ChatGPT or Alma for med school: chatbot helper or tutor that knows what you forgot?
By Roupen Odabashian, MD, FRCPC · Published · Last updated · 14 min read
How we write and check articles: Editorial Policy.

Disclosure: Alma publishes this article and makes one of the study tools it discusses.
If you are a medical student choosing between ChatGPT and Alma, use ChatGPT when you need flexible explanations or brainstorming, and Alma when you need a structured plan that knows what you have already seen and forgotten. ChatGPT answers whatever you ask but does not track your forgetting. Alma builds plans, flashcard schedules and Insights from your own activity to tell you what to study next.
This article focuses on medical students. Alma also serves nursing, NP and PA students, but the examples here are for preclinical courses, Step 1, shelves and Step 2 CK. Alma is our product, so we say plainly where ChatGPT is the better pick.
What is the quick difference between ChatGPT and Alma for med students?
ChatGPT is a general AI assistant that explains, brainstorms and drafts, with voice features in the ChatGPT desktop app for supported paid plans as documented in the ChatGPT Voice docs and pricing page.[1][2] It does not track your study history or schedule review. Alma is a study platform: it reads your own material, writes quizzes and slides, tracks recall with flashcards and quizzes, and uses Insights and Study Schedule to decide what you should do next.
| Feature | ChatGPT (OpenAI) | Alma Study |
|---|---|---|
| Core role | General AI chat assistant across tasks in Chat, Work and Codex[1] | Exam focused study workspace with flashcards, quizzes, Insights and Schedule |
| Study history awareness | ChatGPT Voice uses the current conversation and the selected task or model when you talk to it in the desktop app, and it can resume earlier voice chats, but the documentation describes it in terms of ongoing tasks and conversations rather than long term spaced review.[1] | Tracks flashcard ratings, quiz answers and confidence taps across time inside Alma |
| Voice conversations | ChatGPT Voice is available in the ChatGPT desktop app for Plus, Pro, Business, Edu and Enterprise plans, and Voice in Desktop uses your existing Codex usage budget with costs metered per minute or per credit according to the pricing page.[1][2] | Text tutor bubble on every page, with a separate voice feature where available |
| Pricing model | Free and paid ChatGPT plans with Codex access such as Go, Plus, Pro, Business and Enterprise, each with listed monthly fees and features on the pricing page.[2] | A time limited trial at sign up, then Monthly or Annual plans that differ in AI generation allowances |
| Medical evidence access | Not specified in the pricing or voice documentation[1][2] | Looks up PubMed for evidence sensitive questions, cites returned papers |
| Study planning | The ChatGPT Voice and pricing documentation focuses on conversations, tasks, usage limits and pricing, and does not describe a calendar style exam plan builder.[1][2] | Study Schedule builds a backward plan from your exam dates with editable blocks |
| Best at | Explaining concepts, brainstorming mnemonics, talking through ideas by text or voice | Telling you what to review today from your own material and how to space that work |
Alma is our product, so we say plainly where ChatGPT is the better pick.
How do ChatGPT and Alma actually help you learn, not just feel productive?
Both tools sit on top of learning science: retrieval practice and spaced repetition are high utility strategies that reliably improve learning across ages and materials.[3][4] ChatGPT can help you run retrieval sessions if you prompt it, but it does not schedule or track them for you. Alma bakes retrieval and spacing into flashcards, quizzes, Insights and the Study Schedule.
Large reviews of learning techniques have found that practice testing and distributed practice have high utility for improving student performance across many outcomes.[3][4] Rereading and highlighting, which feel comfortable, show much lower and inconsistent benefits.[4] Many of us slide back into rereading First Aid or lecture notes when stressed.
Metacognitive calibration also matters: we are often overconfident in what we think we know. The Dunlosky review notes that some techniques, such as practice testing and distributed practice, were rated as high utility in part because they help students monitor and improve their learning across different materials and tasks.[3][4] Alma Insights uses Guess or Sure taps and accuracy data to flag where your confidence does not match reality. ChatGPT will usually not do this unless you set up elaborate prompting yourself.
Which tool actually tells you what to study next and how to space it?
If your main problem is deciding what to study each day and how to balance new content with review, Alma is the stronger choice. ChatGPT can draft plans on request, but it does not see your past card ratings or quiz misses, and it cannot automatically adapt a calendar as your recall changes over time.
Alma Study Schedule lets you enter exam dates, weekly hours, days off and multiple exams, then builds a backward plan from your nearest exam. It can handle several exams on the same calendar and exports to your calendar app via an .ics file. The plan is built from your own weak topics, hours and days off. You can use it as a full plan or as a calendar that you edit by hand.
