How Google Might Generate Snippets for Search Results

When you perform a search at Google, and you have a set of search results in front of you, how do you decide what to click upon? How do you judge the page titles, the snippets, and the URLs that you see. How does Google decide what to show you? A little more than a year ago, Google Webmaster Trends Analyst Pierre Far wrote on the Google Webmaster Central Blog a post titled Better page titles in search results. There he told us that Google might sometimes rewrite the titles for web pages when showing them in search results. The post told us that Google might do some changing of titles when those had generic titles such as “home”, or no title at all, or:

We use many signals to decide which title to show to users, primarily the <title> tag if the webmaster specified one. But for some pages, a single title might not be the best one to show for all queries, and so we have algorithms that generate alternative titles to make it easier for our users to recognize relevant pages.

Before we consider how Google might decide when and how to change page titles (in a follow up post to this one), there’s another question about search results that needs some exploration.

How does Google decide upon snippets for search results when it chooses snippets from the content of pages?

Sometimes Google will use the meta description created for a page as a snippet. Sometime Google will pull a sentence or some information from the content of a page instead to display to a searcher. Chances are, if a page has a meta description that is well written, includes the keyword terms or phrases the page is optimized for, and is roughly around 150 characters or so, Google will choose the meta description to display as a snippet. But not always.

Sometimes a page ranks well enough to show in search results for words other than the terms or phrases that a page is optimized for, and those words aren’t all contained within the meta description for the page. Sometimes a page’s meta description isn’t well written and doesn’t include keywords the page is optimized for either. A meta description for a page may be extremely short and not very descriptive, which would make it a poor choice as a snippet. Sometimes a meta description might be identical to every other meta description on a site. Some pages don’t even have meta descriptions. Google could even choose to use content from a page even if the words from a query appear in a meta description.

Last March, Google was granted a patent that provides some hints about when Google might choose content to display from a page, and where it might choose that text from.

If the query terms or phrases that someone searches with are word that tend to appear on pages that have abstracts or lengthy introductions, Google might decide to pull content from the start of a page if the query terms are present.

If query terms or phrases being searched for tend to appear in ranking pages that often have conclusions at the end of a page, Google might choose to pull content to display from near the end of a page. That’s what the patent tells us:

Generation of document snippets based on queries and search results
Invented by Alexandre A. Verstak and Anurag Acharya
Assigned to Google
US Patent 8,145,617
Granted March 27, 2012
Filed: November 18, 2005

Abstract

A document retrieval system generates snippets of documents for display as part of a user interface screen with search results. The snippet may be generated based on the type of query or the location of the query terms in the document. Different snippet generation algorithms may be used depending on the query type. Alternatively, snippets may be generated based on an analysis of the location of the query terms in the document.

We don’t know for certain if the approach in this patent is one that Google had adopted, or if they even used it at one point, and moved on to something new. But it’s worth digging through the patent and seeing how they might make that choice.

Multiple Snippet Generation Algorithms

The patent tells us that Google might decide which algorithm to use to decide upon a snippet based upon its perception of a query.

Google might look at the length of paragraphs that include the query terms and the distance of a paragraph from either the beginning of a document, or the end of the document.

Some documents on the Web might have a lot of metadata associated with them, such as scholarly literature that might include “names of authors, title, publisher, publication date, publication location, citation information, article identifiers such as Digital Object Identifier, PubMed Identifier, SICI, ISBN, and the like, network location (e.g., URL), number of references, number of citations, language, and the like.”

Other documents might have considerably less meta data, such as a blog or an ecommerce page.

There are potentially a couple of different types of rules that might be used to decide upon a snippet – location based rules, and language dependent rules.

Location Based Rules

These rules might be used to generate snippets based on the location of the query terms in page. A paragraph or a portion of a paragraph might be chosen as a snippet based on the length and distance of the paragraph from the start or end of a page.

Every paragraph that includes the query terms may be given a score based on the length of the paragraph and the distance of the paragraph from a predetermined location in the document, such as the beginning or the end of the document. The beginning of the page could be used in the types of documents that “include abstracts, executive summaries or comprehensive introductions” at the start of those documents.

The ends of pages might be used for other types of pages that “include a conclusion or summarization” at their end.

Language Dependent Rules

Some language rules might be used in addition to choose snippets.

