How to Turn First Aid for USMLE Into Flashcards Automatically (2026 Guide)

Making and reformatting Anki cards by hand can consume a large part of an M2 study schedule: highlighting, screenshotting, and pasting into the card editor instead of reviewing the material.

If you live in First Aid (and you do), there's a better way. This guide walks through how to turn First Aid for the USMLE Step 1 into a working Anki-ready deck using Scholarly's AI flashcard generator, how it stacks up against AnKing v12, and how to keep your reviews from melting your brain in the run-up to dedicated.

Why First Aid is still the bible in 2026

Despite AMBOSS, Boards & Beyond, Pathoma, Sketchy, and a UWorld bank that now writes its own explanations longer than the questions, First Aid hasn't lost its grip. The 2026 edition still clocks in north of a thousand pages and remains the only single source that maps almost cleanly to the Step 1 content outline.

Most students treat FA as an index. You annotate it from your videos (BnB, Pathoma, Sketchy), then on second pass you'd ideally encode every high-yield bullet into Anki. That second pass is where the wheels come off.

The manual workflow that eats your M2 year

Here's the typical loop:

1. Read a FA page on, say, glycogen storage diseases.
2. Highlight the high-yield bits.
3. Screenshot the table for image occlusion.
4. Open Anki, create new card, type cloze deletions one by one.
5. Tag it (#step1::biochem::glycogen), add the screenshot, hit save.
6. Repeat for the next bullet. And the next. And the next.

Doing that manually across a full chapter adds up quickly, especially when many of the facts may already exist in AnKing and you do not want to create duplicates.

This is the time-sink Scholarly's PDF to flashcards tool is designed to reduce. If you need Anki cloze syntax specifically, the separate cloze deletion generator can draft cloze text for you to review and paste into Anki.

The AI workflow: First Aid PDF to a reviewable deck

The new flow looks like this:

1. Upload the relevant First Aid chapter PDF (or the whole book) to Scholarly.
2. Choose how many standard question-and-answer cards you want.
3. Let the model parse the chapter, extract the high-yield bullets, and generate cards.
4. Spot-check the output. Edit anything weak.
5. If you are on a paid plan, export the deck as .apkg and import it into Anki.

The model drafts the cards so your time can go into checking accuracy, editing weak questions, and removing low-value material instead of typing every card from scratch.

If you've used the medical flashcards guide for MCAT and USMLE prep, the same idea applies: active recall and spaced repetition are the only two things that actually move the needle. The bottleneck is never "did I review enough" — it's "did I have a card for this in the first place."

How this compares to AnKing, Bros, and Brosencephalon

Let's be honest about the existing deck landscape.

AnKing v12 (the Overhaul successor) is the consensus default. It's enormous — north of 35,000 cards — and tagged to BnB, Pathoma, Sketchy, UWorld, AMBOSS, and FA section numbers. It's the gold standard. It's also overkill if you're trying to get through dedicated in six weeks, and it covers plenty of content beyond what's actually in your edition of FA.

Bros deck / Brosencephalon is the legacy option. Cleaner, smaller, less maintained. Still beloved by some, but the tagging hasn't kept pace with the 2026 FA layout.

Lightyear and Duke Pathoma are video-specific decks — great companions, not replacements.

Scholarly-generated cards from First Aid are something else: they're your FA, not a community average. If your school crammed extra GI pharm into M2, or your annotated FA has scribbles around the immuno tables, the AI cards reflect that. They're meant to slot in alongside AnKing, not replace it.

The honest framing: use AnKing as your core pre-made deck, and use Scholarly to generate supplemental decks from FA pages where AnKing feels thin, from your annotated PDFs, and from any other source (lecture slides, AMBOSS articles, Sketchy notes) you want carded.

Step-by-step: turning a First Aid chapter into a deck

Let's walk through the actual workflow. We'll use the Biochemistry chapter as the example because it's the one most students hate carding manually.

1. Get a clean PDF of the chapter

If you own First Aid digitally (USMLE-Rx, Inkling, or your annotated iPad copy), export the chapter pages as a PDF. Aim for 30-80 pages — small enough to process fast, large enough that you're not re-uploading every twenty minutes.

If your copy is heavily annotated with handwriting, flatten the annotations before exporting. Clean printed text and legible diagrams work best; messy handwriting in the margins can confuse extraction.

2. Upload to the PDF-to-flashcards tool

Drop the PDF into the PDF to flashcards tool. Set the target card count — for a 50-page FA chapter, somewhere between 150 and 300 cards is realistic. Past that you're generating filler.

3. Generate standard recall cards

Scholarly's deck generator produces standard question-and-answer cards. First Aid's concise bullets work well as source material for direct recall questions, while tables and diagrams can provide visual context for questions about relationships, pathways, and labels.

It does not currently generate cloze note types or masked image-occlusion regions. If you need those formats, create them separately in Anki after reviewing the generated Q&A deck.

4. Review and prune

This is the part people skip and shouldn't. Spend 15-20 minutes reading the generated cards. Look for:

• Cards that ask for trivia the NBME doesn't test (random eponyms, obscure ratios).
• Questions where the wording gives away the answer.
• Duplicates of cards you already have in AnKing.

