Anki notes vs cards: why one edit can affect several prompts

Think of the note as the source record containing the fields you maintain. The cards are the questions generated from that record according to its note type. The official getting-started manual explains this relationship. It matters because content editing and reviewing are different actions: a corrected definition can belong to the shared source record, while the forward and reverse questions can demand different kinds of recall. Do not infer that two prompts must have separate content merely because they appear at different moments.

Build a paper sketch before changing a real collection. Draw one box for a vocabulary note with fields for the word, meaning, example sentence, and source. Draw a forward question and a reverse question connected to that box. Label the action each question requires. One may ask for meaning; the other may ask for a word in context. The sketch makes it clear why changing a field can affect several prompts, and why a successful answer in one direction does not prove success in the other.

Imagine a fictional learner who stores a Spanish word, an English gloss, and a sentence from a permitted class handout. The forward card asks for the meaning of the Spanish word. The reverse card shows the English gloss and asks for the Spanish word. The learner then discovers that the gloss is too broad: several Spanish words fit. Editing the shared meaning field may improve one direction while leaving the reverse task ambiguous. The model explains why a shared correction does not automatically produce two well-designed questions.

The learner writes a criterion for each direction. The forward task accepts an explanation of the intended sense. The reverse task needs context that selects the intended word. They inspect both generated prompts after the edit instead of checking only the field editor. This is the practical advantage of distinguishing notes from cards: you can preserve a single traceable source record while evaluating each retrieval task separately. The goal is not maximum prompt multiplication. It is a set of questions whose fronts and answers agree.

A collection can generate more than one review prompt from a source record. Therefore two counts can describe the same material without being equal. Before diagnosing duplicates, identify what the current view is counting. In a mixed collection, some notes may generate a single basic question while others generate several directions or cloze prompts. A numerical difference is not, by itself, evidence that content was imported twice. Inspect a known note and trace its generated questions.

Use a manual count on a tiny sample. Select a few source records whose intended prompt patterns you can explain, then compare the generated questions with that expectation. If one record produces an unexpected prompt, inspect the note type and relevant fields. If two records contain the same content, investigate the import history separately. These are different defects: extra generated prompts versus duplicate source records. A repair that addresses the wrong defect can remove useful study directions while leaving the actual duplication untouched. Keep the sample small enough to reason about without a spreadsheet.

Label your sample with both the source record and the generated task. For example, write vocabulary note A, forward meaning prompt, and reverse production prompt. These labels make a later screenshot or export discussion much clearer than saying the apple card when several prompts share the same word. Precise naming reduces the chance of repairing the wrong object.

A note-type change is a structural operation, so define the outcome first. Write the existing fields, the proposed fields, and the questions you intend to keep. Preserve a recovery copy and use a representative sample. Include material with optional explanations or source references, not only the simplest two-field example. An apparently successful front-and-back preview can hide the loss of a field that was not displayed on either side but still mattered for later verification.

After the test, inspect content and generated tasks separately. Does each field contain the intended information? Does every front ask a clear question? Does the back show the correct answer and relevant context? Are there unexpected empty or redundant prompts? Follow the current editing documentation for the actual operation. Avoid expanding a structural change to the whole collection merely because one familiar card still looks normal. The best test includes the kinds of records most likely to expose a mapping assumption, especially notes with several generated prompts.

When two related prompts appear close together, the first may reveal information needed by the second. That can make the second answer feel easier without testing independent retrieval. Recognizing the shared note helps you identify the relationship. Consider whether the directions represent genuinely different tasks and whether the scheduling and burying choices match your intention. This is distinct from repairing the wording, although both may be needed. A clear question can still be cued by its sibling.

In the vocabulary example, the learner records a reverse failure even after a forward success. They do not declare the record mastered as a single undivided unit. They inspect the reverse context and accepted answer, then decide whether the additional direction serves their course goal. If production is not required, a reverse card may be unnecessary. If production is required, it deserves an appropriate cue and a separate honest response. The source record supports those decisions; it does not tell you that every possible question should be generated.

Text transfer carries fields according to the chosen export and receiving format. It is not equivalent to recreating the native note type and every generated review prompt. Before exporting, decide whether you need a source record or explicit question-and-answer rows. Inspect the actual file in a text or spreadsheet viewer and check a small sample. A field called example may carry context you do not want to discard just because the target app has a simpler two-column interface.

Elly supports a text-content workflow using CSV, TSV, or TXT rather than direct preservation of Anki packages. Native templates, media, scheduling, and review history are separate boundaries. Keep Anki if those behaviors matter and test a permitted source in Elly only for the task you want to compare, such as drafting a clearer explanation. The note/card distinction helps you judge that comparison honestly: moving source words is not the same as moving every question that Anki generated from those words. Verify the receiving questions before relying on them in study.