Googlebot and Document.Write

With JavaScript/AJAX being used to place dynamic content in pages, I was wondering how Google indexed web page content that was placed in a page using the JavaScript "document.write" method. I created a page with six unique words in it. Two were in the plain HTML; two were in a script within the page document; and two were in a script that was externally sourced from a different server. The page appeared in the Google index late last night and I just wrote up the results.

Nonsense words?

[Whiney Mac Fanboy (963289)]

An alert came in in the late evening of March 10th for "zonkdogfology", one of the words in the first pair

zonkdogfology is a real word:

zonk-dog-fol-o-gy zohnk-dog--ful-uh-jee
noun, plural -gies.

1. the name given to articles from zonk where the summary makes no sense whatsoever.

Serious question now - is the author of the article worried that the ensuing slashdot discussion will mention all his other nonsense words? I've no doubt slashdotters will find & mention the other words here, polluting google's index....

Re:Nonsense words?

[by Anonymous Coward]

zonkdogfology is a real word:

It's a perfectly cromulent word, and it's use embiggens all of us.

The Results:

[XanC (644172)]

Save a click: No, Google does not "see" text inserted by Javascript.

Re:The Results:

[temojen (678985)]

And rightly so. You should be hiding & un-hiding or inserting elements using the DOM, never using document.write (which F's up your DOM tree).

How does document.write mess up your DOM tree?

[catbutt (469582)]

I don't believe you.

Re:How does document.write mess up your DOM tree?

[XanC (644172)]

If you're using document.write, you're writing directly into the document stream, which only works in text/html, not an XHTML MIME type, because there's no way to guarantee the document will continue to be valid.

In this day and age, document.write should never be used, in favor of the more verbose but more future-proof document.createElement and document.createTextNode notation.

Re:How does document.write mess up your DOM tree?

[jesser (77961)]

Perhaps more importantly, document.write can't be used to modify a page that has already loaded, limiting its usefulness for AJAX-style features.

Re:How does document.write mess up your DOM tree?

[hackstraw (262471)]

One of the most clever uses of document.write I've seen was something like: document.write("<--") YOU NEED JAVSCRIPT FOR THIS PAGE document.write("--&gt")

Re:

[ultranova (717540)]

Except that the programmer might know what they're doing. But I guess we're getting past the point of trusting people more than machines ;)

Based on all the segfaults, blue screens of death, X-Window crashes, Firefox crashes, code insertion bugs et cetera I've seen, I'd say that no, in general programmers don't know what they're doing, and certainly shouldn't be trusted to not fuck it up. The less raw access to any resource - be it memory or document stream - they are given, the better.

Re:

[Sancho (17056)]

How should 'major', though? When most Firefox-borked sites were coded, Firefox probably had less than 5% (around what Safari had, last I heard). Is 5% enough to overlook? What about 3%? 1%?

If you code to the standard, at least you can blame browsers for their broken implementation.

True

[XanC (644172)]

I doubt Google will notice DOM-created elements, either. But the author should re-test with that. And I would suggest that he post the result only if it turns out Google can see that, because we all assume it can't.

google.com/?q=slashdotting+in+google+dollars

[kale77in (703316)]

I think the actual experiment here is:

• Create a 6-odd-paragraph page saying what everybody already knows.
• Slashdot it, by suggesting something newsworthy is there.
• Pack the page with Google ads.
• Profit.

I look forward to the follow-up piece which details the financial results.

Re:google.com/?q=slashdotting+in+google+dollars

[Scarblac (122480)]

Exactly, this is the typical sort of fluff that Digg seems to love. As far as I know, Slashdot had avoided this particular type of adword blog post crap until now.

Re:

[caluml (551744)]

But with the Firehose, Slashdot will now start using the "wisdom" of crowds to produce the same pap that Digg does.
Shall we all migrate to Technocrat, anyone? It has decent stories.

Re:

[dr.badass (25287)]

As far as I know, Slashdot had avoided this particular type of adword blog post crap until now

It used to be that the web as a whole avoided this crap. Now, it's so easy to make stupid amounts of money from stupid content that a huge percentage of what gets submitted only even exists for the money -- it's like socially-acceptable spam. Digg is by far the worst confluence of this kind of crap, but the problem is web-wide, and damn near impossible to avoid.

Google Pigeon technolog

[sdugoten2 (449392)]

The Google Pigeon [google.com] is smart enough to read through Document.write. Duh!

If they weren't, then they're trying

[AnonymousCactus (810364)]

Google needs to consider script if they want high-quality results. Besides the obvious fact that they'll miss content supplied by dynamic page elements, they could also sacrifice page quality. Page-rank and the like will get them very far, but an easy way to spam the search engines would be to have pages on a whole host of topics that immediately get rewritten as ads for Viagra as soon as they're downloaded by a Javascript-aware browser. It's interesting to know the extent to which they correct for this.

