They look identical, a text box, a medical question, an intelligent response, and they are opposites. AI search answers your questions. An AI tutor asks you questions. One exists to end your uncertainty as fast as possible; the other exists to make you work through uncertainty because the work is the product. Clinicians who conflate them end up using a search engine as a teacher and wondering why nothing sticks.
Search is reactive; tutoring is proactive
Search waits for you. It has no opinion about what you should know, no memory of what you asked last month, no plan for your development. That neutrality is a feature for decision support: at the point of care you want an instrument, not an agenda. A tutor, by contrast, comes with intent. It tracks what you have struggled with, schedules what you are about to forget, raises difficulty as you improve, and pursues your gaps whether or not you feel like visiting them. Reactivity serves today; proactivity builds tomorrow.
Search solves today's problem; tutoring reduces tomorrow's mistakes
The economic difference is temporal. A search interaction is complete when the answer arrives: the dose is confirmed, the threshold checked, the consultation moves on, and the value is fully consumed in the moment. A tutoring interaction is incomplete by design: the question you were made to answer, the misconception that was named, the retrieval you were forced through, all of it pays out later, in the exam you pass and the 3am decision you get right without looking anything up. Search is expenditure; tutoring is investment.
The mechanism is retrieval, and only one of them has it
Underneath the product distinction sits a cognitive one. Durable learning is built by retrieval, generating knowledge from your own memory, and by the feedback that follows an attempt. Search structurally removes retrieval: the answer arrives before you tried. Tutoring structurally requires it: the good tutor will not tell you until you have committed to a position, precisely because the commitment is the learning event. This is why reading ten fluent answers feels productive and tests poorly, a trap we dissect in Searching Isn't Studying.
How to tell which one you are holding
Marketing now blurs the categories, so use behavioural tells rather than labels. Ask the tool a question you already understand and watch what happens. A search engine answers, fully and immediately, and considers the job done; press it and it will elaborate, but it will never turn the question around. A tutor responds with a question, or at minimum makes you commit before it reveals; it references what you got wrong last session; it schedules a return visit; it gets harder as you get better. Three structural tells settle most cases: does it ask first, does it track you over time, and does it plan your future practice? A tool with none of the three is search, whatever its branding says, and should be used, valued and trusted as search. A tool with all three is making a claim about your development that you can hold it to.
You need both, in the right seats
None of this is an argument against search. A clinician without a fast, cited answer engine is worse at their job this afternoon; a clinician without deliberate practice is worse at their job every year from now on. The mistake is seat assignment: search in the learning seat produces illusions of competence, and tutoring in the point-of-care seat produces dangerous delay. The mature workflow is explicit: search when a patient is waiting, tutor when your development is the task, and never let the comfort of the first substitute for the effort of the second.
The hybrid products muddy the water
The honest complication is that the categories are converging inside single products, and mode boundaries are where confusion breeds. A platform can host a search mode and a tutor mode legitimately; the danger is drift within a session, a search conversation that starts quizzing you when a patient is waiting, or a study session where one frustrated just tell me collapses the tutor into an answer engine and quietly ends the learning. Two design features protect the boundary, and users should insist on them: explicit mode labelling, so you always know which contract you are in, and asymmetric friction, easy to drop from tutoring into answers when clinical need demands, deliberately effortful in the other direction. Where the product does not enforce the boundary, enforce it yourself: name the mode before you type, and hold the session to it.
One platform, both seats, clearly labelled
iatroX is built around exactly this separation. Ask iatroX is the search seat: immediate answers grounded in UK national guidance with citations to source. The Q-bank and Socratic Tutor are the tutoring seat: adaptive questions that force retrieval, spaced repetition that schedules it, and a tutor that asks before it answers and diagnoses the misconception behind your errors. Same platform, same trusted sources, two deliberately different tools, because your Tuesday clinic and your career need different things from AI.