Block Web Crawlers with Rails

Search engines “crawl” and “index” web content through programs called robots (a.k.a. crawlers or spiders). This may be problematic for our projects in situations such as:

• a staging environment
• migrating data from a legacy system to new locations
• rolling out alpha or beta features

Approaches to blocking crawlers in these scenarios include:

• authentication (best)
• robots.txt (crawling)
• X-Robots-Tag (indexing)

Problem: duplicate content

With multiple environments or during a data migration period, duplicate content may be accessible to crawlers. Search engines will have to guess which version to index, assign authority, and rank in query results.

For example, we periodically back up our production data to the staging environment using Parity:

production backup
staging restore production

Things search engines do

In order to provide results, a search engine may prepare by doing these things:

1. check a domain’s robots settings (e.g. http://example.com/robots.txt)
2. request a webpage on the domain (e.g. http://example.com/)
3. check the webpage’s X-Robots-Tag HTTP response header
4. cache the webpage (saving its response body)
5. index the webpage (extract keywords from the response body for fast lookup)
6. follow links on the webpage to other webpages and repeat

Steps 1, 2, 3, and 6 are generally “crawling” steps and steps 4 and 5 are generally “indexing” steps.

Solution: authentication (best)

The most reliable way to hide content from a crawler is with authentication such as HTTP Basic authentication:

class ApplicationController < ActionController::Base
  if ENV["DISALLOW_ALL_WEB_CRAWLERS"].present?
    http_basic_authenticate_with(
      name: ENV.fetch("BASIC_AUTH_USERNAME"),
      password: ENV.fetch("BASIC_AUTH_PASSWORD"),
    )
  end
end

This often is all we need for situations such as a staging environment. The following approaches are more limited but may be more suitable for other situations.

Notice we can control whether crawlers are allowed to access content via config in the environment. We can use Parity again to add configuration to Heroku staging:

staging config:set DISALLOWALLWEB_CRAWLERS=true

Solution: robots.txt (crawling)

The robots exclusion standard helps robots decide what action to take. A robot first looks at the /robots.txt file on the domain before crawling it.

It is a de-facto standard (not owned by a standards body) and is opt-in by robots. Mainstream robots such as Googlebot respect the standard but bad actors may not.

An example /robots.txt file looks like this:

User-agent: *
Disallow: /

This blocks (disallows) all content (/) to all crawlers (User-agents). See this list of Google crawlers for examples of user agent tokens.

Globbing and regular expressions are not supported in this file. See what can go in it.

Add Climate Control to the Gemfile to control environment variables in tests:

gem "climate_control"

In spec/requests/robots_txt_spec.rb:

require "rails_helper"

describe "robots.txt" do
  context "when not blocking all web crawlers" do
    it "allows all crawlers" do
      get "/robots.txt"

      expect(response.code).to eq "404"
      expect(response.headers["X-Robots-Tag"]).to be_nil
    end
  end

  context "when blocking all web crawlers" do
    it "blocks all crawlers" do
      ClimateControl.modify "DISALLOW_ALL_WEB_CRAWLERS" => "true" do
        get "/robots.txt"
      end

      expect(response).to render_template "disallow_all"
      expect(response.headers["X-Robots-Tag"]).to eq "none"
    end
  end
end

Google recommends no robots.txt if we want all our content to be crawled.

In config/routes.rb:

get "/robots.txt" => "robots_txts#show"

In app/controllers/robots_txts_controller.rb:

class RobotsTxtsController < ApplicationController
  def show
    if disallow_all_crawlers?
      render "disallow_all", layout: false, content_type: "text/plain"
    else
      render nothing: true, status: 404
    end
  end

  private

  def disallow_all_crawlers?
    ENV["DISALLOW_ALL_WEB_CRAWLERS"].present?
  end
end

If we’re using an authentication library such as Clearance site-wide, we’ll want to skip its filter in our controller:

class ApplicationController < ActionController::Base
  before_action :require_login
end

class RobotsTxtsController < ApplicationController
  skip_before_action :require_login
end

Remove the default Rails robots.txt and prepare the custom directory:

rm public/robots.txt
mkdir app/views/robots_txts

In app/views/robots_txts/disallow_all.erb:

User-agent: *
Disallow: /

Solution: X-Robots-Tag (indexing)

It is possible for search engines to index content without crawling it because websites might link to it. So, our robots.txt technique blocked crawling, but not indexing.

Adding a X-Robots-Tag header to our responses short-circuits the entire process; well-behaved crawlers won’t make HTTP requests at all to content on the domain.

You may have seen meta tags like this in projects you’ve worked on:

<meta name="robots" content="noindex,nofollow">

The X-Robots-Tag header has the same effect as the robots meta tag but applies to all content types in our app (e.g. images, scripts, styles), not only HTML files.

To block robots in our environment, we want a header like this:

X-Robots-Tag: none

The none directive is equivalent to noindex, nofollow. It tells robots not to index, follow links, or cache.

In lib/rack_x_robots_tag.rb:

module Rack
  class XRobotsTag
    def initialize(app)
      @app = app
    end

    def call(env)
      status, headers, response = @app.call(env)

      if ENV["DISALLOW_ALL_WEB_CRAWLERS"].present?
        headers["X-Robots-Tag"] = "none"
      end

      [status, headers, response]
    end
  end
end

In config/application.rb:

require_relative "../lib/rack_x_robots_tag"

module YourAppName
  class Application < Rails::Application
    # ...
    config.middleware.use Rack::XRobotsTag
  end
end

Our specs will now pass.

Conclusion

Our environment’s content can be blocked in three different ways from crawling and indexing by web robots that respect the robots exclusion standard (most importantly Google).

Use authentication to entirely hide it, or robots.txt plus the X-Robots-Tag for more granular control.