


Source: 24.7M public reviews of US dental practices, 2021 to Sept 2026. Made with Python and DuckDB.
A review counts if it has an em dash or one of 53 phrases that barely existed in dental reviews before 2022. I also had random samples read blind with the dates hidden: 0 of 150 from 2021 looked AI-written, 14 of 150 recent ones did.
More charts and the method: https://www.toothlocal.com/reports/ai-reviews-2026/
—
happylytical
25 comments
Clearly the integrity of negative online dentist reviews is the single most important benchmark to improve upon
9 out of 10 dentists recommend the use of AI
if there’s any legit use for LLMs it’s “write this email I don’t give a fuck about”. this is fine
So, are people just getting AI to write their negative reviews or are we discovering a review war between different dentists?
Are these reviews about the guitar collection of different dentists?
I don’t doubt there is an influx of AI written comments, but the same way that AI detection tools are not reliable, you cannot reliably tell if a review is AI.
You address that yourself and you “solve” that issue by asking another AI model if the review is AI? What kind of scientific method is that?
Unless there aren’t any mathematically deterministic markers in the text (that the generative AI model put in theirselves) you are basically just strong-guessing.
Google reviews have been inauthentic for a long time. Like they have been bots with a template or fake accounts run by people out of the country, etc. No surprise that the same people coordinating this have moved to LLMs. It’s frustrating because Google often has enough info on folks to tell who is authentic, but they don’t do anything to clean the platform because they want to claim how hundreds of real people are writing reviews to random local businesses.
the AI detection algorithm is questionable at best
I’m more interested in how much more negative (or not) the reviews and responses to reviews have become since private equity took over dentistry around the same time.
Health care provider here. My office manager got duped into scheduling me for a 40 minute interview with WebMd saying I was going to be profiled as “an exceptional provider”. Which of course was a sales call. I took the call and started to apologize for wasting his time but he cuts me off and says, “Listen _____, LISTEN LISTEN!!TO ME we will use AI to make sure you have at least five good reviews…” i said no thank you and he calls me back from another number. I push it to VM. He calls again. And again. I let it ring and go to VM. He then calls me from a number with a local area code. I answer thinking there is no way WebMD would actually be harassing me and expect me to now pay for anything shitty service they have and ” _____ hey listen, _____ LISTEN TO ME,”
Now I got negative AI reviews on my Google business listing. It’s taken me 12 years to get 16 unsolicited good reviews. Now that’s trashed.
There is literally no way to get a hold of anybody at webmd. They have a single line that goes to voicemail that they never return calls from. My biz attorney wants a 5k retainer to begin sparring with them. I can’t do it right now. Not with Medicare cuts coming inbound.
Data is depressing
So hard to be a human em dash user in this sick world
This ugly ass website burnt my retinas
What exactly is an AI tell? Like, how do you gauge if a phrase “barely existed before 2022?” And then every use of a dash is just assumed to be AI?
I don’t think that the amount rising by 12x is that crazy, when you consider how much more common AI is now than 4 years ago.
I love this sub sometimes. This is so random and interesting.
Surprised so many people are going to angry dentists in the first place.
All angry dentist reviews get a pass forever.
I work in digital. It’s both, but the review war thing is definitely more of a slimey black hat play that is definitely rarer to hear about.
Sounds like a plot by big dentist to discredit their critics.
> It’s not that he pulled out the wrong tooth — that’s human error and could happen anytime — it’s that he didn’t comment on my new moustache.
On the flip side, I went to a new dentist with glowing reviews, but received the hard sell for extensive treatments I didn’t need at all, confirmed by 2nd opinion. I went back and took a harder look at the reviews and it was pretty clear most of the positive ones were ai, and the goal seemed to be to drown out the negative reviews from real people complaining about shady practices.
What tool you used to make these images. It looks great
Every time someone points out em dashes I like to point out that Microsoft products have been “correcting” standard dashes and hyphens to em dashes for like 15 years. I know because it’s mildly infuriating to have to go fix it or turn off the autoformat option every time I use a different pc.
Why wouldn’t the second slide be % of dentist responses that started sounding like AI? And the “where” chart is a heatmap of population density.
The map is pretty from a distance, and it does help with the fact that states vary so much in size, but I find it funny how messed up some of them get. Like look at where North Carolina ended up.
Comments are closed.