It Is 2026. Your Patients Are Using ChatGPT. Here Is How to Use It Too.

About this Guide

A patient sits down and says, **"I asked ChatGPT and it said I probably don't need the crown."** Most dentists feel 1 of 2 things. Irritation, or a small flicker of worry. Both show on your face before you say a word.

This guide replaces that feeling with a plan. The plan has 2 halves.

1. **What to do in the room**, in the next 60 seconds, so the patient leaves better informed instead of quietly shopping your diagnosis.

2. **How to use the same tools yourself**, safely, so you understand what your patients are reading and get real value from it.

It is not a guide about whether AI is good or bad. That argument is already over in your operatory. It is a guide about what to do next.

The quick answer

Treat the AI answer the way you would treat advice from the patient's brother in law who "knows a guy." Do not fight it. Welcome it. Ask to see it. Sort it into 3 piles: right, right but not for you, and wrong. Then add the one thing no chatbot can see, which is this patient's mouth.

**KEY TAKEAWAY**

You are not competing with the chatbot on knowledge. You are competing on information it does not have. You have all of it.

---

Why this is now your problem

• **About 1 in 3.** A KFF poll of 1,343 U.S. adults (February 24 to March 2, 2026) found **32 percent** had asked an AI chatbot a health question in the past year. Monthly use went from 17 percent in June 2024 to **29 percent** in May 2026.
• **They still trust you more.** 56 percent feel confident they can tell true from false in a chatbot answer. For a health professional, the figure is **80 percent**. Your word still outranks the machine, if you do not waste it being defensive.
• **The answers are good enough to be dangerous.** Specialists rated roughly 85 to 93 percent of everyday chatbot answers accurate. On harder questions, accuracy dropped to about 70 to 78 percent, and answers changed week to week.

The chatbot is right often enough to be trusted and wrong often enough to matter. That combination is worse than being wrong all the time.

The typical failure is not a wild claim. It is a correct general statement applied to the wrong mouth. "Small cracks can often be monitored" is true. It is not true for a cracked tooth with a fracture line under a large old filling. The chatbot cannot see the filling. You can.

---

The framework: 5 moves

Move What it does

--- ---

**1. Welcome** "Good, I'm glad you looked." Your face decides whether they hear the rest.

**2. See** Ask for the screen and for what they typed. The answer follows the question.

**3. Sort** 3 piles: right, right but not for you, wrong.

**4. Add** Show what the chatbot could not see: this mouth, this x ray, this history.

**5. Send** Hand them a better question to ask next time, and invite them to bring the answer back.

Each move takes under a minute. Skipping the first 2 is where most of these conversations go wrong.

---

Try one line today

Most dentists reach for a version of "you can't believe everything on the internet." It is true, and it ends the conversation. Try this instead.

**SAY THIS**

"Good, I'm glad you looked. Honestly, I'd rather you come in with questions than sit at home worrying. Tell me what it said, and then let's check it against what I can actually see in your mouth."

And the single most useful sentence in the whole guide, for the end of the visit:

**SAY THIS**

"You're going to look this up again tonight, and that's fine. Bring it back and I'll go through it with you."

---

What you can safely put into a chatbot

Every time you open ChatGPT to help with work, you are making one decision: does this task involve a real patient? That answer puts it in 1 of 3 lanes.

Lane Example tasks What to do

--- --- ---

**Green: no real patient in it** Explain a crown in 6th grade language for a handout. Rehearse a hard conversation. Summarize a CE article. See what a patient reads when they search "cracked tooth." Use it freely.

**Yellow: your practice, not your patients** Draft a hygiene job description, a phone script, an office policy, a new patient welcome letter. Let it draft. You edit every line and sign it.

**Red: a real patient or a clinical call** Anything with a name, chart, photo, x ray, insurance ID or date of birth. Any "should I treat this." Stop. Do not paste it into a personal account, free or paid.

**THE RULE THAT MATTERS MOST**

Federal law requires a signed agreement with any company that handles your patients' information. A personal ChatGPT account does not have one, no matter what plan you pay for. No patient goes into a tool that has not signed paper with your practice.

The green lane is where almost all the value sits, and it carries almost no risk. It is also the fastest way to understand what your patients are reading before they walk in.

---

What is in the full guide

• The complete 5 move method, with word for word scripts for each move
• The 3 pile sort, including the pile most dentists get wrong and why it costs them credibility
• 3 real scenarios worked start to finish: "it said I can just watch it," "it told me I need an implant," "it said it could be cancer"
• 5 prompts you can copy this week to see exactly what your patients are seeing
• The vendor question table: what to ask, what a documented answer looks like, and the red flag version
• Red flags in the room and in the practice, plus what a good setup actually looks like
• A 90 day action plan and a check with a professional table
• Full sources from KFF, Pew, the ADA Health Policy Institute and 4 peer reviewed accuracy studies

**12 minutes to read. Free, no strings.**

Guide Information

Chairlink Guides are general educational and informational resources. They are not medical, dental, legal, financial, employment or other professional advice.