diff --git a/crates/milli/src/heed_codec/compressed_obkv_codec.rs b/crates/milli/src/heed_codec/compressed_obkv_codec.rs
index 05b0d90ac..711a5c049 100644
--- a/crates/milli/src/heed_codec/compressed_obkv_codec.rs
+++ b/crates/milli/src/heed_codec/compressed_obkv_codec.rs
@@ -103,8 +103,10 @@ impl<'a> CompressedKvReaderU16<'a> {
 pub struct CompressedKvWriterU16(Vec<u8>);
 
 impl CompressedKvWriterU16 {
-    // TODO ask for a KvReaderU16 here
-    pub fn new_with_dictionary(input: &[u8], dictionary: &EncoderDictionary) -> io::Result<Self> {
+    pub fn new_with_dictionary(
+        input: &KvReaderU16,
+        dictionary: &EncoderDictionary,
+    ) -> io::Result<Self> {
         let mut compressor = Compressor::with_prepared_dictionary(dictionary)?;
         compressor.compress(input).map(CompressedKvWriterU16)
     }
diff --git a/crates/milli/src/update/new/extract/documents/compression.rs b/crates/milli/src/update/new/extract/documents/compression.rs
index 3b323ff1b..a068b7afd 100644
--- a/crates/milli/src/update/new/extract/documents/compression.rs
+++ b/crates/milli/src/update/new/extract/documents/compression.rs
@@ -3,8 +3,12 @@ use std::sync::atomic::{self, AtomicUsize};
 
 use bumpalo::Bump;
 use heed::RwTxn;
+use rayon::iter::{ParallelBridge, ParallelIterator as _};
+use roaring::RoaringBitmap;
+use zstd::bulk::Compressor;
 use zstd::dict::{from_continuous, EncoderDictionary};
 
+use crate::heed_codec::CompressedKvWriterU16;
 use crate::update::new::document::Document as _;
 use crate::update::new::indexer::document_changes::{
     DocumentChangeContext, DocumentChanges, Extractor, IndexingContext, Progress,
@@ -87,8 +91,13 @@ where
                 Step::PreparingCompressionDictionary,
             )?;
 
+            let mut all_documents_seen = RoaringBitmap::new();
             for data in datastore {
-                let CompressorExtractorData { fields, fields_count, .. } = data.into_inner();
+                let CompressorExtractorData { documents_seen, fields, fields_count, must_stop: _ } =
+                    data.into_inner();
+
+                all_documents_seen |= documents_seen;
+
                 let mut fields_iter = fields.into_iter();
                 for field_count in fields_count {
                     let mut document_fields_size = 0;
@@ -109,10 +118,31 @@ where
 
             let dictionary = from_continuous(&sample_data, &sample_sizes, DICTIONARY_MAX_SIZE)?;
             index.put_document_compression_dictionary(wtxn, &dictionary)?;
+            let dictionary = EncoderDictionary::copy(&dictionary, COMPRESSION_LEVEL);
 
-            todo!("compress (in parallel) all the database documents that are not impacted by the current update");
+            let all_documents = index.documents_ids(wtxn)?;
+            let documents_to_compress = all_documents - all_documents_seen;
+            let pi = documents_to_compress.into_iter().par_bridge().map(|docid| {
+                /// TODO put the rtxn and encoder dict in a ThreadLocal datastore
+                let rtxn = index.read_txn()?;
+                let compressed_document = index.compressed_document(&rtxn, docid)?.unwrap();
+                // The documents are not compressed with any dictionary at this point.
+                let document = compressed_document.as_non_compressed();
+                let compressed = CompressedKvWriterU16::new_with_dictionary(document, &dictionary)?;
+                Ok((docid, compressed)) as crate::Result<_>
+            });
 
-            Ok(Some(EncoderDictionary::copy(&dictionary, COMPRESSION_LEVEL)))
+            // We compress in parallel and sequentially write the documents
+            // in the database using the above parallel iterator.
+            rayon_par_bridge::par_bridge(100, pi, |seq_iter| {
+                for result in seq_iter {
+                    let (docid, compressed_document) = result?;
+                    index.documents.put(wtxn, &docid, &compressed_document)?;
+                }
+                Ok(()) as crate::Result<_>
+            })?;
+
+            Ok(Some(dictionary))
         }
     }
 }
@@ -126,6 +156,8 @@ struct CompressorExtractor {
 
 #[derive(Default)]
 struct CompressorExtractorData<'extractor> {
+    /// The set of documents impacted by this update: deleted, modified, or updated.
+    documents_seen: RoaringBitmap,
     /// The field content in JSON but as bytes.
     fields: Vec<&'extractor [u8]>,
     /// The number of fields associated to single documents.
@@ -160,9 +192,9 @@ impl<'extractor> Extractor<'extractor> for CompressorExtractor {
             }
 
             let change = change?;
-            match change {
-                DocumentChange::Deletion(_) => (),
-                DocumentChange::Update(_) => (),
+            let docid = match change {
+                DocumentChange::Deletion(deletion) => deletion.docid(),
+                DocumentChange::Update(update) => update.docid(),
                 DocumentChange::Insertion(insertion) => {
                     let mut fields_count = 0;
                     for result in insertion.inserted().iter_top_level_fields() {
@@ -176,8 +208,13 @@ impl<'extractor> Extractor<'extractor> for CompressorExtractor {
                         self.extracted_documents_count.fetch_add(1, atomic::Ordering::SeqCst);
                     data.must_stop = previous_count >= self.total_documents_to_extract;
                     data.fields_count.push(fields_count);
+
+                    insertion.docid()
                 }
-            }
+            };
+
+            let is_new = data.documents_seen.insert(docid);
+            debug_assert!(is_new, "We must not see the same documents multiple times");
         }
 
         Ok(())