diff --git a/doc/migrating_from_nokogiri.md b/doc/migrating_from_nokogiri.md
new file mode 100644
index 0000000..883eb70
--- /dev/null
+++ b/doc/migrating_from_nokogiri.md
@@ -0,0 +1,169 @@
+# Migrating From Nokogiri
+
+If you're parsing XML/HTML documents using Ruby, chances are you're using
+[Nokogiri][nokogiri] for this. This guide aims to make it easier to switch from
+Nokogiri to Oga.
+
+## Parsing Documents
+
+In Nokogiri there are two defacto ways of parsing documents:
+
+* `Nokogiri.XML()` for XML documents
+* `Nokogiri.HTML()` for HTML documents
+
+For example, to parse an XML document you'd use the following:
+
+    Nokogiri::XML('<root>foo</root>')
+
+Oga instead uses the following two methods:
+
+* `Oga.parse_xml`
+* `Oga.parse_html`
+
+Their usage is similar:
+
+    Oga.parse_xml('<root>foo</root>')
+
+Nokogiri returns two distinctive document classes based on what method was used
+to parse a document:
+
+* `Nokogiri::XML::Document` for XML documents
+* `Nokogiri::HTML::Document` for HTML documents
+
+Oga on the other hand always returns `Oga::XML::Document` instance, Oga
+currently makes no distinction between XML and HTML documents other than on
+lexer level. This might change in the future if deemed required.
+
+## Querying Documents
+
+Nokogiri allows one to query documents/elements using both XPath expressions and
+CSS selectors. In Nokogiri one queries a document as following:
+
+    document = Nokogiri::XML('<root><foo>bar</foo></root>')
+
+    document.xpath('root/foo')
+    document.css('root foo')
+
+Oga currently only supports XPath expressions, CSS selectors will be added in
+the near future. Querying documents works similar to Nokogiri:
+
+    document = Oga.parse_xml('<root><foo>bar</foo></root>')
+
+    document.xpath('root/foo')
+
+Nokogiri also allows you to query a document and return the first match, opposed
+to an entire node set, using the method `at`. In Nokogiri this method can be
+used for both XPath expression and CSS selectors. Oga has no such method,
+instead it provides the following more dedicated methods:
+
+* `at_xpath`: returns the first node of an XPath expression
+
+For example:
+
+    document = Oga.parse_xml('<root><foo>bar</foo></root>')
+
+    document.at_xpath('root/foo')
+
+By using a dedicated method Oga doesn't have to try and guess what type of
+expression you're using (XPath or CSS), meaning it can never make any mistakes.
+
+## Retrieving Attribute Values
+
+Nokogiri provides two methods for retrieving attributes and attribute values:
+
+* `Nokogiri::XML::Node#attribute`
+* `Nokogiri::XML::Node#attr`
+
+The first method always returns an instance of `Nokogiri::XML::Attribute`, the
+second method returns the attribute value as a `String`. This behaviour,
+especially due to the names used, is extremely confusing.
+
+Oga on the other hand provides the following two methods:
+
+* `Oga::XML::Element#attribute` (aliased as `attr`)
+* `Oga::XML::Element#get`
+
+The first method always returns a `Oga::XML::Attribute` instance, the second
+returns the attribute value as a `String`. I deliberately chose `get` for
+getting a value to remove the confusion of `attribute` vs `attr`. This also
+allows for `attr` to simply be an alias of `attribute`.
+
+As an example, this is how you'd get the value of a `class` attribute in
+Nokogiri:
+
+    document = Nokogiri::XML('<root class="foo"></root>')
+
+    document.xpath('root').first.attr('class') # => "foo"
+
+This is how you'd get the same value in Oga:
+
+    document = Oga.parse_xml('<root class="foo"></root>')
+
+    document.xpath('root').first.get('class') # => "foo"
+
+## Modifying Documents
+
+Modifying documents in Nokogiri is not as convenient as it perhaps could be. For
+example, adding an element to a document is done as following:
+
+    document = Nokogiri::XML('<root></root>')
+    root     = document.xpath('root').first
+
+    name = Nokogiri::XML::Element.new('name', document)
+
+    name.inner_html = 'Alice'
+
+    root.add_child(name)
+
+The annoying part here is that we have to pass a document into an Element's
+constructor. As such, you can not create elements without first creating a
+document. Another thing is that Nokogiri has no method called `inner_text=`,
+instead you have to use the method `inner_html=`.
+
+In Oga you'd use the following:
+
+    document = Oga.parse_xml('<root></root>')
+    root     = document.xpath('root').first
+
+    name = Oga::XML::Element.new(:name => 'name')
+
+    name.inner_text = 'Alice'
+
+    root.children << name
+
+Adding attributes works similar for both Nokogiri and Oga. For Nokogiri you'd
+use the following:
+
+    element.set_attribute('class', 'foo')
+
+Alternatively you can do the following:
+
+    element['class'] = 'foo'
+
+In Oga you'd instead use the method `set`:
+
+    element.set('class', 'foo')
+
+This method automatically creates an attribute if it doesn't exist, including
+the namespace if specified:
+
+    element.set('foo:class', 'foo')
+
+## Serializing Documents
+
+Serializing the document back to XML works the same in both libraries, simply
+call `to_xml` on a document or element and you'll get a String back containing
+the XML. There is one key difference here though: Nokogiri does not return the
+exact same output as it was given as input, for example it adds XML declaration
+tags:
+
+    Nokogiri::XML('<root></root>').to_xml # => "<?xml version=\"1.0\"?>\n<root/>\n"
+
+Oga on the other hand does not do this:
+
+    Oga.parse_xml('<root></root>').to_xml # => "<root></root>"
+
+Oga also doesn't insert random newlines or other possibly unexpected (or
+unwanted) data.
+
+[nokogiri]: http://nokogiri.org/