Sooner or later every Anki user faces this: you have a deck you have been studying for months, with scheduling progress you do not want to lose, and new material that needs to go in. A new lecture block, a second resource, a friend's deck. The naive approach — import everything and sort it out later — creates a mess of duplicates that doubles your daily reviews and poisons your retention data.
Here is how to do it cleanly.
Why duplicates are worse than you think
A duplicate is not just an extra card. When two cards test the same fact, your brain learns the answer from the first card and then breezes through the second, marking it "easy." Anki interprets that as genuine mastery and stretches the interval. Your retention statistics become fiction. Worse, duplicates train recognition instead of recall: you start recognizing the card rather than retrieving the fact, which is exactly what fails on exam day when the question is phrased differently.
The manual merge workflow
1. Search before you make. For each new fact, search your existing deck first using the Anki browser. Search the key terms, not the exact phrasing. If a card covers it, do not make a new one. If a card covers it partially, edit the existing card instead.
2. Tag everything you add. Use a consistent tag scheme like lecture::block3::cardio plus a source tag such as added-2026-10. When the exam passes or the material turns out to be low yield, you can find and suspend the whole batch in seconds.
3. Import into a staging deck first. If you are importing someone else's .apkg, import it into a temporary deck, not your main one. Review the imports there, delete or fix what does not belong, then move the keepers over. This quarantine step catches the formatting disasters and the cards that test nothing.
4. Edit in place, do not replace. When new material updates an existing fact — a changed guideline, a corrected value — edit the existing card. Replacing the card resets its scheduling history. Months of spaced-repetition intervals are worth preserving.
The duplicate problem does not scale manually
This workflow is correct and it is exhausting. Searching a 10,000-card deck for every new fact before making a card turns a 30-minute task into a 3-hour one. Most students give up around week three and go back to importing blindly, which is how decks end up with four cards asking for the normal range of potassium.
The fast way: merge with the comparison built in
SlideToAnki's deck improvement handles the merge as a comparison problem. You upload your existing .apkg and the new material — lecture slides, a second deck, whatever the source is — and it works out what is already covered, what is new, and what conflicts. New cards are proposed only for genuine gaps, so duplicates never enter the deck. Conflicting cards are flagged for your review instead of silently coexisting with the old version. You accept, reject, or edit each proposal, and everything new arrives tagged so you can track it.
The result is the merge you would have done manually if you had infinite patience: no duplicates, scheduling preserved, new material tagged and traceable to its source.
Protect the scheduling
Whatever method you use, remember what you are protecting. A mature deck represents hundreds of hours of reviews encoded as scheduling data. Every workflow decision — edit versus replace, suspend versus delete, staging versus direct import — should be judged by whether it preserves that investment. Duplicates and blind imports destroy it quietly. A careful merge protects it.