Critical Thinking in the Age of AI: Why Clinicians Still Need to Think

I recently hosted a webinar on diagnosing ADHD in children, and somewhere between the clinical cases a bigger question kept surfacing: in a world where AI can answer almost anything in seconds, what is actually left for us to do? My honest answer is that the most human skill of all — thinking critically — matters more now, not less.

We are living through a strange moment. Information has never been cheaper or faster to get hold of. You can ask a model a clinical question and have a fluent, confident, well-structured answer before you have finished your coffee. That is genuinely useful. But it has quietly changed what is scarce — and therefore what is valuable. The bottleneck is no longer access to information. It is the judgement to know whether that information is right, relevant and safe to act on. That judgement is what competence is made of, and it is the one thing AI cannot simply hand you.

This is the written companion to my video on the same question — the one warning doctors and pharmacists that AI is already here. Watch it first if you like, then read on for the argument in full.

A two-day course won’t make you competent — and neither will AI

People often ask me whether a short course will make them competent in a new clinical area. It is the wrong question, and AI has made the same trap easier to fall into. A two-day course can raise your awareness or start you on a path. It cannot make you competent, because competence is not a certificate you earn once — it is the demonstrated ability to perform a specific task safely, in context, right now, and to keep doing so. (I’ve written separately on what clinical competence really means and how to demonstrate it.)

AI tempts us into exactly the same shortcut. Having instant access to an answer feels like knowing. It isn’t. Having the information in front of you is not the same as being competent to act on it — any more than owning a textbook makes you a clinician. The information was never really the hard part. Knowing what to do with it is.

Why AI makes critical thinking more important, not less

Here is the paradox. The better AI gets at producing answers, the more it matters that a human can tell a good answer from a confident-sounding wrong one. A model will give you a plausible response whether or not it is correct. It does not know what it does not know, and it will not hesitate before telling you something subtly mistaken.

So the skill that protects you is the ability to interrogate the answer: Does this make sense? What is the source? What would have to be true for this to be right? What is it missing? You can only ask those questions if you understand the subject well enough to challenge the machine. You will never out-retrieve AI. The point is not to compete with it on recall — it is to do the thing it cannot do, which is reason about what the information means here, for this patient, in this context.

AI doesn’t reason the way you do

It is worth being clear-eyed about what these tools actually are. They are extraordinarily capable language models — they read, interpret and generate text at a level that would have seemed impossible a few years ago. But they do not reason the way humans reason. They do use a form of logic, but not human logic, and not with human context or accountability. Treating their output as literal, settled truth is a mistake.

Two practical things follow from that. First, how you ask determines what you get. The same question, framed two different ways, produces two different answers. Prompting is a skill. Second, the tool is shaped by how you use it over time. A general consumer chatbot is a blunt instrument; tools that are properly set up and trained on a specific domain behave very differently. Good AI is not free or automatic — the serious tools cost real money and real effort to configure, and even then they are an assistant, not an oracle. Use them intelligently, and verify.

The skill that doesn’t go out of date: learning how to learn

In the webinar, the psychiatrist made a point that stuck with me. The diagnostic criteria he trained on years ago have since changed — and if he hadn’t kept updating himself, he’d still be diagnosing some conditions incorrectly today. That is the whole case for lifelong learning in one sentence. Knowledge has a shelf life. Fields move. What was best practice when you qualified may be wrong now.

This is also why our regulators build in revalidation and continuing professional development: you are not signed off as competent for life. You are expected to keep learning, reflecting and updating. AI does not change that — if anything it raises the bar, because the baseline information is now available to everyone, so what distinguishes you is how well you think, judge and keep growing.

How I’d use AI as a clinician, in practice

1. Treat the output as a draft, not a verdict. Use it to get started, never to finish. The first answer is a hypothesis to test, not a decision to copy.
2. Only trust what you can evaluate. If you don’t know enough to judge whether the answer is right, you don’t know enough to use it. Learn the topic.
3. Check it against authoritative sources. Guidelines, frameworks and the primary evidence — not the model’s say-so.
4. Protect patient data. Never put identifiable information into consumer tools, and stay within your organisation’s information governance rules.
5. Keep the responsibility where it belongs. The accountability for the decision is yours, within your scope of practice and your regulator’s standards.

The bottom line

AI is one of the most powerful tools we have ever been handed, and clinicians should absolutely use it. But it raises the value of being human rather than lowering it. It makes information cheap, which makes judgement priceless. The professionals who do well in this era won’t be the ones who memorised the most or the ones who lean on the model hardest — they’ll be the ones who learned how to think: to question, to verify, to reason in context, and to keep learning when the facts change. That has always been what competence really is. AI has just made it impossible to fake.

Where this goes next

If you want to see what this looks like applied to a real clinical topic, the ADHD webinar that prompted this piece is a good example of judgement over recall — separating ADHD from autism, anxiety and other look-alikes is exactly the kind of reasoning no model can do for you. You can register to watch the full breakdown through MedLRN, where I focus on developing and assessing real competencies rather than handing out certificates.

And if you’re developing your own clinical skills, expanding your scope, or just trying to work out how to use AI well without letting it think for you, get in touch.

Book a session with me

Frequently asked questions