AI Is Consuming the Web It Depends On

In What Just Happened to TheNumbers.com Should Worry Us All, Stephen Follows breaks down how AI traffic helped overwhelm one of the film industry’s most valuable independent data sources.

“Only 10% of their traffic is from humans browsing the site.”

By early 2026, The Numbers was being hammered by training crawlers, live AI searches, autonomous agents, and other automated traffic. Its small team was spending approximately 90 percent of its time keeping the existing site running instead of improving it. The thirty-year-old server eventually collapsed on March 5, forcing the team to rebuild a limited version of the site on new infrastructure.

The logs apparently showed more than indiscriminate scraping. Some automated visitors were searching for back doors that could provide early access to unpublished data or potentially manipulate what users saw. Nobody has established who was responsible or exactly what brought the server down.

I would not be surprised if prediction markets had something to do with the targeted interest. Polymarket uses The Numbers as the official resolution source for box-office wagers. One recent opening-weekend market alone recorded nearly $370,000 in trading volume. Getting the final numbers even slightly before everyone else could provide a profitable advantage. That is a credible motive, although there is currently no proof connecting Polymarket or any particular trader to the attack.

The larger story is what AI is doing to the valuable sources it depends upon. Independent websites spend years collecting, verifying, organizing, and publishing useful information. AI systems then retrieve that information at enormous scale, frequently without sending readers back or paying enough to sustain the source.

The Numbers sells licensed access to its structured OpusData service and explicitly prohibits systematic scraping of its public website. Yet the public site still had to absorb an overwhelming amount of automated demand.

There is a nasty contradiction here. AI becomes more useful by consuming reliable human-maintained data, but its appetite may make those same sources too expensive and dangerous to keep online.

When the original sources disappear, we do not merely lose a website. We lose the people maintaining the facts that everyone else, including the AI, treats as canon.

Read Stephen Follows’s full article, What Just Happened to TheNumbers.com Should Worry Us All.