diff --git a/proto/anki/import_export.proto b/proto/anki/import_export.proto
index 7c2679c4a..2e1646ab5 100644
--- a/proto/anki/import_export.proto
+++ b/proto/anki/import_export.proto
@@ -97,22 +97,20 @@ message MediaEntries {
 }
 
 message ImportCsvRequest {
-  message CsvColumn {
-    enum Other {
-      IGNORE = 0;
-      TAGS = 1;
-    }
-    oneof variant {
-      uint32 field = 1;
-      Other other = 2;
-    }
+  // Source column indices for note data. Negative values mean n/a.
+  message Columns {
+    repeated int32 fields = 1;
+    int32 tags = 2;
+    int32 deck = 3;
+    int32 notetype = 4;
   }
   string path = 1;
   int64 deck_id = 2;
   int64 notetype_id = 3;
-  repeated CsvColumn columns = 4;
-  CsvMetadata.Delimiter delimiter = 5;
-  bool is_html = 6;
+  CsvMetadata.Delimiter delimiter = 4;
+  bool is_html = 5;
+  Columns columns = 6;
+  repeated string column_names = 7;
 }
 
 message CsvMetadataRequest {
diff --git a/rslib/src/backend/import_export.rs b/rslib/src/backend/import_export.rs
index 7664db428..89ca3d153 100644
--- a/rslib/src/backend/import_export.rs
+++ b/rslib/src/backend/import_export.rs
@@ -6,14 +6,8 @@ use std::path::Path;
 use super::{progress::Progress, Backend};
 pub(super) use crate::backend_proto::importexport_service::Service as ImportExportService;
 use crate::{
-    backend_proto::{
-        self as pb,
-        export_anki_package_request::Selector,
-        import_csv_request::{csv_column, CsvColumn},
-    },
-    import_export::{
-        package::import_colpkg, text::csv::Column, ExportProgress, ImportProgress, NoteLog,
-    },
+    backend_proto::{self as pb, export_anki_package_request::Selector},
+    import_export::{package::import_colpkg, ExportProgress, ImportProgress, NoteLog},
     prelude::*,
     search::SearchNode,
 };
@@ -93,9 +87,12 @@ impl ImportExportService for Backend {
                 &input.path,
                 input.deck_id.into(),
                 input.notetype_id.into(),
-                input.columns.into_iter().map(Into::into).collect(),
                 delimiter,
                 input.is_html,
+                input
+                    .columns
+                    .ok_or_else(|| AnkiError::invalid_input("missing value"))?,
+                input.column_names,
             )
         })
         .map(Into::into)
@@ -142,17 +139,3 @@ impl From<OpOutput<NoteLog>> for pb::ImportResponse {
         }
     }
 }
-
-impl From<CsvColumn> for Column {
-    fn from(column: CsvColumn) -> Self {
-        match column.variant.unwrap_or(csv_column::Variant::Other(0)) {
-            csv_column::Variant::Field(idx) => Column::Field(idx as usize),
-            csv_column::Variant::Other(i) => {
-                match csv_column::Other::from_i32(i).unwrap_or_default() {
-                    csv_column::Other::Tags => Column::Tags,
-                    csv_column::Other::Ignore => Column::Ignore,
-                }
-            }
-        }
-    }
-}
diff --git a/rslib/src/import_export/text/csv/import.rs b/rslib/src/import_export/text/csv/import.rs
index fd11a2379..11536bec1 100644
--- a/rslib/src/import_export/text/csv/import.rs
+++ b/rslib/src/import_export/text/csv/import.rs
@@ -2,35 +2,35 @@
 // License: GNU AGPL, version 3 or later; http://www.gnu.org/licenses/agpl.html
 
 use std::{
+    collections::HashMap,
     fs::File,
     io::{BufRead, BufReader, Read, Seek, SeekFrom},
 };
 
