The Grind’s editors have spent the last week chasing after fake people, after we got inundated with drafts written by Artificial Intelligence (AI). Here’s what happened.
We’ve been preparing to publish a special Food Issue this month. The Grind hasn’t published much about food so far, but we wanted to spotlight Toronto’s amazing food culture and the people behind it. So we put out a call for pitches.
We got over 50 article pitches. We accepted the ones where the writing was clear, the topic was exciting and the angle was true to our priorities. We decided to give new writers a chance — including some writers who didn’t have a portfolio, but who had a really interesting angle.
We greenlit around 24 article pitches. Many of the articles are as short as 100 words, some up to 1,000 words. Drafts started coming in on time (which in itself was kind of weird, but we didn’t want to look that gift horse in its mouth).
First, we noticed that one writer was based outside Canada. That’s not a deal-breaker in itself, but the story was a very local one: they were writing about intergenerational tensions in two immigrant-run restaurants in Toronto’s suburbs. Their draft had direct quotes from the owners and vividly described the interior as being perfumed with turmeric and cardamom, and seeing a “stained handwritten notebook by the counter.” We scheduled a call to find out if they’d actually been there.
At first the writer defended their piece, claiming they’re a travel writer and an “expert researcher.” “I’ve written about places in China even though I’ve never been there,” they told us.
When we couldn’t find any evidence of the restaurants on Google or social media, the writer agreed to email us links and contact information.
But their email turned out to be an admission that “The characters and places in my article are fictional composites […] based on real themes.” We’re pretty sure they were made up by an AI program after feeding it a prompt.
We cancelled their article immediately. At The Grind, we’re committed to keeping AI slop outta here. Only real stories written by real humans, please.
The new reality is that publications are being sent a lot of AI junk, and some of it is getting through. Earlier this spring, a number of U.S. papers published the same summer reading list where more than half the books were hallucinated by AI.
So we began fact-checking every draft earlier than we normally would. We identified seven articles we strongly suspect were written by AI. They all had a similar feel: too neat, too vague. We learned to read the signs: U.S. instead of Canadian spelling, double-barreled article headlines that didn’t quite match the drafts, the same author writing an eloquent pitch and then awkward follow-up emails, and drafts riddled with em-dashes. (What normal person even knows how to type an em-dash?)
When pushed to provide contact info for their sources, some of the writers who didn’t immediately ghost us went pretty far out of their way to trick us. Two gave us bogus 555-phone numbers — like they use in movies — for the people they had supposedly interviewed, and addresses of places that didn’t exist.
One person gave Fernando a website for a labour organizer’s fictional workplace. The website shows an “under construction” page, but when Fernando called the listed phone number, the person on the other end had no idea what the organization was or why we were disturbing him.
Another gave Saima an email address for an “expert source” who had no digital footprint and whose organization also didn’t seem to exist. The “source” initially answered emails asking to set up a fact-checking video call — which she never showed up to. She stopped answering follow-up emails.
In short, we didn’t vet our authors thoroughly enough. We didn’t scrutinize every new author as closely as we should have.
Most of the time, we work with trusted writers on topics we cover over and over: city council, public transit, Palestine, music, theatre and so on. Many of these writers are people we’ve met in real life, people we’ve worked with for years across different publications.
The Grind is made by a team of volunteers, some of us jenga-ing emails into the bleary-eyed hours around gig work or at the end of paid workdays. Not too long ago, tech billionaires kicked the knees out from under journalism’s old revenue model. Those same tech overlords are now trying to profit off of pushing AI into every corner of life. Meanwhile, journalism jobs keep getting cut, and the few of us who survive see our workloads grow to unmanageable proportions while our pay stagnates.
Because of this AI debacle, we’ll be delaying the print date for The Grind’s food issue by a week. During that time, we’re fact-checking our stories, and we’ve reached out to trusted people to write new articles to replace the AI slop. The food issue will start being distributed across Toronto on Friday, June 20.
As more digitally-fabricated garbage gets sent to newsrooms, we are strengthening our processes to catch it. That includes vetting writers, fact-checking drafts early on, and using AI detection software.
On a broader scale, this experience has been a reminder of the importance of adequately-funded newsrooms. Traditionally, fact-checkers have been interns or junior reporters. Those roles, ironically enough, were some of the first to go when media outlets faced cuts. It means most newsrooms are woefully unprepared to stop the avalanche of AI slop that’s coming.
You can support real journalism — committed to truth and written by humans — by supporting The Grind with a donation of any amount.
AI is yet another challenge, but we’ll handle it.
This article appeared in the 2025 Summer issue.