diff --git a/src/file/extractor/app.cpp b/src/file/extractor/app.cpp
--- a/src/file/extractor/app.cpp
+++ b/src/file/extractor/app.cpp
@@ -164,17 +164,6 @@
         return;
     }
 
-    // HACK: Also, we're ignoring ttext files which are greater tha 10 Mb as we
-    // have trouble processing them
-    //
-    if (mimetype.startsWith(QLatin1String("text/"))) {
-        QFileInfo fileInfo(url);
-        if (fileInfo.size() >= 10 * 1024 * 1024) {
-            tr->removePhaseOne(id);
-            return;
-        }
-    }
-
     // We always run the basic indexing again. This is mostly so that the proper
     // mimetype is set and we get proper type information.
     // The mimetype fetched in the BasicIndexingJob is fast but not accurate
@@ -188,7 +177,14 @@
 
     const QList<KFileMetaData::Extractor*> exList = m_extractorCollection.fetchExtractors(mimetype);
 
+    quint64 filesize = QFileInfo(url).size();
     for (KFileMetaData::Extractor* ex : exList) {
+        // HACK: We're ignoring text files which are greater tha 10 Mb as we
+        // have trouble processing them
+        if ((filesize >= 10*1024*1024) && (ex->extractorProperties()["Id"].toString() == QLatin1String("org.kde.plaintextextractor"))) {
+            qCWarning(BALOO) << "Skipping" << url << ": large plain text file";
+            continue;
+        }
         ex->extract(&result);
     }