Is An Improved Version of Agent Rank Returning to Google?

Sometimes, I run across a patent that provides details on things that Google might do, but only hints at whether or not it might actually be implemented. A few years back in 2007, I wrote about a Google patent for Agent Rank, which described reputation scores for authors (to be used as an alternative to PageRank), and looked like an important part of Google’s social network, Google+. It was referred to in the patent as “Agent Rank” and people commenting upon it started referring to it as “Author Rank”.

It seemed like it was a good description of how some people whom you may have connected with in Google+ were showing up in response to queries they had some expertise within. There may have been issues with Google’s version of Agent Rank that the search engine wanted a second bite at. Google has since removed the Photos that were showing up for authors whom you might be connected to, who may have been highly ranked, seemingly based upon a reputation score, for a topic related to a query that you might perform.

There were people who wrote that while this authorship markup was removed, and author photos associated with it, the author rank scoring system that came with it was still around, like this Search Engine Land article: Google Authorship May Be Dead, But Author Rank Is Not.

A short time after Google stopped having people add authorship markup to pages, and showing author badges next to content from people whom you had connected with in Google+, I ran across a patent that hinted that Google would still show content from people whom you might be connected with in Google+ on topics that those authors might be considered “Authoritative” for a specific topic. I wrote about that patent in a post titled, Has Google Decided that you are Authoritative for a Query?

Today, Google was granted a patent that appears to be related to that one, which shares 4 inventors with that authoritative query patent.

The images from the patent show off profile images and snippets of “prominent people in response to specific queries, which look like they are related to the topic of a query, rather than just a profile image of a person whom you may have connected with. This appears to be an improvement over the Agent Rank approach, because it provides information as to why someone might be displayed as having some level of expertise in response to a query that you perform. That’s much better than just an image of someone whom you may have connected with in Google’s social network, and randomly placed within a circle.

It’s worth looking carefully at this patent, and seeing if it is a followup to Agent Rank or Author Rank, where the rankings of pages in search results might be partially based upon a reputation score associated with an individual on a specific topic.

Authoritative Users and Trigger Query Data

The patent starts off by defining the purpose of search engines, as identifying resources relevant to a searcher’s needs and presenting information about those resources in a way which might be most useful to that searcher.


The patent uses language that we’ve seen in another recent patent from Google, such as one I wrote about in the post How Google May Trigger Answer Box Results For Queries. It refers to receiving trigger query data that identifies one or more trigger queries and one or more sets of authoritative users, each set of those authoritative users are associated with a respective trigger query, The combinations of authoritative users and trigger queries produces a snippet based upon user data. So, when a query from a searcher might result in search results from authoritative users being returned, profiles of those people might show up along with snippets about them.


The patent tells us that the advantages in using it include:

  • Displaying authoritative users and associated snippets in search results enables searchers to discover and connect with other searchers.
  • This connection of people is intended to improve user engagement with search services
  • The snippets shown are relevant to the search query and can be based on recent information
  • The snippets may look as if they were provided from a larger body of text associated with the authoritative user and/or the search query

The patent is:

Generating snippets for prominent users for information retrieval queries
Invented by Bogdan Dorohonceanu, John E. Saalweachter, Kumar Mayur Thakur, Sheng Zhang
Assigned to: Google
US Patent 9,087,130
Granted July 21, 2015
Filed: October 4, 2012


Implementations include receiving trigger query data, the trigger query data identifying one or more trigger queries and one or more sets of authoritative users, each set of authoritative users being associated with a respective trigger query, providing a plurality of trigger query and authoritative user pairs, each trigger query and authoritative user pair identifying a trigger query and an authoritative user from a set of authoritative users associated with the trigger query, for each trigger query and authoritative user pair: generating a snippet based on user data, the user data being associated with the authoritative user in one or more computer-implemented services, each snippet being specific to the trigger query and specific to the authoritative user, and storing one or more snippets in computer-readable memory, each snippet being associated with the trigger query and the authoritative user for which the snippet was generated.

A Snippet Presents a Difference

One place where this differs from the agent rank patent is in the display of a snippet. A snippet is a group of text that provides a pseudo-biography of an authoritative user with respect to a search query. The snippet may include partial sentences, or sentence fragments, making it look like a pseudo-biography, as opposed to a biography with complete sentences.

These snippets appear in search results where an authoritative user is shown with a response to a particular query. The patent defines an authoritative user as “an expert, on one or more topics that can be associated with one or more queries.”

This system of rating a person based upon authoritativeness may include each profile section that “reflects the importance of the particular profile section to other profile sections.” For example, they could be ranked in this order, or others: “occupation, education, work experience, personal biography, hobbies, employment and places lived.”

The digital content posts that an authoritative user makes on the social network can be in the form of posts made on the social network, responses to others posts, comments on their own posts or others posts as well

The snippets that show up together with an image of a person could come from any of this digital content from the authoritative user from within the social network.

