Ten of the Fifteen Books Didn't Exist.

Isabel Allende did not write a novel called Tidewater Dreams. Andy Weir did not write The Last Algorithm. Both titles ran in the Chicago Sun-Times last May, inside a 64-page summer supplement called "Heat Index," on a list of fifteen recommended books. Five of the fifteen were real. All fifteen authors were real people. The rest of the books had been invented and attributed to them.

The same supplement ran in the Philadelphia Inquirer.

A novelist named Rachael King noticed and posted about it. 404 Media broke the story. By Tuesday morning the Sun-Times had a statement out: "It is not editorial content and was not created by, or approved by, the Sun-Times newsroom."

Every word of that is true. It is also the entire problem.

The handoffs

The freelancer who assembled the list, Marco Buscaglia, didn't hide behind anything. "I do use AI for background at times but always check out the material first. This time, I did not and I can't believe I missed it because it's so obvious. No excuses. On me 100 percent and I'm completely embarrassed."

Take him at his word. His failure was one skipped check on one assignment, which is the kind of error every person who has ever shipped work has made. That is not what put fabricated books in front of two cities.

What did that was everything downstream. King Features Syndicate, a Hearst company, produced and distributed the supplement. The Sun-Times inserted it. The Inquirer inserted it. Nobody in that line opened a browser and typed Tidewater Dreams to see whether it existed. The check was trivial and it was not anyone's job, and those two facts are related.

Then Buscaglia said the thing that should stop anyone who runs a content pipeline: "We never get a list of where things ran."

The person closest to the work did not know whose names would end up vouching for it. The people whose names were on it did not know what the work was. That is a chain of custody with no custody in it.

What a masthead is actually for

A masthead is not a logo. It's a promise about process. It says somebody with a job and an editor and a phone number looked at this before you did, and if it turns out to be wrong, there is a person who answers for it. Institutional credibility is that promise kept over and over, for long enough that readers stop consciously checking. It takes a very long time to build and it is spent all at once.

The Sun-Times statement drew a line between editorial content and a syndicated advertorial. Inside the building, that line is real, and it matters. On a subscriber's kitchen table it doesn't exist. One reader put it plainly on Reddit: "As a subscriber, I am livid! What is the point of subscribing to a hard copy paper if they are just going to include AI slop too!?"

He isn't confused about the distinction. He's telling you it was never his job to make it. He bought the masthead. The masthead is the unit of trust, and the paper spent it on a 64-page insert it hadn't read.

Nobody sets out to publish something no one read. It happens because verification is nobody's line item. Every handoff has an owner for cost, an owner for schedule, an owner for format compliance. The question of whether the thing is true belongs to whoever touched it first, and that person is three companies back and has moved on.

Some context worth holding: two months before the reading list ran, the Sun-Times had lost roughly twenty percent of its staff. Thirty people took buyouts, twenty-three of them from the newsroom, saving about $4.2 million a year. I don't raise that to excuse the failure. I raise it because the checking capacity got cut and the page count didn't. That gap is where syndicated filler goes, and it's now a structural feature of the industry rather than an accident at one paper.

This is a shape, not an incident

August 2023. Gannett, the largest local newspaper chain in the country, ran automated high school sports recaps through a tool called LedeAI. The Columbus Dispatch printed this: "The Worthington Christian [[WINNING_TEAM_MASCOT]] defeated the Westerville North [[LOSING_TEAM_MASCOT]] 2-1 in an Ohio boys soccer game on Saturday."

Unfilled template placeholders, in print. The Dispatch's own ethics policy states that AI-generated content "must be verified for accuracy and factuality before used in reporting." The same recaps ran in Louisville, Phoenix, and Milwaukee. The corrections were appended with a note that is almost too on the nose: "This AI-generated story has been updated to correct errors in coding, programming or style."

Earlier that year, CNET published 77 machine-generated stories, several of which required corrections, and reporting found the tool had reproduced existing phrasing closely enough to be called plagiarism.

Different companies, different tools, years apart. The shape doesn't change. Content enters the pipe somewhere upstream and comes out under a name that never inspected it.

The audit almost nobody runs

The useful question is not "where are we using AI." Most organizations can answer that one now, and the answer is reassuring and incomplete. The question is: where does our name go on work we didn't make?

Write the list. Syndicated inserts and licensed content. White-label tools sold under your brand. Agency and freelance deliverables that ship straight to customers. Embedded third party widgets. The vendor chatbot with your logo on it. The co-branded research report. The partner integration your support team can't see into.

Then, for each one, name the last human who looks at it before a customer does. Not the approver of the contract. The reader. If you can't name that person, there isn't one, and you are running the Sun-Times supply chain with better luck so far.

The fix is not a policy banning AI. Buscaglia's tool was incidental; a lazy human researcher would have produced the same list more slowly. The fix is a verification step that belongs to someone, at every point where your credibility gets lent to work you didn't do.

The Sun-Times was right that its newsroom didn't create that reading list. That is precisely what makes it worse. The paper's name went on something nobody in the building had read, and its readers had no way to tell which parts of the paper the promise still covered. You cannot outsource the making and keep the vouching. Whatever your name is printed on, you made.