diff --git a/CHANGELOG.md b/CHANGELOG.md
index 97424e4fab5e..75f457a00310 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -28,6 +28,7 @@
 * [#1899](https://github.com/bbatsov/rubocop/issues/1899): Be careful about comments when auto-correcting in `BracesAroundHashParameters`. ([@jonas054][])
 * [#1897](https://github.com/bbatsov/rubocop/issues/1897): Don't report that semicolon separated statements can be converted to modifier form in `IfUnlessModifier` (and don't auto-correct them). ([@jonas054][])
 * [#1644](https://github.com/bbatsov/rubocop/issues/1644): Don't search the entire file system when a folder is named `,` (fix for jruby and rbx). ([@rrosenblum][])
+* [#1901](https://github.com/bbatsov/rubocop/issues/1901): Properly fix comments that are missing a note. ([@rrosenblum][])
 
 ## 0.31.0 (05/05/2015)
 
diff --git a/lib/rubocop/cop/style/comment_annotation.rb b/lib/rubocop/cop/style/comment_annotation.rb
index c2cf7e65399c..2432c8937e51 100644
--- a/lib/rubocop/cop/style/comment_annotation.rb
+++ b/lib/rubocop/cop/style/comment_annotation.rb
@@ -13,25 +13,34 @@ class CommentAnnotation < Cop
 
         def investigate(processed_source)
           processed_source.comments.each do |comment|
-            margin, first_word, colon, space, note = split_comment(comment)
+            _margin, first_word, colon, space, note = split_comment(comment)
             next unless annotation?(comment) &&
                         !correct_annotation?(first_word, colon, space, note)
 
-            start = comment.loc.expression.begin_pos + margin.length
-            length = first_word.length + colon.to_s.length + space.to_s.length
-            range = Parser::Source::Range.new(processed_source.buffer,
-                                              start,
-                                              start + length)
-            add_offense(range, range)
+            add_offense(comment, :expression)
           end
         end
 
         private
 
-        def autocorrect(range)
+        def autocorrect(comment)
+          margin, first_word, colon, space, note = split_comment(comment)
+          start = comment.loc.expression.begin_pos + margin.length
+
           lambda do |corrector|
-            annotation_keyword = range.source.split(/:?\s+/).first
-            corrector.replace(range, annotation_keyword.upcase << ': ')
+            if note.nil?
+              start += first_word.length
+              range = Parser::Source::Range.new(comment,
+                                                start,
+                                                comment.loc.expression.end_pos)
+              corrector.remove(range)
+            else
+              length = first_word.length + colon.to_s.length + space.to_s.length
+              range = Parser::Source::Range.new(comment.loc.expression.source,
+                                                start,
+                                                start + length)
+              corrector.replace(range, "#{first_word.upcase}: ")
+            end
           end
         end
 
diff --git a/spec/rubocop/cop/style/comment_annotation_spec.rb b/spec/rubocop/cop/style/comment_annotation_spec.rb
index 78d93e194dd5..7acce99fa9ea 100644
--- a/spec/rubocop/cop/style/comment_annotation_spec.rb
+++ b/spec/rubocop/cop/style/comment_annotation_spec.rb
@@ -39,9 +39,9 @@
     it 'marks the annotation keyword' do
       inspect_source(cop, '# TODO:make better')
       formatter.report_file('t', cop.offenses)
-      expect(output.string).to eq(["t:1:3: C: #{described_class::MSG}",
+      expect(output.string).to eq(["t:1:1: C: #{described_class::MSG}",
                                    '# TODO:make better',
-                                   '  ^^^^^',
+                                   '^^^^^^^^^^^^^^^^^^',
                                    ''].join("\n"))
     end
   end
@@ -71,6 +71,11 @@
     expect(cop.offenses.size).to eq(1)
   end
 
+  it 'autocorrects when missing the note' do
+    corrected = autocorrect_source(cop, '# HACK:')
+    expect(corrected).to eq('# HACK')
+  end
+
   it 'accepts upper case keyword with colon, space and note' do
     inspect_source(cop, '# REVIEW: not sure about this')
     expect(cop.offenses).to be_empty