 use crate::{
+    backend_proto::import_export::import_csv_request::Columns as ProtoColumns,
     import_export::{
-        text::{
-            csv::{metadata::Delimiter, Column},
-            ForeignData, ForeignNote,
-        },
+        text::{csv::metadata::Delimiter, import::NotetypeForString, ForeignData, ForeignNote},
         NoteLog,
     },
     prelude::*,
 };
 
 impl Collection {
+    #[allow(clippy::too_many_arguments)]
     pub fn import_csv(
         &mut self,
         path: &str,
         deck_id: DeckId,
         notetype_id: NotetypeId,
-        columns: Vec<Column>,
         delimiter: Delimiter,
         is_html: bool,
+        columns: ProtoColumns,
+        column_names: Vec<String>,
     ) -> Result<OpOutput<NoteLog>> {
-        let notetype = self.get_notetype(notetype_id)?.ok_or(AnkiError::NotFound)?;
-        let fields_len = notetype.fields.len();
         let file = File::open(path)?;
-        let notes = deserialize_csv(file, &columns, fields_len, delimiter, is_html)?;
+        let mut ctx = ColumnContext::new(columns, column_names, is_html, self);
+        let notes = ctx.deserialize_csv(file, delimiter)?;
 
         ForeignData {
             // TODO: refactor to allow passing ids directly
@@ -43,21 +43,170 @@ impl Collection {
     }
 }
 
