diff --git a/src/bin/search.rs b/src/bin/search.rs
index 86abe752c..c3fd7cd66 100644
--- a/src/bin/search.rs
+++ b/src/bin/search.rs
@@ -58,7 +58,7 @@ fn main() -> anyhow::Result<()> {
         let before = Instant::now();
 
         let query = result?;
-        let documents_ids = index.search(&rtxn, &query)?;
+        let (_, documents_ids) = index.search(&rtxn, &query)?;
         let headers = match index.headers(&rtxn)? {
             Some(headers) => headers,
             None => return Ok(()),
diff --git a/src/bin/serve.rs b/src/bin/serve.rs
index f8e84555d..f9e18b315 100644
--- a/src/bin/serve.rs
+++ b/src/bin/serve.rs
@@ -152,14 +152,21 @@ async fn main() -> anyhow::Result<()> {
             let before_search = Instant::now();
             let rtxn = env_cloned.read_txn().unwrap();
 
-            let documents_ids = index.search(&rtxn, &query.query).unwrap();
+            let (words, documents_ids) = index.search(&rtxn, &query.query).unwrap();
 
             let mut body = Vec::new();
             if let Some(headers) = index.headers(&rtxn).unwrap() {
                 // We write the headers
                 body.extend_from_slice(headers);
 
-                let re = Regex::new(r"(?i)(hello)").unwrap();
+                let mut regex = format!(r"(?i)\b(");
+                let number_of_words = words.len();
+                words.into_iter().enumerate().for_each(|(i, w)| {
+                    regex.push_str(&w);
+                    if i != number_of_words - 1 { regex.push('|') }
+                });
+                regex.push_str(r")\b");
+                let re = Regex::new(&regex).unwrap();
 
                 for id in documents_ids {
                     let content = index.documents.get(&rtxn, &BEU32::new(id)).unwrap();
diff --git a/src/lib.rs b/src/lib.rs
index 6af75e875..9384155b7 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -4,7 +4,7 @@ mod iter_shortest_paths;
 mod query_tokens;
 
 use std::borrow::Cow;
-use std::collections::HashMap;
+use std::collections::{HashSet, HashMap};
 use std::hash::BuildHasherDefault;
 use std::time::Instant;
 
@@ -96,10 +96,10 @@ impl Index {
         }
     }
 
-    pub fn search(&self, rtxn: &heed::RoTxn, query: &str) -> anyhow::Result<Vec<DocumentId>> {
+    pub fn search(&self, rtxn: &heed::RoTxn, query: &str) -> anyhow::Result<(HashSet<String>, Vec<DocumentId>)> {
         let fst = match self.fst(rtxn)? {
             Some(fst) => fst,
-            None => return Ok(vec![]),
+            None => return Ok(Default::default()),
         };
 
         let (lev0, lev1, lev2) = (&LEVDIST0, &LEVDIST1, &LEVDIST2);
@@ -342,7 +342,10 @@ impl Index {
             }
         }
 
-        debug!("{} candidates", documents.iter().map(RoaringBitmap::len).sum::<u64>());
-        Ok(documents.iter().flatten().take(20).collect())
+        debug!("{} final candidates", documents.iter().map(RoaringBitmap::len).sum::<u64>());
+        let words = words.into_iter().flatten().map(|(w, _)| String::from_utf8(w).unwrap()).collect();
+        let documents = documents.iter().flatten().take(20).collect();
+
+        Ok((words, documents))
     }
 }