diff --git a/code/index/java/nu/marginalia/index/index/CombinedIndexReader.java b/code/index/java/nu/marginalia/index/index/CombinedIndexReader.java
index ba6748fc..27a631f5 100644
--- a/code/index/java/nu/marginalia/index/index/CombinedIndexReader.java
+++ b/code/index/java/nu/marginalia/index/index/CombinedIndexReader.java
@@ -35,13 +35,24 @@ public class CombinedIndexReader {
     }
 
     public IndexQueryBuilderImpl newQueryBuilder(IndexQuery query) {
-        return new IndexQueryBuilderImpl(reverseIndexFullReader, query);
+        return new IndexQueryBuilderImpl(reverseIndexFullReader, reverseIndexPriorityReader, query);
     }
 
     public QueryFilterStepIf hasWordFull(long termId) {
         return reverseIndexFullReader.also(termId);
     }
 
+    public QueryFilterStepIf hasWordPrio(long termId) {
+        return reverseIndexPriorityReader.also(termId);
+    }
+
+
+    /** Creates a query builder for terms in the priority index */
+    public IndexQueryBuilder findPriorityWord(long wordId) {
+        return newQueryBuilder(new IndexQuery(reverseIndexPriorityReader.documents(wordId)))
+                .withSourceTerms(wordId);
+    }
+
     /** Creates a query builder for terms in the full index */
     public IndexQueryBuilder findFullWord(long wordId) {
         return newQueryBuilder(
diff --git a/code/index/java/nu/marginalia/index/index/IndexQueryBuilderImpl.java b/code/index/java/nu/marginalia/index/index/IndexQueryBuilderImpl.java
index 92dce62a..0f63fdbc 100644
--- a/code/index/java/nu/marginalia/index/index/IndexQueryBuilderImpl.java
+++ b/code/index/java/nu/marginalia/index/index/IndexQueryBuilderImpl.java
@@ -11,6 +11,7 @@ import nu.marginalia.index.query.filter.QueryFilterStepIf;
 public class IndexQueryBuilderImpl implements IndexQueryBuilder  {
     private final IndexQuery query;
     private final ReverseIndexReader reverseIndexFullReader;
+    private final ReverseIndexReader reverseIndexPrioReader;
 
     /* Keep track of already added include terms to avoid redundant checks.
      *
@@ -21,10 +22,12 @@ public class IndexQueryBuilderImpl implements IndexQueryBuilder  {
     private final TLongHashSet alreadyConsideredTerms = new TLongHashSet();
 
     IndexQueryBuilderImpl(ReverseIndexReader reverseIndexFullReader,
+                          ReverseIndexReader reverseIndexPrioReader,
                           IndexQuery query)
     {
         this.query = query;
         this.reverseIndexFullReader = reverseIndexFullReader;
+        this.reverseIndexPrioReader = reverseIndexPrioReader;
     }
 
     public IndexQueryBuilder withSourceTerms(long... sourceTerms) {
diff --git a/code/index/java/nu/marginalia/index/index/StatefulIndex.java b/code/index/java/nu/marginalia/index/index/StatefulIndex.java
index dcaf5d7a..74ca220f 100644
--- a/code/index/java/nu/marginalia/index/index/StatefulIndex.java
+++ b/code/index/java/nu/marginalia/index/index/StatefulIndex.java
@@ -4,6 +4,9 @@ import com.google.inject.Inject;
 import com.google.inject.Singleton;
 import it.unimi.dsi.fastutil.longs.*;
 import nu.marginalia.api.searchquery.model.compiled.aggregate.CompiledQueryAggregates;
+import nu.marginalia.index.query.filter.QueryFilterAllOf;
+import nu.marginalia.index.query.filter.QueryFilterAnyOf;
+import nu.marginalia.index.query.filter.QueryFilterStepIf;
 import nu.marginalia.index.results.model.ids.CombinedDocIdList;
 import nu.marginalia.index.results.model.ids.DocMetadataList;
 import nu.marginalia.index.model.QueryParams;
@@ -138,8 +141,16 @@ public class StatefulIndex {
             for (int i = 1; i < elements.size(); i++) {
                 head.addInclusionFilter(combinedIndexReader.hasWordFull(elements.getLong(i)));
             }
-
             queryHeads.add(head);
+
+            // If there are few paths, we can afford to check the priority index as well
+            if (paths.size() < 4) {
+                var prioHead = combinedIndexReader.findPriorityWord(elements.getLong(0));
+                for (int i = 1; i < elements.size(); i++) {
+                    prioHead.addInclusionFilter(combinedIndexReader.hasWordPrio(elements.getLong(i)));
+                }
+                queryHeads.add(prioHead);
+            }
         }
 
         // Add additional conditions to the query heads