Delete aggressively. A leaner deck you actually finish beats a comprehensive deck you abandon by week three.

5. Export to Anki

On a paid plan, hit export, download the .apkg, and import it into Anki via File → Import. The deck arrives as new basic question-and-answer notes. Scholarly tags, citations, and review history are not included, so add any Anki organization you need after import.

"But doesn't AI miss high-yield content?"

Yes, sometimes. Here's the honest answer.

The model is good at extracting facts that are explicitly in the source. It's less good at knowing which of those facts are NBME-favorite testing points versus background context. That's a learned skill from doing hundreds of UWorld questions, not something a model picks up from FA alone.

Two ways to handle this:

1. Spot-check against AnKing tags. If you're already running AnKing, the FA-tagged subset is your high-yield filter. Cross-reference the cards Scholarly generated with AnKing's FA-tagged cards on the same topic. If Scholarly produced something AnKing skipped, ask yourself: is this actually high-yield, or did AnKing skip it for a reason?

2. Lean on UWorld. When you miss a UWorld question, generate a card from the explanation using the same tool. Those cards are the highest-yield cards in your deck because they're attached to a tested concept by definition.

The combination — AI for volume, UWorld + spot-checks for prioritization — is how you get the best of both worlds without spending 400 hours typing.

Spaced repetition cadence for Step 1

Quick math on review load, because this is where decks die.

Anki's classic SM-2 algorithm and the newer FSRS scheduler both follow the same principle: review intervals stretch as you get cards right. FSRS, which became Anki's default in version 23.10, uses your own review history to time each card rather than applying a fixed heuristic, so it generally schedules fewer reviews for the same retention target.

Practical numbers if you're starting cards 12 months out from Step 1:

• 25 new cards/day → ~150-200 reviews/day at steady state on FSRS, ~90 min total.
• 50 new cards/day → ~300-400 reviews/day, ~3 hours.
• 100 new cards/day → don't.

If you're inside dedicated (6 weeks out), stop adding new cards from FA. Drill what you have. The cards you make today won't get enough reps to be useful by test day.

When to use AI cards alongside AnKing vs replacing it

A simple decision tree:

• You haven't started a pre-made deck yet and you're more than 9 months out: Use AnKing as your base. Generate Scholarly cards from FA for topics where AnKing feels thin (often: rare biochem pathways, classical music of pharmacology side effects, some immuno).
• You're 4-9 months out and behind on AnKing: Don't start fresh. Use Scholarly to generate a focused FA deck (300-500 cards covering only your weakest organ systems) and run it in parallel with whatever AnKing subset you can finish.
• You're inside dedicated: Generate cards only from UWorld mistakes and FA pages you keep forgetting. No new big decks.

Exporting to Anki: the actual file workflow

On paid plans, Scholarly exports a standard .apkg containing basic question-and-answer notes. Import is straightforward, but the export does not preserve Scholarly tags, citations, or review history. A few practical tips:

• Tag after import. Add #FA::biochem::glycogen-style hierarchical tags in Anki if you want to filter by topic during dedicated.
• Don't merge decks blindly. Keep your AI-generated FA deck as a separate top-level deck (Scholarly::FA::Biochem) rather than dumping cards into AnKing. If a card turns out to be low-yield, you want to suspend the whole subdeck, not hunt through 35k cards.
• Sync via AnkiWeb early. Sync after every import. If your import goes sideways you can roll back from the web copy.

FAQs

Does Scholarly export to .apkg? Yes, on paid plans. A generated deck exports as a standard Anki .apkg file containing basic question-and-answer notes. Scholarly tags, citations, and review history are not included.

Can I use this alongside AnKing instead of replacing it? That's the recommended workflow. AnKing handles the breadth; Scholarly handles depth on the specific FA pages, lecture slides, or AMBOSS articles where you want denser coverage. Keep them as separate subdecks so you can suspend independently.

Does it generate image occlusion cards? No. Tables and diagrams can be used as source material for standard question-and-answer cards, but Scholarly does not currently create pre-marked image-occlusion regions or export an image-occlusion note type.

What about image quality on FA tables? Legible source pages give the model better context. If a scan or photograph is blurry, crop or rescan it before generating cards; the exported Anki deck is not a table-screenshot or image-occlusion export.

How many cards per day will I actually have to do? Rough estimate: a 1,500-card FA-derived deck, started 9 months out, with FSRS scheduling and 92% retention target, settles around 120-180 daily reviews. Combined with AnKing's FA-tagged subset (~6,000 active cards), expect 250-400 total reviews/day during M2. That's roughly 90-120 minutes if you're not getting buried in lapses.

Is there a free version? Yes. The free plan includes limited AI creation so you can try the workflow. Paid plans add higher generation and upload limits and include Anki export.

TL;DR

You do not have to type every recall question from scratch. Upload a First Aid chapter to Scholarly's PDF-to-flashcards tool, generate a question-and-answer deck, review it against the source and AnKing's FA tags, and use a paid plan to export the deck to .apkg if you want to study in Anki.