diff --git a/lib/email/html_cleaner.rb b/lib/email/html_cleaner.rb
index 19e4b417327..b4333ae91d2 100644
--- a/lib/email/html_cleaner.rb
+++ b/lib/email/html_cleaner.rb
@@ -74,9 +74,20 @@ module Email
     private
 
     def add_newlines(doc)
+      # Replace <br> tags with a markdown \n
       doc.xpath('//br').each do |br|
-        br.replace(Nokogiri::XML::Text.new("\n", doc))
+        br.replace(new_linebreak_node doc)
       end
+      # Surround <p> tags with newlines, to help with line-wise postprocessing
+      # and ensure markdown paragraphs
+      doc.xpath('//p').each do |p|
+        p.before(new_linebreak_node doc)
+        p.after(new_linebreak_node doc, 2)
+      end
+    end
+
+    def new_linebreak_node(doc, count=1)
+      Nokogiri::XML::Text.new("\n" * count, doc)
     end
 
     def trim_process_node(node)
diff --git a/lib/email/receiver.rb b/lib/email/receiver.rb
index e9d09024cd7..03094c3a3af 100644
--- a/lib/email/receiver.rb
+++ b/lib/email/receiver.rb
@@ -159,7 +159,8 @@ module Email
                  (l =~ /via #{SiteSetting.title}(.*)\:$/) ||
                  # This one might be controversial but so many reply lines have years, times and end with a colon.
                  # Let's try it and see how well it works.
-                 (l =~ /\d{4}/ && l =~ /\d:\d\d/ && l =~ /\:$/)
+                 (l =~ /\d{4}/ && l =~ /\d:\d\d/ && l =~ /\:$/) ||
+                 (l =~ /On \w+ \d+,? \d+,?.*wrote:/)
 
         # Headers on subsequent lines
         break if (0..2).all? { |off| lines[idx+off] =~ REPLYING_HEADER_REGEX }