Of course, there are much more subtle ways of changing content once it's been put out there. One might imagine a script that waits 10 seconds and then removes all relevant content and displays Viagra instead. Who knew web search would be restricted by the halting problem? I wonder how far Google goes...

Re:

[TubeSteak (669689)]

Page-rank and the like will get them very far, but an easy way to spam the search engines would be to have pages on a whole host of topics that immediately get rewritten as ads for Viagra as soon as they're downloaded by a Javascript-aware browser.

Of course, there are much more subtle ways of changing content once it's been put out there. One might imagine a script that waits 10 seconds and then removes all relevant content and displays Viagra instead.

Google tends to nuke those sites from orbit once it disc

Re:If they weren't, then they're trying

[gregmac (629064)]

You have to also remember though, that often the content generated dynmically is going to be of no use to a search engine, it will often be user-specific - there's obviously some reason it's being generated that way.

And if pages are designed using AJAX and dynamic rendering just for the sake of using AJAX and dynamic rendering.. well, they deserve what they get :)

How did this make the front page?

[by Anonymous Coward]

It should be pretty obvious that no search engine should interpret javascript, let alone remotely sourced javascript. I was actually hoping this guy would show me wrong and demonstrate otherwise, but to my disappointment this was just another mostly pointless blog post.

Re:

[Jake73 (306340)]

Yeah, I was kinda shocked, really. I always wondered how people with bad blogs were able to break into the mainstream and gather regular readers. I guess they just try like hell to get picked up on Slashdot/Digg/etc with some worthless blog post.

Re:

[NoTheory (580275)]

Why should it? I mean, isn't the point of javascript content that responds dynamically to the intentions of an agent? The googlebot, although an extremely complicated AI agent, isn't intentional. It doesn't know what it's doing on a site, and so i figure probably shouldn't just be allowed out to wreak havoc. Also, wouldn't that allow one an opportunity to fork-bomb the googlebot then as well?

Re:

[zobier (585066)]

Also, wouldn't that allow one an opportunity to fork-bomb the googlebot then as well?

JavaScript doesn't have fork AFAIK.

The setTimeout function can do a similar thing.

Re:

[Bitsy Boffin (110334)]

From memory, setTimeout forms a time-delayed but synchronous entry into the execution stream, you will not get two threads in the same javascript code pile running simultaneously, the timeout will not fire until the execution stream is idle.

Re:

[VGPowerlord (621254)]

Because JavaScript can create content. Since 99% of people run with it enabled, they will see this content, so it makes sense to index it.

Did you know that 99% of all statistics are made up?

I can source some Javascript statistics: W3Schools reports [w3schools.com] that, as of January 2007, 94% of their audience has Javascript turned on, a significantly lower statistic than you are reporting. Not only that, but it is actually the highest percentage since they started recording them binannually in late 2002.

It's a moot poi

Re:

[vidarh (309115)]

Because doing so without massive limitations would involve the halting problem. A search engine simply CAN'T determine whether a certain piece of javascript will terminate in the general case. In lots of special cases, yes (such as when there's no control constructs, or the control constructs can't possibly cause loops or recursion etc.) and they could use timeouts etc. or only execute the first "n" steps of an interpreter, yes. But all of it would mean essentially crippling the feature.

And for what? So t

Google request external JavaScript file?

[JAB Creations (999510)]

Check your access log to see if Google actually requested the external JavaScript file. If it didn't there would be no reason to assume Google is interested in non-(X)HTML based content.

Re:

[The Amazing Fish Boy (863897)]

I have actually seen some reports [google.com] of a "new" Googlebot requesting the CSS and Javascript. The rumour I heard was that it was using the Gecko rendering engine or something along those lines. This was some time ago. I'm not sure what ever became of this.

Doesn't work; Good (kind of)

[The Amazing Fish Boy (863897)]

FTFA:

Why was I interested? Well, with all the "Web 2.0 technologies that rely on JavaScript (in the form of AJAX) to populate a page with content, it's important to know how it's treated to determine if the content is searchable.

Good. I am glad it doesn't work. Google's crawler should never support Javascript.

The model for websites is supposed to work something like this:

• (X)HTML holds the content
• CSS styles that content
• Javascript enhances that content (e.g. provides auto-fill for a textbox)

In other words, your web page should work for any browser that supports HTML. It should work regardless of whether CSS and/or Javascript is enabled.

So why would Google's crawler look at the Javascript? Javascript is supposed to enhance content, not add it.

