diff --git a/scrape_youtube.py b/scrape_youtube.py
new file mode 100644
index 0000000..66b7916
--- /dev/null
+++ b/scrape_youtube.py
@@ -0,0 +1,156 @@
+#!/usr/bin/env python
+
+"""Scrape PyCon UK's youtube channel for videos and link them in the session.
+
+Fragile regular expressions are used for parsing the youtube channel and the
+session files. Use only with care and version control!
+
+Without third party libraries only few videos can be retrieved from youtube.
+Youtube's official API is not used to avoid need of user credentials. If
+youtbe-dl is installed, all videos are found (but it is not very fast).
+
+Matching videos to sessions is done via a poor-man fuzzy match. A title
+identifier is created by stripping all characters from title except a-z and A-Z
+and converting to lower case. This identifier is used for matching videos
+and sessions.
+
+TODO:
+ - improve fuzzy matching
+ - options via argparse
+
+"""
+
+import re
+import time
+import logging
+from glob import iglob
+from urllib.request import urlopen
+
+try:
+    import youtube_dl
+    use_youtube_dl = True
+except ImportError:
+    use_youtube_dl = False
+
+
+YOUTUBE_URI = 'https://www.youtube.com/channel/UChA9XP_feY1-1oSy2L7acog/videos'
+YOUTUBE_TITLE_PREFIX = 'PYCON UK 2017'
+YOUTUBE_URI_PREFIX = 'https://www.youtube.com'
+YOUTUBE_REGEX = '<a .* href="(.*)">{prefix}: (.*)</a>'.format(
+    prefix=YOUTUBE_TITLE_PREFIX)
+
+SESSION_FILES = './sessions/**/*.md'
+
+# title might contain quotes, they get stripped in title_identifier() anyway
+TITLE_REGEX = 'title: (.*)'
+
+
+logging.basicConfig(level=logging.DEBUG,
+                    format='%(asctime)s [%(levelname)s] - %(message)s')
+
+
+def title_identifier(title):
+    """See module docstring."""
+    title_id = re.sub('[^a-zA-Z0-9]', '', title).lower()
+    title_id = re.sub('keynote', '', title_id)
+    return title_id
+
+
+def retrieve_youtube_videos_poor_man():
+    with urlopen(YOUTUBE_URI) as response:
+        # TODO hardcoded encoding might not be the best idea
+        html = response.read().decode('utf-8')
+    return {title_identifier(title): YOUTUBE_URI_PREFIX + uri
+            for uri, title in re.findall(YOUTUBE_REGEX, html)}
+
+
+def retrieve_youtube_videos_youtube_dl():
+    logging.info("Downloading information from youtube (this might take a few "
+                 "minutes)...")
+    with youtube_dl.YoutubeDL({}) as ydl:
+        videos = ydl.extract_info(YOUTUBE_URI, download=False)
+
+    def strip_prefix(title):
+        return title[len(YOUTUBE_TITLE_PREFIX) + 2:]
+
+    return {
+        title_identifier(strip_prefix(video['title'])): video['webpage_url']
+        for video in videos['entries']
+        if video['title'].startswith(YOUTUBE_TITLE_PREFIX)
+    }
+
+
+def retrieve_youtube_videos():
+    """Download list of youtube videos from channel and return a dictionary of
+    the form {title_identifier: video_uri}."""
+    if use_youtube_dl:
+        return retrieve_youtube_videos_youtube_dl()
+    else:
+        return retrieve_youtube_videos_poor_man()
+
+
+def main():
+    t0 = time.time()
+
+    youtube_videos = retrieve_youtube_videos()
+
+    logging.info("Found %d links to youtube videos in channel.",
+                 len(youtube_videos))
+
+    no_video_found = []
+    videos_added = []
+    video_already_there = []
+
+    for fname in iglob(SESSION_FILES, recursive=True):
+        with open(fname, 'r') as f:
+            session_content = f.read()
+
+        title = re.search(TITLE_REGEX, session_content)
+        if not title:
+            logging.error("No title found in %s", fname)
+            continue
+
+        title = title.groups()[0]
+
+        title_id = title_identifier(title)
+        youtube_uri = youtube_videos.pop(title_id, None)
+
+        if not re.search('video:( )*\n', session_content):
+            # TODO check if there is really a link already and if it is
+            # identical to scraped link
+            video_already_there.append(title)
+            continue
+
+        if youtube_uri is None:
+            no_video_found.append(title)
+            continue
+
+        session_content = re.sub('video:( )*\n', 'video: {}\n'.format(
+            youtube_uri), session_content)
+
+        # TODO this is not safe, should write to tmp file first, but it is a
+        # GIT repository anyway...
+
+        logging.debug("Adding video link %s to session file %s...",
+                      youtube_uri, fname)
+        with open(fname, 'w') as f:
+            # video already there?
+            f.write(session_content)
+
+        videos_added.append(title)
+
+    logging.info("Added %d video links to sessions.", len(videos_added))
+    logging.info("%d video links were already there.", len(video_already_there))
+    logging.info("No video found for %d sessions.", len(no_video_found))
+    if len(youtube_videos):
+        # this prints the title identifier, which is a bit ugly
+        no_session = "\n".join("    {}: {}".format(title_id, uri)
+                               for title_id, uri in youtube_videos.items())
+        logging.error("No session file found for %d videos:\n%s",
+                      len(youtube_videos), no_session)
+
+    logging.info("Scraping took %s seconds.", time.time() - t0)
+
+
+if __name__ == '__main__':
+    main()