diff --git a/README.md b/README.md
index 5a33077..bcbc989 100644
--- a/README.md
+++ b/README.md
@@ -1,15 +1,15 @@
 # YouTube Maze
 
-Some YouTube creators have made video-gamebooks during which viewers can make choices that influence the course of the narration.
+Some YouTube creators have made video-gamebooks during which viewers are invited to make choices that influence the course of the narration.
 This is generally implemented by breaking up the video into short interlinked segments.
-This tool automatically generates a visual representation of networks underlying such games by automatically exploring the video maze starting from a root video.
+This tool automatically generates visual representations of the networks underlying such games.
 
 [See it live →](https://youtube-maze.delab.re)
 
 ## How Does it Work?
 
-The [algorithm](lib/explore.mjs) starts at the first video and explores the links recursively using a breadth-first search approach, keeping track of the links as it goes.
-The [Graphviz](http://www.graphviz.org/) library is used to automatically lay out the graph so that as few edges cross as possible.
+The [exploration algorithm](lib/explore.mjs) automatically downloads metadata about each video from YouTube and then proceeds to a breadth-first search to generate a graph which is laid out using [Graphviz](http://www.graphviz.org).
+Only the links in the end screens and in the in-video cards are considered for the exploration.
 
 ## Running Locally
 
diff --git a/public/index.html b/public/index.html
index b8daf5c..74b09aa 100644
--- a/public/index.html
+++ b/public/index.html
@@ -9,29 +9,36 @@
         <h1>YouTube Maze</h1>
 
         <p>
-            Some YouTube creators have made video-gamebooks during which viewers can make choices that influence the course of the narration.
+            Some YouTube creators have made video-gamebooks during which viewers are invited to make choices that influence the course of the narration.
             This is generally implemented by breaking up the video into short interlinked segments.
-            This tool automatically generates a visual representation of networks underlying such games by automatically exploring the video maze starting from a root video.
+            This tool automatically generates visual representations of the networks underlying such games.
         </p>
 
-        <p>
-            <a href="https://forge.delab.re/matteo/youtube-maze">See the source code →</a>
-        </p>
-
-        <h2>Generate a graph</h2>
-
-        <form action="/mazes" method="POST">
-            <p>
-                <input type="text" name="url" placeholder="Enter the URL of a YouTube video to generate the graph…">
-                <input type="submit" value="Generate">
-            </p>
-        </form>
-
         <h2>Some examples</h2>
 
         <ul>
             <li><a href="/mazes/EZGra6O8ClQ">e-penser — “1 avril 2017”</a></li>
             <li><a href="/mazes/XM1ssJ8yxdg">Defakator — “Le YouTube Game”</a></li>
         </ul>
+
+        <h2>Generate a graph</h2>
+
+        <form action="/mazes" method="POST">
+            <p>
+                <input type="text" name="url" placeholder="Enter the URL of any YouTube video to start from…">
+                <input type="submit" value="Generate">
+            </p>
+        </form>
+
+        <h2>More information</h2>
+
+        <p>
+            The exploration algorithm automatically downloads metadata about each video from YouTube and then proceeds to a breadth-first search to generate a graph which is laid out using <a href="http://www.graphviz.org">Graphviz</a>.
+            Only the links in the end screens and in the in-video cards are considered for the exploration.
+        </p>
+
+        <p>
+            <a href="https://forge.delab.re/matteo/youtube-maze">See the source code →</a>
+        </p>
     </body>
 </html>
diff --git a/public/style.css b/public/style.css
index 6cfa2d5..577b0cd 100644
--- a/public/style.css
+++ b/public/style.css
@@ -62,7 +62,7 @@ body
     line-height: var(--base-unit);
     color: var(--text-color);
 
-    width: 15cm;
+    width: 16cm;
     margin: calc(4 * var(--base-unit)) auto;
     background: var(--background-color);