how it happened
- whoAI writing toolunder a staff byline
- the gapthin editingfluent prose, unchecked facts
- the result41 correctedprogram paused
what everyone assumed
A trusted brand on a piece of writing is a promise about how it was made. The reasonable belief is that content carrying your name has passed through the process that earns the trust: a person who knows the subject, an editor who checks the claims, the ordinary friction that catches mistakes before they reach a reader. When a tool can produce that content faster, it is reasonable to assume the same checks still hold and that scale is the only thing that has changed.
what happened
For roughly two months the program ran, article after article going live under a staff byline, the AI handling the volume the brand had decided it could handle. The pieces were live and indexed and being read as authoritative explanations of personal finance. Nothing in the output announced that the usual checks had thinned, because confident, fluent prose looks the same whether or not its facts are right, and a reader has no way to see the difference from the page.
step by step
- AI-written articles go live under a staff bylinePersonal finance explainers, published at volume for about two months.
- the usual checks thin outFluent prose looks the same whether or not it is right.the assumption breaks here
- January 2023errors are found in a compound interest pieceThat prompts a full audit.
- 41 of 77 articles correctedThe program is paused and disclosure revised.
how it surfaced
In January 2023 the program was reported on, and a closer look at one article about compound interest turned up factual errors, which prompted a full audit. CNET’s editor-in-chief confirmed that corrections had been issued on 41 of the 77 AI-assisted articles, more than half, some of them substantial. The outlet paused the program and revised how it disclosed AI involvement.
Fluent and correct are not the same property, and an AI tool delivers the first reliably while the second still depends on a person checking.
Scale multiplies whatever the process actually does, so a thin review does not get safer when there is more of it, it gets thinner.
picture it at a company like yours
The company here is made up. The other two paths are not: they happened, and each links to the story.
- in this storythe day it mattersA reader spots a basic error and checks the rest.what it costs youHalf your articles need corrections, and readers now doubt the ones that were right.
- or it goes like thisit happenedMoffatt v. Air Canada, 2024howA customer-facing bot promises something your policy does not offer.what leavesmoney you now owewho ends up with itthe customer, backed by a tribunal
- or it goes like thisit happenedDPD chatbot, 2024howA bot is talked into mocking your company and the chat is shared.what leavesyour reputation, in screenshotswho ends up with ita million strangers in a day
sources
your turn
what you can do about it
this week
Take one AI-assisted piece you have already published and check its specific factual claims against a source, and see whether your process would have caught a wrong one.
Someone in your company has used an AI tool to produce something that goes out under your name, a help-center article, a customer email, a published explainer, and it reads well enough that nobody felt the need to check the facts in it. The question the quiet months ask is what your review step actually verifies once the writing is good enough to pass on confidence alone.
go deeper
if you're not sure where to start
Run the offboarding checklist on one leaver. It covers every account and token to close when someone leaves.
open the checklist →