4

Google's John Mueller now says that Googlebot renders the pages it crawls to determine what content is on them. They do this because so many sites are now heavily JavaScript centric and this allows them to better understand AJAX content.

I was also thinking that they might be using this combat a historical weakness in their PageRank algorithm. The original PageRank algorithm passed an equal amount of juice to each of the links on the page. Later I saw some evidence links higher on the page (in source code order) might pass more PageRank than lower links (say in the footer). Now that they are rendering pages they could use that rendereing to pass more PageRank to links:

  • Above the fold
  • That are larger (take up more rendered pixels)
  • In places that are more likely to be clicked (heatmap)

It would allow them to easily ignore links that would otherwise be spammy:

  • Rendered off the page
  • Hidden with CSS

This tool would allow Google to implement the reasonable surfer model of interlinked web pages. This would do a better job of identifing high quality content than the random surfer model on which the original PageRank algorithm was based.

Is there any evidence that Google is now using rendered link prominence as a signal to determine how much PageRank to pass across each link on a page?

  • I will check this out, however, my understanding all these years is that "juice" (and I hate that word as it applies here) or value passed by the link is constant across all links on the page but that links higher up on the page or high within content blocks (header tags and content between header tags as an example) are considered more important for weight. That would mean that searches that might match a link text/URL/URI would rank higher if considered more important but the link value passed to any page would remain constant. There are two mechanisms at play here. – closetnoc May 14 '15 at 18:07
  • Remember that rendering is an effort to do two things: one, capture all the content that is being served via new technologies that would have previously been unseen; and two, get more a humans perspective of the page. I rather suspect that much of the mechanics are the same between rendering and previous efforts setting aside advances as a result of rendering. I see rendering for a human view as primarily checking for content above the fold, too many ads served, bad site development, and so on. Fairly basic normal stuff. There may be more to it these days. I have not looked in a while. – closetnoc May 14 '15 at 18:18
  • I rather suspect that there is a whole laundry list of uses for rendering that Google is taking advantage of. I guess I'm a bit old school for still thinking about PageRank, but I may make some links on my sites more prominent to get them noticed by Googlebot better. – Stephen Ostermiller May 14 '15 at 18:22
  • You are right that there is likely a laundry list, but I am not sure too many of the items are sinister at all. It is often that people take a darker view of Googles motives than actually exists. In fact, as clever as Google is, people often give it far too much credit for being technically advanced rather than simple and effective. I would not be surprised if the links that are considered more important for search can potentially pass more value to the target page. It certainly would be an advantage thinking in those terms even if it does not!! It could not hurt (much)! – closetnoc May 14 '15 at 18:44
2

I suspect you are simply misunderstanding the "renders the pages" phrase. Googlebot stills views pages as raw text, it doesn't see pages as "text rendered into pixels". Googlebot is running the javascript/css/whatever it finds to discover(aka "render") text that was previously not visible to it in the raw html text source (ie because it was pulled in via AJAX calls).

You can experiment yourself using the Fetch as Googlebot tool https://support.google.com/webmasters/answer/6066468?hl=en&ref_topic=6066464

If you browse with js disabled for a while you'll notice many sites simply display nothing or garbage. Google wants to be able to index those sites properly and thus has no choice but to run at least some js.

Consider that it is very much in Google's financial interests to get people to present content as text to them. OCR and machine vision are expensive operations.

to address your comments

I've seen evidence that Google is indexing text that is only produced by JavaScript.

Yes, they are definitely doing that. But, bear in mind they admit they cannot run all js. If they were actually rendering their crawl data to pixels exactly like a browser then there would be no limitations on what js they could run. Therefore they must be using algorithms that understand the visual impact of "most" js but not all of it.

Also, Google's mobile algorithm and associated tool clearly shows that they are looking at rendered pixels to make decisions that affect ranking. One warning from the mobile friendly tool is that two links are rendered too close together.

It must be possible to calculate the position and sizes of all those elements without actually going through the process of converting positional element data to pixels.

The "mobile-friendly" metrics the smartphone googlebot appears to be using are ...

  • viewport configured

  • distance between touch elements

  • font size

  • flash usage

  • content size relative to viewport

These all seem easily calculated using just numbers as opposed to analysing a rendered bitmap.

In addition, "Fetch as Google" in webmaster tools now has a "Fetch and Render" option in addition to the usual "Fetch" option. It will then show a screenshot of your page pixel for pixel as rendered by Googlebot plus another screenshot rendered as a browser user agent.

The tool has to be visual to be of any use to human webmasters. They state the main purpose of the tool is to discover if extra resources such as js and css are blocked for the googlebot. If the little thumbnail looks different to what you expect your site to look like, then they invite you to check your robots.txt to see if it is blocking resources from the "smartphone googlebot". Just becuase the tool is visual doesn't mean the normal googlebot activity operates in exactly the same way. They specifically call it a "simulation".

to summarise and paraphrase MY understanding of your question

"is there evidence of googlebot using these examples of ranking signals on the assumption that they are only available if googlebot is rendering to pixels which I assume it's now doing"

I cannot credibly answer a definitive "no" as only the opposite could be proven if evidence was found and published. But I would address the assumptions and intent of your question with these points

I think the assumption googlebot is rendering to pixels is unfounded and unnecessary.

Even if googlebot is rendering to pixels there is no indication they are being used for your suggested ranking signals.

Even if your suggested ranking signals are being used there might be no need for googlebot to render to pixels to detect them.

Even if someone proved your suggested ranking signals were being used that would not prove googlebot was rendering to pixels.

Some hopefully useful references to consider

http://www.googlewebmastercentral.blogspot.ie/2014/05/understanding-web-pages-better.html

".....help our algorithms understand that the pages are optimized for mobile."

https://support.google.com/webmasters/answer/6066467?rd=1

"It’s a small-scale simulation of the real deal"

and

"fetch and render mode tells Googlebot to crawl and display your page as browsers would display it to your audience"

  • I've seen evidence that Google is indexing text that is only produced by JavaScript. I can see passwords generated by JavaScript in the search results for my password creator site. If I then search for those passwords, I find they are indexed. – Stephen Ostermiller May 20 '15 at 16:23
  • Also, Google's mobile algorithm and associated tool clearly shows that they are looking at rendered pixels to make decisions that affect ranking. One warning from the mobile friendly tool is that two links are rendered too close together. – Stephen Ostermiller May 20 '15 at 16:24
  • In addition, "Fetch as Google" in webmaster tools now has a "Fetch and Render" option in addition to the usual "Fetch" option. It will then show a screenshot of your page pixel for pixel as rendered by Googlebot plus another screenshot rendered as a browser user agent. – Stephen Ostermiller May 20 '15 at 16:42
  • Thanks for expanding your answer to address my points, its worth an up vote now. – Stephen Ostermiller May 20 '15 at 23:43
  • thanks, they were good points and addressing them improved the answer i think – John McNamara May 28 '15 at 18:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.