted publications, attributes its recommendations, and creates a verifiable chain of evidence linking the claim to its source.

Citation architecture is only as good as the sources it cites. An AI system attributing a recommendation to a publication that has itself traded editorial rigor for content volume has not resolved the slop problem; it has given the slop a credible byline.

The brands that have spent decades building authority on the promise of trusted, expert guidance will now be tested in ways they have not been before. The competition is no longer only for readers. It’s the sources LLMs use when a traveler asks what to do, which hotel is worth the price, and which brands deliver on their promises. That competition will be won by editorial quality: by journalists who have been to the places they write about, by editors who hold the line on standards, and media outlets that treat the recommendation as a matter of genuine accountability.

The Shed at Dulwich succeeded because TripAdvisor had outsourced trust entirely to the crowd, a trust that could be manufactured. The crowd-sourced review was the defining travel mechanism of the 2000s and 2010s, and its fundamental weakness was always the same: no named expert, no editorial standard, no institution with a reputation at stake.

The citation-driven AI model, at its best, puts named and reputable institutions back at the center of how travelers receive guidance. Media outlets with genuine expertise, editorial independence, and a commitment to accuracy stand to become more influential in this environment, provided they do the work that has always distinguished them from the noise.

Butler opened The Shed at Dulwich for one night in November 2017, after eight months of tricking the internet. He served microwaved meals, surrounded the guests with actors loudly praising the food, and charged nobody. On their way out, one couple asked whether, having dined once, it would now be easier to get a reservation.

Travelers have always needed someone they could trust. The question is which media outlets will hold themselves to the standard the moment now requires.