Reading time: 4 minutes

Today in Brief

A few weeks ago, a journalist sat down in a dental chair.

Not just any journalist. Joanna Stern spent years as the Wall Street Journal's tech columnist. She just spent a full year letting AI run her life and tested more than a hundred AI tools for a book about it.

The dentist pulls up her x-rays. There's an AI on the screen (it's Pearl). It lights up her scan.

The verdict: she needs periodontal treatment. Four sessions, thousands of dollars. Maybe not covered by insurance…

She said that she never had a problem with her teeth.

So she does what any smart person does and gets a second opinion. Then a third. Then a fourth.

All of them disagree. "The AI is saying that, but honestly, it's really not that bad. Better home care and you'll be fine."

She never did the treatment.

And then she wrote about it, in a book a lot of people are reading right now.

The internet did what the internet does. The headline became: "Dentists Are Using AI to Scare Patients Into Unnecessary Work."

I wrote a quick reaction to it on LinkedIn. Tens of thousands of people saw it.

But there's a part of this story almost nobody is talking about. And it's the part that actually matters for your practice.

Here's what you absolutely need to know, and what you should do about it.

(TL;DR at the end)

Here’s the part the headline burried

Because here's the thing about that same journalist, in that same book.

A few chapters earlier, she sat in a different chair. For a mammogram. With a different AI reading the images.

And she loved it.

She called it scrupulous. Tireless. The kind of second pair of eyes that catches what a human on their tenth scan of the day might miss.

Same person, same year and same kind of technology (a computer trained to spot patterns in medical images).

Hero in radiology. Villain in the dental chair.

Stop on that for a second.

If the technology is basically the same in both rooms, then the technology is not the problem.

Something around it is.

So what’s different between the two rooms

The big one is this: who gets paid.

In radiology, the person who finds the problem is almost never the person who profits from fixing it. The radiologist reads the scan and hands it off. The treatment happens somewhere else, with someone else's budget.

In your world, it's all one chair. The same person finds the bone loss, recommends the treatment, presents the quote, and does the work. Sometimes with a monthly production target on the wall behind them.

The AI didn't create that setup. It just dropped a very convincing new exhibit right into the middle of it.

There's a second reason too, and it's older than any of these tools.

A quick trip back to 1997

Reader's Digest ran an investigation. A reporter carried his own x-rays to 50 dentists across 28 states.

The treatment plans for the same mouth ranged from basically nothing to around $30,000.

No AI anywhere!

That's the part everyone quotes.

Here's the part almost nobody does: 15 of those 50 dentists missed his one real problem completely.

Read that again.

The pre-AI world was never honest dentists versus greedy ones. It was wild variance in both directions. Some invented thousands in work that wasn't there. Others missed disease that was.

The diagnosis itself was the shaky part.

Which is the entire case for a good detection tool, not against it. A consistent second reader is exactly what catches the tooth those 15 dentists missed. It's also what makes the $30,000 plan look like the outlier it is.

The catch? That same detector, run through a production quota, becomes the thing that surfaces four sessions of perio on a patient who just needed to floss better.

The tool is neutral but the dashboard around it is not.

One line worth knowing

The FDA cleared these tools to do exactly one job: detect and measure. Caries. Calculus. Bone loss, down to the millimeter.

That's the whole cleared function.

"3.2mm of bone loss on tooth #19" is a measurement.

"Four sessions, several thousand dollars, let's book the first one today" is a human decision. Outside anything a regulator ever signed off on.

The machine detects but the dentist always decides.

Which is why the most damning line in the whole story isn't about AI at all. It's a manager grilling the staff:

"Why didn't you sell the periodontal treatment?"

That sentence works perfectly with zero AI in the room.

So what do you actually do with this?

Here's the move. And it falls straight out of everything above.

The FDA drew a line: the machine detects, the human decides. Your job in the chair is to make that line audible. Say it out loud, in an order the patient can actually follow.

Three beats.

1. Show before you speak. Turn the screen toward them and et them look for two seconds before you say a word. The image earns a kind of trust your sentence never can.

2. Split "what I see" from "what I'd do." Two separate sentences. "Here's what the scan shows: bone loss here, here, and here." Pause. "Here's what I'd suggest, and here's why." You're doing out loud exactly what the regulator does on paper. You separate the finding from the decision. And the patient stops feeling like a sale is being laundered through a machine.

3. Always name the "let's watch it" option. Even when you'd treat. Especially then.

Now here's the part worth underlining, because it was hiding in the story everyone shared.

Remember the four dentists who disagreed with the AI? The ones Stern actually trusted?

Look at what they said. "It's really not that bad. With better home care, it can get better."

They offered her the watch and wait path!

That's the whole move. The "villain" article you saw on LinkedIn literally contains the playbook for the dentist who wins. The ones who kept her trust were simply the ones who gave her a door that wasn't "pay me now."

Same finding and same AI on the screen. Two completely different patients walking out!

Which one are you building?

There are two versions of AI dentistry. Patients can already tell them apart.

One sends them out to collect four second opinions and write a bestseller about it.

The other turns the same screen into the most honest conversation they've ever had in a dental chair.

And here's what my co-founder (a dentist, 20 years in the chair) will tell you:

❝

Patients say yes to complex treatment far more often when they genuinely understand it, not when they feel cornered into it.

The transparency play isn't the soft option. It's the one that gets cases accepted.

Radiology's culture made its AI a hero. The finder wasn't the seller, and trust was already high.

Dentistry gets to decide what its own AI becomes. That's being settled right now, one operatory at a time.

The version in the headline is one option.

It's not the only one. And it's definitely not the one worth building.

TL;DR

A well-known tech journalist was told by an AI-assisted dentist that she needed thousands in perio work. Four other dentists disagreed. The headline became "AI helps dentists scam patients."
Same journalist, same book: she loved the AI reading her mammogram. Same kind of tech, opposite verdict. So the problem isn't the AI.
The real difference is incentives. In radiology, the finder isn't the seller. In dentistry, one chair detects, recommends, sells, and treats.
Dentistry's trust problem is old. In 1997, 50 dentists quoted the same mouth from roughly $0 to $30k, and 15 missed the one real problem. Detection AI is a fix for that variance, not the cause of it.
The FDA cleared these tools to detect and measure only. The leap to "four sessions, thousands of dollars" is 100% a human call.
What to do: make the detect/decide line audible. Show the image first. Say what you see and what you'd do as two separate sentences. Always name the watch and wait option. The dentists who kept Stern's trust did exactly that.