A Word About Medication Dosages and ChatGPT

I had an experience with ChatGPT recently that highlighted something important, and since I’ve seen it happen more than once, I thought it was worth mentioning.

As a veterinarian, I deal with medication dosages constantly. Whether it’s amikacin in ornamental koi, metronidazole for Giardia in dogs, or treatment protocols discussed in human medicine, I’ve spent years reading research papers and learning from respected authors and specialists.

The other day I asked ChatGPT a simple question:

“What is the dosage of amikacin in ornamental carp?”

The answer came back immediately.

It was also dangerously wrong.

Not slightly wrong. Not “close enough.” Wrong enough that I would never have considered using it.

So I responded that the dosage contradicted information published by a recognized authority on the subject.

ChatGPT’s response?

“You’re right. Thank you for catching that.”

It then proceeded to give me a completely different answer.

Now, before anyone sharpens their pitchforks, let me say this: I use ChatGPT all the time. It’s an incredible tool. It can save time, organize information, explain concepts, and sometimes point you toward literature you might not have found otherwise.

But here’s the lesson:

ChatGPT can be wrong.

And sometimes it can be wrong with the confidence of a teenager explaining why he doesn’t need directions.

The issue isn’t that mistakes happen. Every veterinarian, physician, scientist, teacher, and engineer has made mistakes. The issue is that AI often presents both correct and incorrect information with exactly the same level of confidence.

When it comes to life-impacting information—especially medication dosages—healthy skepticism is still your friend.

If an answer doesn’t sound right, challenge it.

Ask it to check again.

Ask where the information came from.

Ask whether there are more recent sources.

Ask what evidence supports the recommendation.

Sometimes it will stand by its answer.

Sometimes it will completely change its answer.

Both responses are useful.

I suspect AI will eventually transform medicine and veterinary medicine in ways we can barely imagine. Some aspects of practice may become highly automated. Some tasks will become easier. Some jobs will change dramatically.

But for now, judgment, experience, and critical thinking remain essential.

So by all means, use ChatGPT.

I certainly do.

Just don’t let a chatbot become the final authority on a medication dosage any more than you’d let your cousin’s neighbor’s barber be the final authority on open-heart surgery.

Trust, but verify.

Especially when somebody’s life is swimming, barking, meowing, or breathing on the other end of the recommendation.