Confident — even when it's wrong
The most dangerous trait of AI isn't that it makes mistakes. Everyone does. The danger is that it's wrong convincingly — in the same calm, certain tone it uses for correct things. If you don't know this, it's easy to believe something that simply isn't true.
This phenomenon has a name: a hallucination.
What a hallucination is
A hallucination is when AI generates information that sounds convincing but is incorrect or entirely made up. It can "invent" a scientific paper with an author and a year, cite a wrong dose, or get a fact wrong with complete confidence.
Why it happens
An LLM doesn't "know" facts the way you do. It predicts the most likely next word based on a huge amount of text. Most of the time that most-likely word is also correct — but not always. When there isn't enough data, AI doesn't say "I don't know"; it fills the gap with something that "sounds right."
Typical mistakes in practice
- Invented sources — it cites a paper or guideline that doesn't exist
- Wrong numbers — doses, concentrations, percentages
- Outdated information — the model won't know the latest guidelines unless it was recently updated
- Mixed-up details — it merges two correct facts into one incorrect one
How to protect yourself
- Treat AI as a bright intern, not a professor — useful for a draft, but verify everything.
- Ask for the source — ask "where did you get this?" and check it yourself.
- Never take doses or drugs "on its word" — compare all medical numbers against an official source.
- Cross-check tools — if two independent tools give the same answer, it's more likely correct (but still no guarantee).
Where to NEVER trust blindly
- Clinical decisions and diagnoses
- Drug doses and interactions
- Legal and regulatory advice
- Anything that directly affects patient safety
Conclusion
AI is an excellent assistant and a poor authority. Use it to speed up the tedious parts — writing, summarizing, ideas — but keep the final word for yourself. Your knowledge and clinical judgment are exactly what no model has.
Want responsible, practical guides to AI in dentistry? Subscribe to the newsletter — I write from practice, with both feet on the ground.