-fn deserialize_csv(
-    mut reader: impl Read + Seek,
-    columns: &[Column],
-    fields_len: usize,
-    delimiter: Delimiter,
-    is_html: bool,
-) -> Result<Vec<ForeignNote>> {
-    remove_tags_line_from_reader(&mut reader)?;
-    let mut csv_reader = csv::ReaderBuilder::new()
-        .has_headers(false)
-        .flexible(true)
-        .comment(Some(b'#'))
-        .delimiter(delimiter.byte())
-        .from_reader(reader);
-    deserialize_csv_reader(&mut csv_reader, columns, fields_len, is_html)
+/// Column indices for the fields of a notetype.
+type FieldSourceColumns = Vec<Option<usize>>;
+
+struct ColumnContext<'a, C: NotetypeForString> {
+    tags_column: Option<usize>,
+    deck_column: Option<usize>,
+    notetype_column: Option<usize>,
+    /// Source column indices for the fields of a notetype, identified by its
+    /// name or id as string. The empty string corresponds to the default notetype.
+    notetype_fields: HashMap<String, FieldSourceColumns>,
+    /// CSV column labels. Used to identify the source columns for the fields
+    /// of a notetype.
+    column_names: HashMap<String, usize>,
+    /// How fields are converted to strings. Used for escaping HTML if appropriate.
+    stringify: fn(&str) -> String,
+    col: &'a mut C,
+}
+
+impl<'a, C: NotetypeForString> ColumnContext<'a, C> {
+    fn new(
+        columns: ProtoColumns,
+        column_names: Vec<String>,
+        is_html: bool,
+        col: &'a mut C,
+    ) -> Self {
+        Self {
+            tags_column: columns.tags.try_into().ok(),
+            deck_column: columns.deck.try_into().ok(),
+            notetype_column: columns.notetype.try_into().ok(),
+            notetype_fields: notetype_fields_map(&columns.fields),
+            column_names: column_names_map(column_names),
+            stringify: stringify_fn(is_html),
+            col,
+        }
+    }
+
+    fn deserialize_csv(
+        &mut self,
+        mut reader: impl Read + Seek,
+        delimiter: Delimiter,
+    ) -> Result<Vec<ForeignNote>> {
+        remove_tags_line_from_reader(&mut reader)?;
+        let mut csv_reader = csv::ReaderBuilder::new()
+            .has_headers(false)
+            .flexible(true)
+            .comment(Some(b'#'))
+            .delimiter(delimiter.byte())
+            .from_reader(reader);
+        self.deserialize_csv_reader(&mut csv_reader)
+    }
+
+    fn deserialize_csv_reader(
+        &mut self,
+        reader: &mut csv::Reader<impl Read>,
+    ) -> Result<Vec<ForeignNote>> {
+        reader
+            .records()
+            .into_iter()
+            .map(|res| {
+                res.map_err(Into::into)
+                    .and_then(|record| self.foreign_note_from_record(&record))
+            })
+            .collect()
+    }
+
+    fn foreign_note_from_record(&mut self, record: &csv::StringRecord) -> Result<ForeignNote> {
+        let notetype = self.gather_notetype(record);
+        let deck = self.gather_deck(record);
+        let tags = self.gather_tags(record);
+        let fields = self.gather_note_fields(record, &notetype)?;
+        Ok(ForeignNote {
+            notetype,
+            fields,
+            tags,
+            deck,
+            ..Default::default()
+        })
+    }
+
+    fn gather_notetype(&self, record: &csv::StringRecord) -> String {
+        self.notetype_column
+            .and_then(|i| record.get(i))
+            .unwrap_or_default()
+            .to_string()
+    }
+
+    fn gather_deck(&self, record: &csv::StringRecord) -> String {
+        self.deck_column
+            .and_then(|i| record.get(i))
+            .unwrap_or_default()
+            .to_string()
+    }
+
+    fn gather_tags(&self, record: &csv::StringRecord) -> Vec<String> {
+        self.tags_column
+            .and_then(|i| record.get(i))
+            .unwrap_or_default()
+            .split_whitespace()
+            .filter(|s| !s.is_empty())
+            .map(ToString::to_string)
+            .collect()
+    }
+
+    fn gather_note_fields(
+        &mut self,
+        record: &csv::StringRecord,
+        notetype: &str,
+    ) -> Result<Vec<String>> {
+        let stringify = self.stringify;
+        Ok(self
+            .get_notetype_fields(notetype)?
+            .iter()
+            .map(|opt| opt.and_then(|idx| record.get(idx)).unwrap_or_default())
+            .map(stringify)
+            .collect())
+    }
+
+    fn get_notetype_fields(&mut self, notetype: &str) -> Result<&FieldSourceColumns> {
+        Ok(if self.notetype_fields.contains_key(notetype) {
+            // borrow checker doesn't allow to use `if let` here
+            // https://users.rust-lang.org/t/solved-borrow-doesnt-drop-returning-this-value-requires-that/24182
+            self.notetype_fields.get(notetype).unwrap()
+        } else {
+            // TODO: more specific errors
+            let nt = self
+                .col
+                .notetype_for_string(notetype)?
+                .ok_or(AnkiError::NotFound)?;
+            let map = self.build_notetype_fields_map(&nt);
+            if map[0].is_none() {
+                return Err(AnkiError::NotFound);
+            }
+            self.notetype_fields
+                .entry(notetype.to_string())
+                .or_insert(map)
+        })
+    }
+
+    fn build_notetype_fields_map(&mut self, notetype: &Notetype) -> FieldSourceColumns {
+        notetype
+            .fields
+            .iter()
+            .map(|f| self.column_names.get(&f.name).copied())
+            .collect()
+    }
+}
+
+fn column_names_map(column_names: Vec<String>) -> HashMap<String, usize> {
+    HashMap::from_iter(column_names.into_iter().enumerate().map(|(i, s)| (s, i)))
+}
+
+fn notetype_fields_map(default_fields: &[i32]) -> HashMap<String, FieldSourceColumns> {
+    let default_fields = default_fields.iter().map(|&i| i.try_into().ok()).collect();
+    let mut notetype_fields = HashMap::new();
+    notetype_fields.insert(String::new(), default_fields);
+    notetype_fields
+}
+
+fn stringify_fn(is_html: bool) -> fn(&str) -> String {
+    if is_html {
+        ToString::to_string
+    } else {
+        htmlescape::encode_minimal
+    }
 }
 
