When ChatGPT Signs a Real Cartoonist's Name, Is That a False Citation?

When ChatGPT Signs a Real Cartoonist's Name, Is That a False Citation?

Yes, in every way that matters for citation tracking. When ChatGPT puts a real New Yorker cartoonist's signature on a cartoon that person never drew, it is crediting a source that did not produce the work. Nieman Lab documented more than 15 artists affected this way. The same failure shows up when answer engines credit brands and publishers with claims they never made.

What exactly went wrong with the New Yorker-style cartoons?

ChatGPT generated single-panel cartoons in The New Yorker's style and, in some cases, signed them with the names of real contributors. Andrew Deck's Nieman Lab investigation, published October 5, 2026, identified more than 15 cartoonists whose names appeared without their permission or payment.

The reported names include:

• Brendan Loper, whose pen name "BLOPER" appeared on a viral cartoon of Dolly Parton and Tim Curry at heaven's gates
• Harry Bliss, Emily Flake, Joe Dator, Peter Vey, Jason Adam Katzenstein, Liza Donnelly and Ellis Rosen
• Pat Byrnes, whose name Nieman Lab found on more than a dozen outputs
• Earlier contributors such as George Booth, Saul Steinberg and Warren Miller

Readable signatures tended to match real contributors, and some images reproduced the magazine's logo. Loper says he did not draw the Parton cartoon.

AV Club, TechSpot, Futurism and AI Weekly followed between October 5 and 7. Late-October newsletters revived the story, with no new statements from OpenAI or Condé Nast.

Why should a signature be treated as a citation?

A signature is a claim about origin, which is exactly what a citation is. When a reader sees "BLOPER" in the corner of a panel, they read it the way they read a footnote: this came from that person. Emily Flake made the comparison directly, likening it to having a quote attributed to her that she never said.

Seen through a citation lens, the cartoon case breaks down into three distinct errors:

1. Fabricated authorship. The output names a creator who did not make it.
2. Borrowed authority. The name and the logo lend credibility the output did not earn.
3. Lost provenance. Nobody outside OpenAI can trace where the signature came from, so the false credit can't be checked against a source.

Each has a parallel in text answers: a publisher named as the origin of a statistic it never printed, or several sources blended with no trail back.

Where did the signatures come from?

Nobody outside OpenAI knows. The company declined to tell Nieman Lab whether its models had ingested the signatures during training. What the reporting does establish:

• The licensing picture is murky. Condé Nast and OpenAI signed a multi-year deal in 2024 with undisclosed terms. A New Yorker spokesperson told Nieman Lab that Condé Nast has never allowed any LLM developer to train on its cartoons, and that using the logo alongside cartoons is not permitted under any LLM deal.
• The artists hold their rights. Nieman Lab reviewed templated contributor contracts and found the work is not work for hire, and none of the contracts allow licensing for AI training.
• Scraping is a theory, not a finding. Nieman Lab raised the possibility that the signatures came from images scraped from the open web, citing court filings in The New York Times v. OpenAI that describe paywall circumvention.

That gap is the provenance problem in miniature: if the path from source to output can't be inspected, nobody can audit the credit.

How did OpenAI respond, and did it fix the credit?

OpenAI's response was partial. A spokesperson thanked users for flagging bugs. After Nieman Lab reached out, some prompts started returning a guardrail warning about similarity to third-party content. At the time of publication, though, ChatGPT still sometimes signed generic cartoons with real names.

The legal path is unclear too. Cornell law professor James Grimmelmann told Nieman Lab that attribution plays only a small role in fair use analysis. Right-of-publicity claims usually require commercial use, and Nieman Lab found no sign the images were sold.

So the people who were miscredited can't count on the law or the vendor to correct the record. They have to find the errors themselves.

What does this mean for brands and publishers tracking AI citations?

It means a citation count alone tells you little. If you only track mentions, a false credit looks like a win.

Misattribution patterns worth watching for:

• A model credits your brand with a statistic, quote or study you never published.
• Your original research gets attributed to a competitor or an aggregator.
• Your logo shows up on generated images or "reports" you didn't make.
• Creators get linked to your brand when you've never worked with them.

Publishers face a version of the same risk when answers absorb their reporting and credit it loosely, or not at all.

How can you detect misattribution in AI answers?

You detect it by checking what models credit to you against what you actually published, on a regular schedule. A practical routine:

1. Build an attribution prompt set. Write prompts that ask ChatGPT, Gemini, Perplexity and Claude who said, found or made the things in your category. Run them monthly and log the answers. Model Share tracks how often and how accurately models mention you compared with competitors.
2. Separate correct citations from false ones. Tag each mention as accurate, wrong, or credited to someone else. That's the difference between a visibility metric and a citation-integrity metric.
3. Verify creator and partner links. Creator Audit helps confirm which creators are actually tied to your brand, so invented associations stand out.
4. Make provenance easy to read. Dated press releases, clear authorship, and structured product facts give models something correct to cite. The GEO guide covers how to structure that material.
5. Document and escalate. Keep timestamped screenshots of false credits and report them. The cartoonists' experience shows that notifying the vendor can prompt a fix, even an incomplete one.

For the full story timeline, artist list and legal background, see MarkGrid's full breakdown of the fake New Yorker cartoons.