How to export selected Anki notes with a scope you can explain

Write the transfer purpose in one sentence. For example, "Prepare the personally authored definitions from the sampling lecture for a text-mapping test" identifies a source, ownership boundary, and destination task. "Export statistics" does not. State what should remain outside the subset, such as shared material without redistribution permission or image prompts that cannot survive the intended text workflow. The exclusions are part of the scope, not an inconvenience to remove after exporting.

Create a small ledger of known inclusions and near misses. Choose one record that must be present and one similar record that must not. If the selection will be shared, check whether source references or personal notes contain private information. If it will be used privately, preserve enough provenance to verify answers later. A source-based transfer can lose value when references are dropped merely to make the table shorter. Decide the meaning and audience before choosing the export controls. That gives later preview checks a concrete expectation to test.

Build the query in small steps using actual labels and fields from your collection. The official searching manual is the reference for syntax. Inspect the known positive and near-miss examples after each restriction. A plausible result count does not establish the correct membership. If a card is unexpectedly absent, compare its actual metadata and state with the conditions you requested. Do not assume the note was deleted because it fails a restriction irrelevant to the export task.

In the fictional lecture, one note mentions sampling only in a comparison while another belongs to the source group but uses a different term on its front. A broad word search can include the first and miss the intent of the second. The learner uses verified source metadata and inspects the resulting records. If metadata is incomplete, they repair a small candidate list rather than pretending the query is authoritative. Search gathers candidates; the original source and transfer purpose establish whether those candidates belong. Keep the tested query with its explanation so the subset can be reproduced.

A source record can generate more than one review prompt. Decide whether the receiving task needs the underlying fields or explicit question-and-answer directions. The official exporting manual describes the text options; inspect the actual result rather than assuming its row count equals the number of review prompts you remember. A note-level subset can be correct while the receiving question design still needs a decision about which direction to ask.

For the lecture example, a definition note has a forward meaning question and a reverse production question. The learner wants only a clear definition task for the text trial. They record that decision and inspect the exported fields. They do not duplicate rows automatically to mimic every native prompt, because the reverse cue may be ambiguous without template context. If both directions matter, design and verify each explicitly. The distinction prevents a small migration sample from silently becoming a different question set merely because the native note type generated several fronts.

Preserve the native collection and export a representative subset according to the documented controls. Inspect the file in a plain-text or spreadsheet viewer with the intended separator and encoding. Compare rows against the ledger. Check the known inclusion, the exclusion, and examples containing punctuation, optional fields, or meaningful characters. An export completing without an error does not establish that the selected scope or field order matches the destination.

Read each sample row as a question task. Is the accepted answer complete? Is a source reference still associated with the right record? Does a quoted example remain inside its field? Is an image filename being mistaken for available visual content? A text export can faithfully expose a reference without carrying the referenced media into the receiving workflow. Record those cases and keep them out of a text-only trial unless the new task is answerable without the missing asset. The exported file should be inspectable evidence, not a black box whose success is inferred from its existence.

Map the source fields deliberately and import only the sample first. Compare the resulting front, answer, supporting explanation, and source with the original. Do not rely solely on matching row totals: a shifted field can preserve the count while corrupting every question. Include the hardest row shape in the test. If a receiving template reveals the answer on the front, the parsing may be correct while the question design is wrong. Diagnose those separately.

In the fictional lecture subset, the learner finds that an example sentence is being treated as the accepted answer rather than supporting context. They revise the field map and retest the same sample before importing more. Another note needs a qualifier on its front to distinguish two definitions. They repair the question and compare it with the source. This is the point of a deliberately small subset: the learner can examine each outcome closely enough to identify meaning changes. A broad transfer would make the same defect harder to inspect and more expensive to repair.

Use a receiving-stage ledger with one line per representative source record. Record the expected front, accepted answer, and retained source. Mark an excluded direction explicitly rather than leaving its absence unexplained. If a later collaborator asks why the sample contains fewer prompts than the native deck, the ledger answers the question. This keeps an intentional editorial subset from being mistaken for an incomplete or failed export.

Expand when the scope, permissions, field mapping, and resulting tasks match the written acceptance test. Keep the ledger, tested query, export choices, and native recovery copy. If a later chapter introduces a new note type or media dependency, add it to the sample rather than assuming the previous test covers it. Repeating a procedure should mean repeating the checks, not merely clicking the same button sequence. The source and destination can change even when the filename pattern stays familiar.

Elly supports CSV, TSV, and TXT content workflows, not preservation of Anki templates, media, scheduling, or review history. Keep the native collection for those behaviors and evaluate a permitted source-based task independently. A successful selected transfer might help draft clearer questions while Anki remains the review system. The useful outcome is a verified subset that serves its stated purpose. Finish by noting what was transferred, what was deliberately excluded, and how you would recreate the same sample without relying on memory of the original editing session.