Case study
Chicago Sun-Times and Philadelphia Inquirer: A Summer Reading List of Books That Don’t Exist
In May 2025, two big US newspapers printed a summer reading list. Ten of the fifteen books didn’t exist. A freelance writer had used AI, and nobody along the way checked.
What are the labels? The Evidentiality Framework asks an AI to mark each claim it writes: (g) generated, its own work; (u) given, passed to it by someone else; or (m) checked against a named source. The marks stay on a claim while people work with it. How the labels work.
What Happened
- A freelance writer produced a summer section, “Heat Index”, for King Features, a company owned by the media group Hearst that sells ready-made content to newspapers.
- It ran in The Philadelphia Inquirer on 15 May 2025 and in the Chicago Sun-Times that weekend. Its reading list paired real authors with made-up books, such as “The Rainmakers” by Percival Everett. Ten of the fifteen books were fake.
- The writer said: “I do use AI for background at times but always check out the material first. This time, I did not.”
- The Sun-Times said the section “was not created by, or approved by, the Sun-Times newsroom”. Its owner later said the section was never shown to Sun-Times journalists for review. The Inquirer’s publisher called it “a violation of our own internal policies and a serious breach”.
- King Features ended its relationship with the writer. It said he had broken its AI policy, which he disputed. The Sun-Times’ owner stopped buying special sections from King Features.
Follow the Claim
This is an illustration of how the labels would have worked, not a test. It only holds if the conditions under “What Would Have Had to Be True” held.
Key: as it happened, the type gets bigger as the claim sounds more certain. With labels: red (g) generated: written by the AI; green (u) given: passed on, with who said it; blue (m) checked against a named source.
The book does not exist. The title and author are as printed, as reported by WBEZ.
How the Labels Could Have Helped
- The warning travels. The writer’s unchecked list reaches King Features still marked as unchecked.
- Printing waits for a check. Unchecked claims don’t go into a finished section.
- Each step can see the last. Everyone down the chain can see what the step before checked and didn’t.
What Would Have Had to Be True
The four conditions every case shares: the AI tool used the labels; it labelled its own work correctly (the weakest link: in our tests, AI sometimes mislabels its own work); the label stayed on when the text was copied; and someone owned a rule that unchecked claims don’t go further. In this case:
- The rule is owned by the editors at King Features, and by each newspaper for what it prints under its name.
What Already Existed
- Editing exists. Newspapers edit what their journalists write. This section was bought in and skipped that review.
- The seller had rules. King Features said the writer broke its AI policy. He disputed that.
- A simpler check would have caught it: searching for each book. Labels add one thing: the writer’s own checking, or lack of it, travels to every buyer.
What the Labels Wouldn’t Have Caught
- Bought-in content. The section wasn’t reviewed by the papers’ journalists. That is a business decision labels can’t change.
- Openness about tools. Labels only work if the writer uses them.
- The rest of the section. Other pages had their own problems. Each claim needs its own label.
Further Reading
- WBEZ: the Sun-Times’ review of the special section
- WBEZ: bought-in content in the Sunday Sun-Times
- AP: a newspaper’s summer book list recommends nonexistent books
- Axios Philadelphia: the Inquirer’s AI-generated reading list
- A.V. Club: the Sun-Times and Inquirer summer guide
Next: All case studies · How the labels work