 /// If the reader's first line starts with "tags:", which is allowed for historic
@@ -77,161 +226,150 @@ fn remove_tags_line_from_reader(reader: &mut (impl Read + Seek)) -> Result<()> {
     Ok(())
 }
 
-fn deserialize_csv_reader(
-    reader: &mut csv::Reader<impl Read>,
-    columns: &[Column],
-    fields_len: usize,
-    is_html: bool,
-) -> Result<Vec<ForeignNote>> {
-    reader
-        .records()
-        .into_iter()
-        .map(|res| {
-            res.map(|record| ForeignNote::from_record(record, columns, fields_len, is_html))
-                .map_err(Into::into)
-        })
-        .collect()
-}
-
-impl ForeignNote {
-    fn from_record(
-        record: csv::StringRecord,
-        columns: &[Column],
-        fields_len: usize,
-        is_html: bool,
-    ) -> Self {
-        let mut note = Self {
-            fields: vec!["".to_string(); fields_len],
-            ..Default::default()
-        };
-        for (&column, value) in columns.iter().zip(record.iter()) {
-            note.add_column_value(column, value, is_html);
-        }
-        note
-    }
-
-    fn add_column_value(&mut self, column: Column, value: &str, is_html: bool) {
-        match column {
-            Column::Ignore => (),
-            Column::Field(idx) => {
-                if let Some(field) = self.fields.get_mut(idx) {
-                    *field = if is_html {
-                        value.to_string()
-                    } else {
-                        htmlescape::encode_minimal(value)
-                    };
-                }
-            }
-            Column::Tags => self.tags.extend(value.split(' ').map(ToString::to_string)),
-        }
-    }
-}
-
 #[cfg(test)]
 mod test {
-    use std::io::Cursor;
+    use std::{io::Cursor, sync::Arc};
 
     use super::*;
+    use crate::notetype::all_stock_notetypes;
+
+    macro_rules! import {
+        ($options:expr, $csv:expr) => {{
+            let reader = Cursor::new($csv);
+            let delimiter = $options.delimiter;
+            let mut ctx = $options.ctx();
+            ctx.deserialize_csv(reader, delimiter).unwrap()
+        }};
+    }
 
     macro_rules! assert_imported_fields {
-        ($options:expr,$csv:expr, $expected:expr) => {
-            let reader = Cursor::new($csv);
-            let notes = deserialize_csv(
-                reader,
-                &$options.columns,
-                $options.fields_len,
-                $options.delimiter,
-                $options.is_html,
-            )
-            .unwrap();
+        ($options:expr, $csv:expr, $expected:expr) => {
+            let notes = import!($options, $csv);
             let fields: Vec<_> = notes.into_iter().map(|note| note.fields).collect();
             assert_eq!(fields, $expected);
         };
     }
 
+    struct MockNotetypeForString(HashMap<String, Arc<Notetype>>);
+
+    impl NotetypeForString for MockNotetypeForString {
+        fn notetype_for_string(&mut self, name_or_id: &str) -> Result<Option<Arc<Notetype>>> {
+            Ok(self.0.get(name_or_id).cloned())
+        }
+    }
+
+    impl MockNotetypeForString {
+        fn new() -> Self {
+            Self(HashMap::from_iter(
+                all_stock_notetypes(&I18n::template_only())
+                    .into_iter()
+                    .map(|nt| (nt.name.clone(), Arc::new(nt))),
+            ))
+        }
+    }
+
     struct CsvOptions {
-        columns: Vec<Column>,
-        fields_len: usize,
         delimiter: Delimiter,
         is_html: bool,
+        columns: ProtoColumns,
+        column_names: Vec<String>,
+        mock_col: MockNotetypeForString,
     }
 
     impl CsvOptions {
         fn new() -> Self {
             Self {
-                columns: vec![Column::Field(0), Column::Field(1)],
-                fields_len: 2,
                 delimiter: Delimiter::Comma,
                 is_html: false,
+                columns: ProtoColumns {
+                    tags: -1,
+                    deck: -1,
+                    notetype: -1,
+                    fields: vec![0, 1],
+                },
+                column_names: vec![],
+                mock_col: MockNotetypeForString::new(),
             }
         }
 