That digital content might be processed, to clean it up before it’s used as a snippet. This can include “removing redundant punctuation, offensive words, hyperlinks and the like. It might be sorted based upon the importance of the different profile sections, or may come from one of the users digital content.

Potential Snippets may include snippet pieces that might be relevant to a particular triggering query.

The patent provides some examples of these snippet pieces.


Synonyms might be used from these different snippet pieces to build better snippets:



The related patent I mentioned above told us that Google may associate particular queries with particular authoritative users. This patent doesn’t tell us how those associations are made, but rather just focuses upon the snippets that might be shown with these users. Those snippets provide information that show off an expertise in the query these authoritative users are shown for.

I searched to see if there seemed to be any related patents at the USPTO that might be related, and show off more aspects of the process described in these patents. I didn’t see any, but it’s possible they are there.

The Agent Rank patent provided a lot of information about the possibility of the use of a reputation score being used to rank content created by those users. These patents don’t tell us about how those might be ranked, but tells us about how potential snippet content from a person’s profile or content that they post might be ranked to use in snippets for different queries

These patents may indicate a second attempt to present something like Agent Rank, but Google seems to be not presenting too much of the processes behind the scenes. The kind of reputation scoring found in the original agent rank patent provided a way to rank content based upon reputation scores that might be associated with writers, and their expertise on topics. That is an interesting way to rank and order content found on the Web, and it would be nice to see that approach kept with this version of Authoritative Users of a social network.

Article Name
Is An Improved Version of Author Rank Returning to Google?

17 thoughts on “Is An Improved Version of Agent Rank Returning to Google?”

  1. Thanks Bill! Yet another useful article. Specially the “Snippets” part. The way you write makes complicated stuff so much easier to understand.
    Thanks again, and greetings from Denmark.

  2. I allways need a big cup of coffee, a cigar and time to think about it, every time I read a post here. And that’s a reason for coming back, time and time again. 🙂

  3. The way this is presented reminds me of, and looks a lot more like, the “People Who Know” feature that Bing introduced back in 2012. Remember that?

    At the time I wrote “As social results are not included within the main search results there is less of a need to establish an explicit authorship structure” so Bing was effectively giving us a version of Author Rank that relied on a combination of authors producing consistent output and social signals to determine what we might like to see without restricting itself to any single identity scheme.

  4. Hi Colin,

    That’s definitely a good reason to look at patents from the different search engines, to see how they might approach the same problems, because chances are they’ll come out with different approaches. I don’t remember the people who Know feature, but I just found an interesting patent to write about for today that it might be worth comparing to, so thank you! It is interesting seeing Bing following in the footsteps of Google with something similar to Agent Rank.

  5. Hi Andreas,

    I do try to find stuff that makes me think, and share it to make others think about it, too. I’m working on some stuff that hopefully will have us all asking a lot of questions about how best to apply and use it., which should be fun. 🙂

  6. With Google’s crack down on spammy bloggers and articles, this makes sense that Google would want to identify authors that know their stuff. I wonder how big of a role G+ will play in this if this is rolled out?

  7. Google may have removed the author pictures from search results but it would be naive to think that google does not do any kind of author ranking behind the scenes.

  8. I have been long waiting for this. The internet business is going over the roof. Everybody expects high quality rich articles. But when we have to bargain for payroll, for us writers, it is really depressing when we have to go through long interviews to prove ourselves.

    At last, may be google agent rank now can help us!!!

  9. Author rank was both exciting and frustrating as Google Plus attracted more and more SEOs. frustrating because their was far too much speculation based on something I always thought had some inherent problems in practice (easy to abuse). Exciting because if we could all build expertise instead of links for expertise it seems a better way to rank well.

    Long live Agent Rank. Shall we call it 2.0 if there is enough buzz about the new version with rich snippets?

  10. Hi Deepak,

    Without those author pictures, many believed that author ranking gone for good. I’m encouraged to see signs that it may still be around.

  11. Hi Jose,

    I think agent rank has potential to benefit people who write on the Web, and help them earn more. We’ll see if things evolve that way. I’m hoping they do.

  12. Hi Eric,

    I’m convinced that Agent Rank is less prone to abuse than linking is.

    I haven’t seen any actual signs on pages that it is back, but I hope it does return.

  13. Google did dabble with using Authors as a ranking signal some years back (I believe before G+).
    They also worked on approaches to identify relationships between content and likely authors.

    If you look at the life of Authorship – Google had enough data to mine and form algorithms to refine their ability to associate content to authors.
    This may mean that they don’t really need the authorship markup.

    If you look at the structure/pattern of social interactions, they are similar to link graphs (just often more frequent/rapid, and often overlap).
    They could easily use it as another layer of personalisation – or they could aim higher and push certain authors higher in the SERPs if they are recognised as being authoritative.

    If Google did this incrementally – no one would likely spot it unless they were continuously testing and tracking.

Comments are closed.