diff --git a/pelican/readers.py b/pelican/readers.py
index fb2ccfc4..046aeeb6 100644
--- a/pelican/readers.py
+++ b/pelican/readers.py
@@ -37,7 +37,6 @@ except ImportError:
 from pelican.contents import Page, Category, Tag, Author
 from pelican.utils import get_date, pelican_open
 
-
 logger = logging.getLogger(__name__)
 
 METADATA_PROCESSORS = {
@@ -104,6 +103,12 @@ class PelicanHTMLTranslator(HTMLTranslator):
     def depart_abbreviation(self, node):
         self.body.append('</abbr>')
 
+    def visit_image(self, node):
+        # set an empty alt if alt is not specified
+        # avoids that alt is taken from src
+        node['alt'] = node.get('alt', '')
+        return HTMLTranslator.visit_image(self, node)
+
 
 class RstReader(Reader):
     enabled = bool(docutils)
@@ -391,6 +396,41 @@ def read_file(base_path, path, content_class=Page, fmt=None,
     content, reader_metadata = reader.read(path)
     metadata.update(reader_metadata)
 
+    # create warnings for all images with empty alt (up to a certain number)
+    # as they are really likely to be accessibility flaws
+    if content:
+        # find images with empty alt
+        imgs = re.compile(r"""
+            (?:
+                # src before alt
+                <img
+                [^\>]*
+                src=(['"])(.*)\1
+                [^\>]*
+                alt=(['"])\3
+            )|(?:
+                # alt before src
+                <img
+                [^\>]*
+                alt=(['"])\4
+                [^\>]*
+                src=(['"])(.*)\5
+            )
+            """, re.X)
+        matches = re.findall(imgs, content)
+        # find a correct threshold
+        nb_warnings = 10
+        if len(matches) == nb_warnings + 1:
+            nb_warnings += 1 # avoid bad looking case
+        # print one warning per image with empty alt until threshold
+        for match in matches[:nb_warnings]:
+            logger.warning('Empty alt attribute for image {} in {}'.format(
+                           os.path.basename(match[1] + match[5]), path))
+        # print one warning for the other images with empty alt
+        if len(matches) > nb_warnings:
+            logger.warning('{} other images with empty alt attributes'.format(
+                           len(matches) - nb_warnings))
+
     # eventually filter the content with typogrify if asked so
     if content and settings and settings['TYPOGRIFY']:
         from typogrify.filters import typogrify