Alma Insights sits on top of that plan and your activity. It takes your flashcard Again, Hard, Good, Easy ratings, Learning Hub quiz answers, recall tests and Guess or Sure confidence taps and turns them into topic mastery estimates. Topics are labeled Needs work, Getting there, Solid or Too early to judge, with internal cutoffs based on your estimated mastery and a minimum number of answers before a topic is judged. The Readiness tile appears once enough topics can be judged, and the Do this now card points you to a concrete weak topic with a button to review cards.
Sections like About to forget, Is your studying working, Does feeling sure match being right and Alma noticed give you an overview of memory decay, trends and misconceptions, and they feed you back into the right cards or quizzes. For a detailed breakdown of how this works, see Alma’s own explainer on Insights at this article.
ChatGPT can be asked for a study plan each day; for example, you might tell it how much time you have and when your exam is and ask it to make a plan for today. It will write one, but it is operating on what you type, not on stored data about what you reviewed yesterday or last week.
How do ChatGPT and Alma differ for practice questions and flashcards?
ChatGPT is excellent at drafting custom practice questions, explanations and mnemonics on demand. It can walk you through a case orally with ChatGPT Voice in the desktop app, which supports natural turn taking so that you can interrupt during a response, ask a follow up or change direction.[1] It does not give you a persistent question bank tied to your own files, and it does not space questions automatically over months. Alma centers your own content and uses FSRS for long term flashcard scheduling.
Alma flashcards use an FSRS scheduler. FSRS is an open spaced repetition algorithm that combines machine learning with memory formulas and allows a desired retention in the 0.70 to 0.97 range.[5] Cards are graded Again, Hard, Good, Easy, and intervals adjust accordingly. A failed card can return later in the same session, and you can undo a mistaken rating. Alma shows the interval each rating will give, and it support features like image occlusion for anatomy and histology.
ChatGPT does not have a dedicated FSRS powered flashcard system documented in the ChatGPT Voice or pricing sources we have, which focus on conversations, models and usage limits rather than long term flashcard scheduling.[1][2] You can certainly ask it to quiz you, or to turn a block of text into question and answer pairs, then paste those into Alma or another flashcard tool. That is a strong pairing: let ChatGPT help you phrase prompts and examples, then use Alma to schedule and track your recall over time, as discussed in this FSRS article.
Is ChatGPT enough for Step 1 and Step 2 CK, or do you need something like Alma?
ChatGPT alone can help you understand pathology pathways, differential diagnoses and mechanisms, especially if you iterate on prompts and ask it to re explain concepts in different ways. However, the evidence base for learning tells us that structured retrieval and spacing are more important than endlessly re explaining content.[3][4] Alma focuses on those.
Research in cognitive and educational psychology has found that practice tests and distributed practice can be especially effective learning techniques that improve student performance across many materials and criterion tasks.<sup>[3][4] Alma makes it easy to build exam style quizzes from your own notes and to push that information into spaced flashcards and a Schedule that counts down to your exam.
For clerkships and Step 2 CK, your problem often shifts from raw facts to application and prioritisation. ChatGPT is very strong at talking through “why not B or C” explanations, generating sample patient conversations, or role playing consults and handoffs. Alma complements this with Practice Patients: simulated stations that take you through history, physical, investigations, diagnosis and plan, with marking at the end. That is closer to OSCE style work, though it is not an official exam simulator.
If your main barrier is understanding (for example, why uncompensated respiratory alkalosis changes ionised calcium), ChatGPT and Alma’s tutor are both helpful. If your main barrier is forgetting or being scattered about what to do each night, a system that tracks topics and makes a concrete Do this now recommendation, like Alma, solves a different and more structural problem.
What does the evidence say about AI, errors and what Alma can and cannot claim?
Any AI that generates medical content, including ChatGPT and Alma’s tutor, can be wrong or incomplete. You should treat them as study partners, not authorities. For safety sensitive questions, cross check against primary resources and guidelines, and do not use them for clinical decision making.
Recent research on large language models in clinical diagnosis suggests that they are not simply regurgitating training data but appear to apply generalizable reasoning across both older and newer cases.[6] That is encouraging, but the same paper and related work do not claim that these models are error free.
Dunlosky and colleagues highlight that practice testing and distributed practice improve learning, but they also stress that these are tools, not guarantees of performance across all contexts.[4] Alma’s design choices, such as using FSRS for flashcards and building Insights on top of retrieval data, align with that evidence base, but Alma itself has not been tested in a trial of exam outcomes.
