diff --git a/src/calibre/ebooks/pdf/pdftohtml.py b/src/calibre/ebooks/pdf/pdftohtml.py
index 96fb28e923..3599c4a477 100644
--- a/src/calibre/ebooks/pdf/pdftohtml.py
+++ b/src/calibre/ebooks/pdf/pdftohtml.py
@@ -125,8 +125,9 @@ def pdftohtml(output_dir, pdf_path, no_images, as_xml=False):
 
 def parse_outline(raw, output_dir):
     from lxml import etree
+    from calibre.ebooks.oeb.parse_utils import RECOVER_PARSER
     raw = clean_xml_chars(xml_to_unicode(raw, strip_encoding_pats=True, assume_utf8=True)[0])
-    outline = etree.fromstring(raw).xpath('(//outline)[1]')
+    outline = etree.fromstring(raw, parser=RECOVER_PARSER).xpath('(//outline)[1]')
     if outline:
         from calibre.ebooks.oeb.polish.toc import TOC, create_ncx
         outline = outline[0]