-        fn add_column(mut self, column: Column) -> Self {
-            self.columns.push(column);
-            self
-        }
-
-        fn columns(mut self, columns: Vec<Column>) -> Self {
-            self.columns = columns;
-            self
-        }
-
-        fn fields_len(mut self, fields_len: usize) -> Self {
-            self.fields_len = fields_len;
-            self
-        }
-
-        fn delimiter(mut self, delimiter: Delimiter) -> Self {
-            self.delimiter = delimiter;
-            self
-        }
-
-        #[allow(clippy::wrong_self_convention)]
-        fn is_html(mut self, is_html: bool) -> Self {
-            self.is_html = is_html;
-            self
+        fn ctx(&mut self) -> ColumnContext<MockNotetypeForString> {
+            ColumnContext::new(
+                self.columns.clone(),
+                std::mem::take(&mut self.column_names),
+                self.is_html,
+                &mut self.mock_col,
+            )
         }
     }
 
     #[test]
     fn should_allow_missing_columns() {
-        let options = CsvOptions::new().add_column(Column::Field(2)).fields_len(4);
-        assert_imported_fields!(
-            options,
-            "front,back\nfoo\n",
-            &[&["front", "back", "", ""], &["foo", "", "", ""]]
-        );
+        let mut options = CsvOptions::new();
+        assert_imported_fields!(options, "foo\n", &[&["foo", ""]]);
     }
 
     #[test]
     fn should_respect_custom_delimiter() {
-        let options = CsvOptions::new().delimiter(Delimiter::Pipe);
+        let mut options = CsvOptions::new();
+        options.delimiter = Delimiter::Pipe;
         assert_imported_fields!(options, "fr,ont|ba,ck\n", &[&["fr,ont", "ba,ck"]]);
     }
 
     #[test]
     fn should_ignore_first_line_starting_with_tags() {
-        let options = CsvOptions::new();
+        let mut options = CsvOptions::new();
         assert_imported_fields!(options, "tags:foo\nfront,back\n", &[&["front", "back"]]);
     }
 
     #[test]
     fn should_respect_column_remapping() {
-        let options =
-            CsvOptions::new().columns(vec![Column::Field(1), Column::Ignore, Column::Field(0)]);
+        let mut options = CsvOptions::new();
+        options.columns.fields = vec![2, 0];
         assert_imported_fields!(options, "front,foo,back\n", &[&["back", "front"]]);
     }
 
     #[test]
     fn should_ignore_lines_starting_with_number_sign() {
-        let options = CsvOptions::new();
+        let mut options = CsvOptions::new();
         assert_imported_fields!(options, "#foo\nfront,back\n#bar\n", &[&["front", "back"]]);
     }
 