What the evidence does and does not show: there is strong support for retrieval practice, spacing and calibration as learning strategies.[3][7][4] There is early evidence that large language models can engage in clinical reasoning rather than pure memorization.[6] There is no published study showing that Alma or ChatGPT improve USMLE scores or pass rates, and Alma’s own interface explicitly states that Insights do not predict exam scores or chances. Use these tools to implement evidence based methods, but rely on primary exam resources for content and on official policies for what counts on test day.
How should a med student actually use ChatGPT and Alma together day to day?
The most realistic answer for many students is “use both, but for different jobs”. ChatGPT is a flexible explainer and idea generator. Alma is the memory and planning layer that tracks what you did, what you missed and what you should see next. Here is a concrete way to combine them across a normal study day.
| Task | Use ChatGPT for | Use Alma for |
|---|---|---|
| Before lecture | Quick concept overviews, plain language explanations | Setting exam dates and a high level plan in Study Schedule |
| After lecture | Clarifying confusing slides, drafting initial mnemonics | Uploading slides to Learning Hub, generating quizzes and slides |
| Daily review | Occasional custom questions if you want variety | Running FSRS flashcard sessions and notebook based quizzes |
| Planning | Brainstorming rough weekly goals | Letting Insights and Do this now pick topics, then following the calendar |
| Clinical skills | Role playing patient conversations or oral exams | Practice Patients for structured stations with marking |
A simple workflow could look like this:
- Use ChatGPT to explain a tricky concept from today’s renal lecture until it finally clicks. Ask it for a couple of analogies and a mnemonic.
- Upload the lecture PDF into Alma Learning Hub. Generate a quiz and a slide deck with a number of questions and slides that fit your goals. Skim the slides, then take the quiz.
- Turn missed questions into Alma flashcards. Set aside time to review Due Today cards under the Flashcards screen.
- Open Alma Insights, check the Do this now card and About to forget list, then follow those recommendations for a focused study session.
- Once in a while, adjust your Study Schedule using the exam countdown and your real life time constraints so the plan stays realistic. Alma will adapt blocks around your weak topics.
To keep both tools safe and honest, use prompts like “Explain your reasoning step by step and list any assumptions” and “What sources would you check this against” with either ChatGPT or Alma. When an answer matters, cross check with your primary texts, official question banks and trusted guidelines, as covered in this textbook plus AI guide and this piece on judging AI tools.
Who should primarily use ChatGPT, who should use Alma, and who really needs both?
Some students can get by with ChatGPT plus a traditional flashcard app and a question bank, especially if they already have strong habits for scheduling review and tracking weak areas. Others are better served by a platform that handles planning, recall tracking and content transformation in one place. Your situation matters.
ChatGPT is a better primary tool if you are still early M1, experimenting, and your main need is explanations. It is also a strong companion for clerkships when you want to simulate patient interactions or practice oral reasoning. Its voice mode lets you talk through ideas, coordinate tasks and interrupt naturally in the ChatGPT desktop app.[1]
Most students who are serious about efficiency will likely keep both in their toolkit: ChatGPT for flexible chat and creative prompts, Alma for structured long term memory and planning. Alma is our product, so it is important to say clearly that ChatGPT remains the more general tool, with wider uses outside medicine and a broader ecosystem, while Alma is focused and narrower but deeper for health professions studying.
Frequently asked questions
Can I replace question banks like UWorld with ChatGPT or Alma?
No. Neither uses real exam questions, and both are study aids, not official banks. Use them alongside established question banks and primary texts.
Is ChatGPT safe to use for medical study?
It is useful for explanations and brainstorming, but it can be wrong. Treat it as a tutor, not an authority, and verify important content.
Does Alma guarantee a higher Step 1 or Step 2 CK score?
No. Alma has not been tested in outcome trials and does not predict or guarantee exam scores. It implements evidence based methods but results vary.
Can I use Alma without giving up Anki or my current system?
Yes. You can import existing decks, upload your usual PDFs and use Insights and Schedule on top of your current materials and habits.
Is ChatGPT cheaper than Alma for med students?
ChatGPT and Alma use different pricing models that can change. As of the month you read this, check each provider’s current pricing page.
References
- ChatGPT Voice | ChatGPT Learn
- Pricing | ChatGPT Learn
- Improving Students’ Learning With Effective Learning Techniques: Promising Directions From Cognitive and Educational Psychology – Association for Psychological Science – APS
- Improving Students’ Learning With Effective Learning Techniques - Psychological Science in the Public Interest - APS
- fsrs4anki/docs/tutorial.md at main · open-spaced-repetition/fsrs4anki · GitHub
- Reasoning or reciting? A temporal contamination audit of large language models in clinical medicine - PubMed
- Distributed practice in verbal recall tasks: A review and quantitative synthesis - PubMed