Now, that's not saying many people don't (incorrectly) use Javascript to add content to their pages. But maybe when they find out search engines aren't indexing them, they'll change their practices.

The only problem I can see is with scam sites, where they might put content in the HTML, then remove/add to it with Javascript so the crawler sees something different than the end-user does. I think they already do this with CSS, either by hiding sections or by making the text the same color as the background. Does anyone know how Google deals with CSS that does this?

Re:

[doormat (63648)]

I thought I remember a while ago about some search engine using intelligence to ignore hidden text (text with the same or a similar color as the background). Of course the easy work around for that is to use an image for your background and then that may fool the bot, but who knows, they could code to accomidate that too.

Regardless, I'm pretty sure you'd get banned from the search engines for using such tactics.

Re:

[zobier (585066)]

I thought I remember a while ago about some search engine using intelligence to ignore hidden text (text with the same or a similar color as the background). Of course the easy work around for that is to use an image for your background and then that may fool the bot, but who knows, they could code to accomidate that too.

You could use OCR to detect that (and to index images used for text content).

Re:

[cgenman (325138)]

In other words, your web page should work for any browser that supports HTML. It should work regardless of whether CSS and/or Javascript is enabled.

Define "work". A web page without formatting is going to be useless to anyone who isn't a part-time web developer. To them, it's just going to be one big, messy looking freak out... akin to a television show whose cable descrambler broke. Sure all the "information" is there, somewhere, but in such a horrible format that a human being can't use it.

Web pages ar

Re:

[WNight (23683)]

I don't know about you, but I write my webpages so that when the style goes away, the page still views in a basic 1996 kind of style. Put the content first and your index bars and ads last then use CSS to position them first, visibly. This way if a blind user or someone without style sheets sees the site it at least reads in order.

Re:

[caluml (551744)]

View, Page Style, No Style in Firefox will show you what your page looks like to browsers/spiders.

Re:

[The Amazing Fish Boy (863897)]

Define "work". A web page without formatting is going to be useless to anyone who isn't a part-time web developer.

How's this? Disable CSS on Slashdot. First you get the top menu, then some options to skip to the menu, the content, etc. Then you get the menu, then the content. It's very easy to use it that way.

To them, it's just going to be one big, messy looking freak out... akin to a television show whose cable descrambler broke. Sure all the "information" is there, somewhere, but in such a horrible

Re:

[Animats (122034)]

The model for websites is supposed to work something like this:

If only. Turn off JavaScript and try these sites:

• Ford Motor Company [ford.com]
• Jeep [jeep.com]
• Credit Suisse [credit-suisse.com]

Re:

[catbutt (469582)]

Who's talking about Java?

Re:Doesn't work; Good (kind of)

[VGPowerlord (621254)]

In actuality, it says "Use a text browser such as Lynx to examine your site, because most search engine spiders see your site much as Lynx would. If fancy features such as JavaScript, cookies, session IDs, frames, DHTML, or Flash keep you from seeing all of your site in a text browser, then search engine spiders may have trouble crawling your site." – Webmaster Guidelines [google.com], Technical Guidelines section, bullet point 1.

Re:

[Rakishi (759894)]

Huh? He's talking about browser generated content, most dynamic content is server side generated (like slashdot but I think slashdot may have flat files as cache for speed reasons). No one said that nice xml file can't be generated by the server when the page is called.

I would make normal links, then use JS on top

[The Amazing Fish Boy (863897)]

So, what do you have to say about websites that have their entire user-interfaces built with content that gets filled by javascript asynchronously from a single html page?

If I understand you, you something like this: The site has two parts, a menu and content. When you click a menu item, rather than being taken to a new URL, it executes Javascript which fetches only the new content from the web server, then replaces the content section. So the URL doesn't change.

It's a nice improvement. Less bandwidth used, and a quicker interface.

Unfortunately, it's not often done right. The way I would do it is to first make the menu work like it normally would. Make each menu item a link to a new page. Then you apply Javascript to the menu item. Something like this:

// menuLink is the DOM element for each menu link.
// (i.e. get it from document.getElementById(), etc.)
menuLink.onclick = function() { getNewContent(); return false; }

(FYI, this is how I do pop-up windows, too.)

Putting it behind a login screen doesn't solve all the problems. You're right that it won't be searchable anyway, but people with older browsers or screen readers won't be able to access it.

I think Gmail actually offers two versions. One for older browser that uses no (or little?) Javascript, and the other which almost everyone else (including me) uses and loves. But I'm not sure how easy it would be to maintain two versions of the same code like that. I also don't think it's nice for the end user to have to choose "I want the simple version", though it may encourage them to update to a newer browser, I guess.