     #[test]
     fn should_escape_html_entities_if_csv_is_html() {
-        assert_imported_fields!(CsvOptions::new(), "<hr>\n", &[&["&lt;hr&gt;", ""]]);
-        let with_html = CsvOptions::new().is_html(true);
-        assert_imported_fields!(with_html, "<hr>\n", &[&["<hr>", ""]]);
+        let mut options = CsvOptions::new();
+        assert_imported_fields!(options, "<hr>\n", &[&["&lt;hr&gt;", ""]]);
+        options.is_html = true;
+        assert_imported_fields!(options, "<hr>\n", &[&["<hr>", ""]]);
+    }
+
+    #[test]
+    fn should_parse_tag_column() {
+        let mut options = CsvOptions::new();
+        options.columns.tags = 2;
+        let notes = import!(options, "front,back,foo bar\n");
+        assert_eq!(notes[0].tags, &["foo", "bar"]);
+    }
+
+    #[test]
+    fn should_parse_deck_column() {
+        let mut options = CsvOptions::new();
+        options.columns.deck = 2;
+        let notes = import!(options, "front,back,foo bar\n");
+        assert_eq!(notes[0].deck, "foo bar");
+    }
+
+    #[test]
+    fn should_parse_notes_according_to_their_respective_notetypes() {
+        let mut options = CsvOptions::new();
+        options.columns.notetype = 3;
+        options.column_names = ["Front", "Back", "Text", "notetype"]
+            .iter()
+            .map(ToString::to_string)
+            .collect();
+        assert_imported_fields!(
+            options,
+            "front,back,Basic (and reversed card)\n,,foo,Cloze\n",
+            &[&["front", "back"], &["foo", ""]]
+        );
     }
 }
diff --git a/rslib/src/import_export/text/csv/metadata.rs b/rslib/src/import_export/text/csv/metadata.rs
index aac09e322..45b3d2886 100644
--- a/rslib/src/import_export/text/csv/metadata.rs
+++ b/rslib/src/import_export/text/csv/metadata.rs
@@ -9,7 +9,10 @@ use std::{
 use strum::IntoEnumIterator;
 
 pub use crate::backend_proto::csv_metadata::Delimiter;
-use crate::{backend_proto::CsvMetadata, error::ImportError, prelude::*};
+use crate::{
+    backend_proto::CsvMetadata, error::ImportError, import_export::text::import::NotetypeForString,
+    prelude::*,
+};
 
 impl Collection {
     pub fn get_csv_metadata(
diff --git a/rslib/src/import_export/text/csv/mod.rs b/rslib/src/import_export/text/csv/mod.rs
index d3a6fb0dd..f6a8e3f16 100644
--- a/rslib/src/import_export/text/csv/mod.rs
+++ b/rslib/src/import_export/text/csv/mod.rs
@@ -3,10 +3,3 @@
 
 mod import;
 mod metadata;
-
-#[derive(Debug, Clone, Copy)]
-pub enum Column {
-    Field(usize),
-    Ignore,
-    Tags,
-}
diff --git a/rslib/src/import_export/text/import.rs b/rslib/src/import_export/text/import.rs
index deb0eef8e..e0893353d 100644
--- a/rslib/src/import_export/text/import.rs
+++ b/rslib/src/import_export/text/import.rs
@@ -143,6 +143,20 @@ impl<'a> Context<'a> {
     }
 }
 
+pub(super) trait NotetypeForString {
+    fn notetype_for_string(&mut self, name_or_id: &str) -> Result<Option<Arc<Notetype>>>;
+}
+
+impl NotetypeForString for Collection {
+    fn notetype_for_string(&mut self, name_or_id: &str) -> Result<Option<Arc<Notetype>>> {
+        if let Some(nt) = self.get_notetype_for_id_string(name_or_id)? {
+            Ok(Some(nt))
+        } else {
+            self.get_notetype_by_name(name_or_id)
+        }
+    }
+}
+
 impl Collection {
     pub(super) fn deck_id_for_string(&mut self, deck: &str) -> Result<Option<DeckId>> {
         if let Ok(did) = deck.parse::<DeckId>() {
@@ -153,14 +167,6 @@ impl Collection {
         self.get_deck_id(deck)
     }
 
-    pub(super) fn notetype_for_string(&mut self, notetype: &str) -> Result<Option<Arc<Notetype>>> {
-        if let Some(nt) = self.get_notetype_for_id_string(notetype)? {
-            Ok(Some(nt))
-        } else {
-            self.get_notetype_by_name(notetype)
-        }
-    }
-
     fn get_notetype_for_id_string(&mut self, notetype: &str) -> Result<Option<Arc<Notetype>>> {
         notetype
             .parse::<NotetypeId>()