Examples of language dependent rules:

  • How much of the paragraph are punctuation characters
  • Whether the paragraph ends with punctuation or proposition
  • Whether any of the words in the paragraph is overly long
  • The number of bold or italicized words in the paragraph

Some paragraphs (in part or full) that include all of the query terms might rank poorly as choices for snippets for other reasons. These paragraphs might end up with a score of 0 because they:

  • Are shorter than a certain threshold
  • Are mostly punctuation, or have punctuation aboce a certain threshold
  • Fail language specific rules
  • Contain italicized or bold words above a certain threshold
  • Are too far from the start or the end of a page, based upon query type (a query that tends to show result pages that include abstracts, or one that tends to have results with conclusions)

A snippet that is chosen might be from the paragraph with the highest score on a page, or the “first paragraph to score above a threshold amount.”

A snippet might be chosen from that paragraph, and it might be:

  • A predetermined number of words of the selected paragraph, such as the first predetermined number of words of the paragraph (e.g., 25 words)
  • The first one or more sentences of the paragraph (e.g., 3 sentences)
  • A middle portion (e.g., 50 words) containing at least one of the query terms
  • The entire paragraph

Conclusion

I started off this post with a mention that Google may sometimes change titles or snippets for pages in search results when they believed that doing so might result in more clicks through to a page when it’s shown in search results. We will get to that in the next post or two. What’s important here is that Google does have a process in place to decide where it might take text from when it shows a snippet that isn’t from the meta description on a page.

Google might choose a snippet from the paragraphs weighted from the top of a page, if the search results pages returned for the query tend to be pages that that include abstracts and meta data such as an author’s name, a publisher’s name, and so on. This could be true on a query for the name of a scholar, for instance.

Google might choose a snippet starting from paragraphs at the bottom of the page, if the search results pages returned for the query are pages that often end with conclusions.

Other signals, such as the lengths of paragraphs, amount of punctuation, bold and italics, and more can also influence the choice Google makes.

Why is the decision as to what snippet to show in search results so important?

The snippet for a page in search results may determine whether or not someone might click through from a search results page.

Share

27 thoughts on “How Google Might Generate Snippets for Search Results”

  1. Interesting stuff. I’m looking forward to the future posts you hinted at as I’m seeing lots of title changes in the SERPs lately, especially rearranging order to emphasize brands.

  2. Great stuff Bill.

    I’m in the process if writing and article about SERP engagement and looking at how to improve click through rate through optimized titles, meta, rich snippets, universal media etc. but the Google folks never want to make it that easy :-)

    We (guys I work with) started to notice the rewriting last year, easiest to spot in brand search results where often the clickable link is often far shorter than the actual title, out conclusion was it was a close a match to an definitive intent to search on a brand name, and with that focused of intent there was no need (if fact is could be perceived as an incremental distraction), so that Google was just showing the brand as the clickable link, rather than the complete title tag.

    As noted, great deep dive into how Google shows search results, believe the aim is always to present something that closer matches Google’s interpretation of the searchers intent… Improving click through signals and extrapolating to more relevant results.

    Cheers

  3. That’s good to know. Frankly speaking previously I thought Google takes snippets from meta-descp only. Yeah sometimes I used to get wrong snippets in the search results for my website. I tried to sort it out. Now I come to know the secrets.

    Bill thanks for sharing :)

  4. Hi Bill, hope you are fine and thanks for the insight. Actually, Titles are the key in the SERP’s and according to what the user sees, he will click or not. In terms of SEO, it is relevant to take time and think about the titles of his website. And or the meta description as well!

    Besides, it is true that your original title may appear differently according to the search terms. Like Jim, I recently saw many changes in the display of the titles in the SERP’s.

  5. I always thought they took from the meta-description when figuring out the snippets. Funny enough, since I’ve been monitoring my sites more closely with Webmaster Tools, I’ve noticed more consistant results; at least showing the intended results I wanted.

    Can’t wait for the next two parts! :D

  6. I’ve recently noticed Google playing around with titles of websites by including the brand name of the website at the front, something I found quite interesting. I originally noticed this for the company I work for, which is only an SME, so it’s definitely an argument against people who claim that Google are only out for the big companies, but only a small argument.

