diff --git a/tests/diff/README.md b/tests/diff/README.md
new file mode 100644
index 0000000..34c2210
--- /dev/null
+++ b/tests/diff/README.md
@@ -0,0 +1,15 @@
+`comparediffs` provides a tools for downloading a PDF from two different sources, running pdfparanoia on the files, comparing the outputs byte-for-byte,
+and reporting the results.
+
+Typical usage is to first establish two `ssh` tunnels to hosts with access to the literature, e.g. via
+`ssh -D 1080 host1` and `ssh -D 1081 host2`. You can then invoke `comparediffs` via
+
+    ./comparediffs localhost:1080 localhost:1081 < urls
+
+where urls is a file containing one URL per line (e.g. the example file in this directory).
+
+`comparediffs` creates a subdirectory `pdf/`, in which is stores PDFs. It won't try to download the same PDF twice, so if you fix pdfparanoia you'll
+need to clean out some or all of this subdirectory.
+
+It's easy to see which PDFs pdfparanoia failed on, as it leaves copies of the scrubbed files with suffixes `.1.cleaned.pdf` and `.2.cleaned.pdf`.
+When pdfparanoia succeeds (or isn't even needed, because the downloaded files were identical), the scrubbed files are removed.