Algorithms, Search, and Recommendations

The Reader and some of our emails recommend posts and websites based on a number of different algorithms.

We have two goals with the algorithms that we use:

  1. Help people find websites that they want to follow and keep up to date with.
  2. Help web site builders and owners to find an audience.

We use and test a multiple data sources for building these algorithms. Below we describe where we use your data and how.

Locations Where We Recommend Content

We recommend content in many places and use different algorithms for each:

  • Reader Search (for both posts and sites) and try to find the most popular content that matches your search. The results can also be sorted reverse chronologically.
  • Reader Post Recommendations (shown on the search page and in the main Reader stream) are made based on what you have recently liked or commented on, and by using collaborative filtering (if you and another user both liked a post, then we are more likely to recommend to posts the other user liked to you).
  • Reader “More on” Related Posts (shown at the bottom of all posts in the Reader) is mostly popular content on that is similar to the current post.
  • Reader Tag Pages list recent posts that have a particular tag. To prevent spam, posts with more than 15 tags are excluded from the tag pages.
  • Related Posts only show posts from within the same site. They are mostly based off of the title and content of the posts, but there is some boosting by likes and comments.

↑ Table of Contents ↑

Data Sources

The above algorithms are often being improved, and what content we show depends on a complex combination of factors. Here are examples of the types of information we may use to make our recommendations:

  • The title, content, tags, and categories of posts.
  • Other text from the site, such as usernames and logins; site names and the host name (
  • Total number of likes and comments.
  • Who has liked and commented on a post.
  • Total number of followers.
  • Who has followed a site.
  • How recently a post was published.
  • How often or recently a site has posted.
  • The content of what you have liked and commented on.
  • Whether posts have links, images, or videos.
  • How often a site has been rejected from being recommended in the Reader.

Content we filter out from our algorithms:

  • Sites we think may be spam.
  • Sites that have mature content.
  • Sites with potentially objectionable content.
  • Content that is not in your language.

You can view your most recent post likes at When you comment in the Reader or on any website where you use your account to comment, that comment history will be used in our recommendations. Commenting on a site with your account will look something like this:

To avoid your comments being included in our algorithms, you can comment anonymously by logging out.


Not quite what you're looking for?

Get Help