  7. Hi Bill, this is one of the best posts on snippets I’ve come across. However, when you suggest ‘ Google may sometimes change titles or snippets for pages in search results when they believed that doing so might result in more clicks through to a page when it’s shown in search results’, I find myself incredulous. That Google would do anything altruistic remains a mythological concept. They are self-serving, first and foremost…

  8. For SEO, I’ve always read (and been told) that the content at the top of the page is the most important and also that if the page is large, the end sections may not even get crawled. Perhaps I need to pay more attention to the conclusions of pages on my site!

  9. Hi Bill, I thought I understood what you were saying, but then found I didn’t really – or at least I’m a bit confused – probably my fault.

    I get it that Google might not use our skillfully crafted Meta Descriptions if the Google Search query targets keywords that are not our primary optimized keywords in an article or our Meta descriptions.

    But then you discuss characteristics that Google uses to score our First and last paragraphs – which is useful to know – thanks. But -

    I’m confused by the link between why Google might choose a different Meta description, and why Google will not choose from first and last paras with things like too many italics and bolds.

    What happens when Google wants to choose a different meta description from the one we specified, AND the first and/or last paragraph scores Zero?

    I don’t see how the one relates to the other – would Google just give up then, and not list our article at all?

  10. Hi Carolm.

    When Google decides to use content from a page instead of the meta description, it might score the different paragraphs on a page where it finds all of the query terms used to find the page.

    For queries that tend to start with things like abstracts or long introductions (as seen in pages that show up in search results), it might start exploring paragraphs starting from the top, and working their way down the page.

    For queries that don’t start off with abstracts or introductions, but end with long conclusions (again, as pages that show up in search results for that query), Google might start exploring paragraphs from the bottom of the page, and working their way up.

    The snippet Google will show won’t necessarily be from the first paragraph or the last one, but instead the one that ranks highest based upon a few different scoring rules, or the first paragraph it sees that scores above a certain threshold based on those rules.

  11. Piers, I don’t think it’s necessarily altruistic (at least toward site operators) for Google to want to increase CTRs. These are pages that Google is returning in response to a query, so they want to present them in a way that indicates that they were right to choose them. If the meta descriptions and titles we’ve written fail to demonstrate (in Google’s view) that our pages are truly relevant, Google’s going to find its own way to demonstrate it in order to let the searcher know that it knows what it’s doing.

  12. I have noticed that some times Google search result includes meta description of about 150 characters and also some times it includes 3 to 4 lines from the relevant paragraph. In fact Google is much smarter and improving the search results day by day..

  13. I am back again reading your blog. I dare say you have the best blog about SEO I have seen. I can tell you put a lot of thought into it. Thank you for your resources. Happy friday:)

  14. Great post Bill! It’s interesting that Google has decided to do this. To me it seems like these are more steps that should encourage people to build pages around themes or concepts not necessarily one keyword phrase.

  15. I appreciate the amount of effort you put into writing this post. My takeaway is that there is no way to control the snippet.

    You mentioned many variables that Google may use to pull meta… I guess the best I can do at this point is always write a keyword rich well constructed meta description and add citations when I can.

    If I missed something let me know. Thank you.

  16. On one of my old sites, Google made ​​a snippet – that I was scared to go on it …. and customers and even more so)

  17. When I do a search, I never click on the first few links and never click on sponsored links unless my search is particularly hard to find or pin down. And I try for a long tailed phrase search. It’s interesting what you said about Snippets although I’m trying to get my head around it, I just don’t seem to be getting it.

  18. That very interesting! I wonder how they determine which is going to get top spot? A rich title might have some bearing but it probably reverts back to how many links they have.

  19. Google has many, many ways to generate snippets. You may see snippets for navigational pages with small bits of every item in the list, sometime starting with one line from the meta description.
    It takes a lot of time to have those generated snippets though, but this can lead to great CTR, and give some insight on the meta description you should actually use for best results.

  20. My cousin experimented by creating a full website 8 months ago and he didn’t put any meta descriptions to test this Google’s “auto-snippets”. He didn’t actually like what Google’s doing for his pages on SERPs. So maybe the best thing to do is (still) write high-quality meta descriptions. This is the most ignored thing when creating web contents… but if you can invest some time and skill in doing this, maybe it would be worth it.

  21. Interesting topic. At least now we got an idea how it works for google everytime we hit the search button and how it does the trick. I actually have wondered before how does it work. This one is an enlightenment for all the questions in my head. Thanks for the answers.

  22. Why do we see these snippets for some results in the Google search and not for others. Does it have something to do with Structured Data or Schema ?

Comments are closed.