If the NYT Can’t Catch AI Slop, You Won’t Either
The New York Times issued a correction over the weekend. As Simon Willison flagged on Saturday, a Times reporter had quoted Canadian Conservative leader Pierre Poilievre calling someone “turncoats.” He never said it. An AI tool summarized his views, used “turncoats” in the summary, and the reporter treated the summary as a verbatim quote. The editors’ note: “The reporter should have checked the accuracy of what the A.I. tool returned.”
If the most-edited paper in the world can fall for that, your part-time staffer running ChatGPT in the back room on a Saturday afternoon will fall for it too. Right now, somewhere in your store, AI is writing things on your behalf — product descriptions, customer reply emails, social captions, holiday card copy, vendor outreach. Most of it is fine. Some of it is making things up that sound real.
AI doesn’t lie. It invents things that sound true. That’s worse, because nothing reads wrong.
Three checks before anything AI-assisted goes live:
1. Verify every date, number, and proper name. “Family-owned since 1978” should match your actual founding year. Hand-stitched in Italy should mean Italy. AI guesses these constantly and confidently.
2. Never let AI invent a quote. If a product description says a customer “raved” about a candle, that customer must exist and must have said it. If it’s real, attribute it. If it isn’t, cut it.
3. Read it all out loud to yourself before publishing anything. Hallucinations look smooth on the page. Hearing the words is what catches them.
It’s really easy to make this mistake. The Times made it. Just take the extra few minutes to get things right — your human review counts.
xoxo,
❤️ AmyFay
What’s happening in the world. What it means for your store.
Monday through Thursday, from AmyFay Chandler. More in Friday Retail Tidbits, every Friday morning.
