Tuesday, November 01, 2011 at 3:53 PM
We generally advise to use GET for fetching resources a page needs, and this is by far our preferred method of crawling. We’ve started experiments to rewrite POST requests to GET, and while this remains a valid strategy in some cases, often the contents returned by a web server for GET vs. POST are completely different. Additionally, there are legitimate reasons to use POST (e.g., you can attach more data to a POST request than a GET). So, while GET requests remain far more common, to surface more content on the web, Googlebot may now perform POST requests when we believe it’s safe and appropriate.
We take precautions to avoid performing any task on a site that could result in executing an unintended user action. Our POSTs are primarily for crawling resources that a page requests automatically, mimicking what a typical user would see when they open the URL in their browser. This will evolve over time as we find better heuristics, but that’s our current approach.
Let’s run through a few POSTs request scenarios that demonstrate how we’re improving our crawling and indexing to evolve with the web.
Examples of Googlebot’s POST requests
- Crawling a page via a POST redirect
<form name="foo" action="request.php" method="post"> <input type="hidden" name="bar" value="234"/>
- Crawling a resource via a POST XMLHttpRequest
In this step-by-step example, we improve both the indexing of a page and its Instant Preview by following the automatic XMLHttpRequest generated as the page renders.
- Google crawls the URL, yummy-sundae.html.
- Google begins indexing yummy-sundae.html and, as a part of this process, decides to attempt to render the page to better understand its content and/or generate the Instant Preview.
- During the render, yummy-sundae.html automatically sends an XMLHttpRequest for a resource, hot-fudge-info.html, using the POST method.
This page is about a yummy sundae.
- The URL requested through POST, hot-fudge-info.html, along with its data payload, is added to Googlebot’s crawl queue.
- Googlebot performs a POST request to crawl hot-fudge-info.html.
- Google now has an accurate representation of yummy-sundae.html for Instant Previews. In certain cases, we may also incorporate the contents of hot-fudge-info.html into yummy-sundae.html.
- Google completes the indexing of yummy-sundae.html.
- User searches for [hot fudge sundae].
- Google’s algorithms can now better determine how yummy-sundae.html is relevant for this query, and we can properly display a snapshot of the page for Instant Previews.
General advice for creating crawlable sites is found in our Help Center. For webmasters who want to help Google crawl and index their content and/or generate the Instant Preview, here are a few simple reminders:
- Prefer GET for fetching resources, unless there’s a specific reason to use POST.
You can test whether resources are blocked through Webmaster Tools “Labs -> Instant Previews.”
To verify that you're not accidentally cloaking, you can use Instant Previews within Webmaster Tools, or try setting the User-Agent string in your browser to something like:
Mozilla/5.0 (compatible; Googlebot/2.1;
- Remember to include important content (i.e., the content you’d like indexed) as text, visible directly on the page and without requiring user-action to display. Most search engines are text-based and generally work best with text-based content. We’re always improving our ability to crawl and index content published in a variety of ways, but it remains a good practice to use text for important information.
If you’d like to prevent content from being crawled or indexed for Google Web Search, traditional robots.txt directives remain the best method. To prevent the Instant Preview for your page(s), please see our Instant Previews FAQ which describes the “Google Web Preview” User-Agent and the nosnippet meta tag.
We’ll continue striving to increase the comprehensiveness of our index so searchers can find more relevant information. And we expect our crawling and indexing capability to improve and evolve over time, just like the web itself. Please let us know if you have questions or concerns.