(Of course this is all "ideally speaking", I realize there are deadlines to meet and I violate some of my own guidelines sometimes. I still think they're good practices, though.)

Re:

[The Amazing Fish Boy (863897)]

I agree with you that with a small enough user base (or one that is adequately controlled), you can cut some corners, especially if time is a constraint. Generally I would say whenever your users could reasonably demand their browser work. That is, if the site is going to be publicly accessible, I would not make Javascript a requirement. I'm not sure what the actual "limit" I would put on the number of users would be; I think that would vary from project to project.

It's a matter of project requirements,

Accessibility?

[BladeMelbourne (518866)]

The bottom line is your web sites should probably degrade nice enough when JavaScript is not enabled. It might not flow as nice, the user may have to submit more forms, but the core functionality should still work and the core content should still be available.

DDA / Section 508 / WCAG - the no JavaScript clause makes for a lot of extra work - but it is one that can't be avoided on the (commercial) web application I architect. (Friggin sharks with laser beems for eyes making lawsuits and all.)

Document.write() is not the way to go

[Max Romantschuk (132276)]

Document.write() is executed as the page loads. Most AJAX-style implementation rely on either the innerHTML-property or creating nodes through the DOM. Testing these would tell us much more than testing Document.write().

From TFA:

[bennomatic (691188)]

So, some friends and I have been bantering back and forth about how Google treats content that has been inserted into a page using Javascript. So I decided to do an experiment. This page has six nonsense words. Two are hardcoded into the page via straight HTML. Two are inserted via Javascript, but the script is part of the page HTML. The last two are inserted via Javascript, but the script is on a remote server. The purpose of the test is to see three things... * The time lapse between when the words

(tagging beta)

[Jack Schitt (649756)]

I predict that from now on, zonkdogfology will be a common tag for all articles that relate to google search...

Google doesn't, but it's possible

[Animats (122034)]

I'd thought Google would be doing that by now. I've been implementing something that has to read arbitrary web pages (see SiteTruth [sitetruth.com]) and extract data, and I've been considering how to deal with JavaScript effectively.

Conceptually, it's not that hard. You need a skeleton of a browser, one that can load pages and run Javascript like a browser, builds the document tree, but doesn't actually draw anything. You load the page, run the initial OnLoad JavaScript, then look at the document tree as it exists at that point. Firefox could probably be coerced into doing this job.

It's also possible to analyze Flash files. Text which appears in Flash output usually exists as clear text in the Flash file. Again, the most correct approach is to build a psuedo-renderer, one that goes through the motions of processing the file and executing the ActionScript, but just passes the text off for further processing, rather than rendering it.

Ghostscript [ghostscript.com] had to deal with this problem years ago, because PostScript is actually a programming language, not a page description language. It has variables, subroutines, and an execution engine. You have to run PostScript programs to find out what text out.

OCR is also an option. Because of the lack of serious font support in HTML, most business names are in images. I've been trying OCR on those, and it usually works if the background is uncluttered.

Sooner or later, everybody who does serious site-scraping is going to have to bite the bullet and implement the heavy machinery to do this. Try some other search engines. Somebody must have done this by now.

Again, I'm surprised that Google hasn't done this. They went to the trouble to build parsers for PDF and Microsoft Word files; you'd think they'd do "Web 2.0" documents.

Re:

[dargaud (518470)]

OCR is also an option. Because of the lack of serious font support in HTML, most business names are in images. I've been trying OCR on those, and it usually works if the background is uncluttered.

Yes, and it should work like that too. If the background is so cluttered as to make the OCR difficult, then chances are the human will have trouble reading it too. I suggested that during a job interview witha *cough* serious search engine: use a secondary crawler reporting as a normal IE/firefox, load a page usin

Re:

[VGPowerlord (621254)]

Again, I'm surprised that Google hasn't done this. They went to the trouble to build parsers for PDF and Microsoft Word files; you'd think they'd do "Web 2.0" documents.

Does Google run macros in Word documents? No? Then why are you even comparing this? I can parse a PDF document or a Word document without having to have a script interpreter running.

I imagine that the Googlebot crawler is a rather simplistic program that only knows how to:
1. Read robots.txt
2. Read meta tags (robot tags in particular)
3. Fi

If you want to see

[BrynM (217883)]

If you want to see through a search engine's eyes, open the page in Lynx [browser.org]. The funniest part about showing that method to another developer is when they think Lynx is broken because the page is empty. "It didn't load. How do I refresh the page? This browser sucks." Heh. Endless fun.

(method does not account for image crawlers)

Re:

[DrXym (126579)]

I do this also on my pages. I also mangle certain things like affiliate links, AdSense ids etc. not for any particular reason except I don't like the idea of any search engine inadvertantly indexing them.