Table 2.

Thematic analysis of interviews with themes based on the structure of the technology acceptance model (TAM) and subthemes mapped to benefits, facilitators, and barriers with example quotes (N=15).

TAM theme and subtheme			Prevalence, n (%)		Aim mapping		Example quote
Actual system use
	Topic summarization	15 (100)		Benefit		“Instead of getting a textbook, I can ask ChatGPT to summarise something for me in X words and read it under a minute or two.”
	Google or search engine replacement	13 (86.7)		Benefit		“It’s just better than Googling it.”
	Quiz or exam question creation	6 (40)		Benefit and facilitator		“I give it a topic...and say ‘make some test questions,’ and I specify the difficulty and style as well.”
	History taking role-play	5 (33.3)		Benefit and facilitator		“It’s a nice environment to test out sentences or questions before a real patient.”
	Email drafting or administrative tasks	3 (20)		Benefit		“I’ve used it for writing emails and admin tasks like that.”
	Article or research summary	7 (46.7)		Benefit		“It enabled me to spend less time reading tons of papers to find an answer.”
Perceived usefulness
	Efficiency and time saving	13 (86.7)		Benefit		“It’s more of a natural conversation as opposed to Google, where you have to scroll through lots of sites.”
	Clarifying difficult concepts	5 (33.3)		Benefit		“If there’s a difficult concept, I ask ChatGPT to summarise it in simple terms.”
	Mnemonic or flash card generation	3 (20)		Benefit		“It can come up with mnemonics and memory aids—though they’re not always very good.”
	Change in educational assessment needs	10 (66.7)		Facilitator and barrier		“If everyone is using ChatGPT to study, assessments may need to be adjusted...”
Perceived ease of use
	Device flexibility	6 (40)		Facilitator		“It fits into my revision routine because I can use it on my phone in clinic.”
	Technical simplicity and low learning curve	7 (46.7)		Facilitator		“You just type your question and it gives you what you want.”
Attitudes toward use
	Overreliance	8 (53.3)		Barrier		“It’s almost like an addiction...outsourcing every little bit of thinking instead of working something out.”
	Collaborative learning via peer demonstration	9 (60)		Facilitator		“After my friend showed me, I started using it for new things.”
	Lack of awareness of large language model features	8 (53.3)		Barrier and facilitator		“That would be useful, but I’ve never used it for that before.”
	Hallucinations or inaccurate answers	15 (100)		Barrier		“I stopped using it...because I found it to be inaccurate, and I don’t want to be learning the wrong things.”
	Generation of fake or nonexistent references	8 (53.3)		Barrier		“It almost makes up sources out of thin air. You paste the reference in a browser, and it doesn’t exist.”
	Difficulty trusting without prior knowledge	9 (60)		Barrier		“I think you need to have a bit of an understanding already to make sure what you’re being told is right.”
	Privacy and data concerns	5 (33.3)		Barrier		“I’m wary of using ChatGPT in research because I don’t understand all the copyright implications.”
	Out-of-context output	6 (40)		Barrier		“Sometimes the information is just out of context, so you have to clarify...get more specific with prompts.”
	Preference for official or older resources	7 (46.7)		Barrier		“I have access to sources which are more reliable than ChatGPT, like older years’ notes or textbooks.”
	Inadequate for guideline or recommendation queries	9 (60)		Barrier		“I don’t think I’d use it for treatment guidelines. Easier to get it from NICE or textbooks.”