diff --git a/src/main/scala/com/gravity/goose/network/HtmlFetcher.scala b/src/main/scala/com/gravity/goose/network/HtmlFetcher.scala
index 34ebf44a7..22c9ed010 100644
--- a/src/main/scala/com/gravity/goose/network/HtmlFetcher.scala
+++ b/src/main/scala/com/gravity/goose/network/HtmlFetcher.scala
@@ -221,6 +221,29 @@ object HtmlFetcher extends AbstractHtmlFetcher with Logging {
         }
       }
       else {
+        /*
+         * Sometimes guessContentTypeFromStream() fails because there is junk before the start of the HTML.
+         *
+         * This code looks at the first 1k characters to see if there is something that looks like HTML lurking beyond some
+         * initial junk, before giving up and saying it is not HTML.
+         */
+
+        val excerpt = htmlResult.substring(0, Math.min(htmlResult.length(), 1024)).toLowerCase()
+
+        val idx1 = excerpt.indexOf("<!doctype html")
+        if (idx1 != -1) {
+          return Some(htmlResult.substring(idx1))
+        }
+
+        val idx2 = excerpt.indexOf("<html")
+        if (idx2 != -1) {
+          val temp = htmlResult.substring(idx2)
+
+          if (temp.contains("<body")) {
+            return Some(temp)
+          }
+        }
+
         throw new NotHtmlException(cleanUrl)
       }
     }
