Google Announces Page PreFetching (again) - beta

[engine]

Google just announced Instant Pages which works within Chrome (Beta) to load sites almost instantly. Looks like it pre-loads the top query using pre-rendering.


Related:
Banning Google FireFox PreFetching: [webmasterworld.com...]

[edited by: Brett_Tabke at 2:11 pm (utc) on Aug 3, 2011]

[HuskyPup]

[googlewebmastercentral.blogspot.com...]

[indyank]

Why is this restricted to chrome?

[netmeg]

Cause Google wants more people using Chrome? I think they did say the technology is open for someone to make a plugin for FF if they wanted to.

[anand84]

What does it mean for website owners? Now if my site ranks number 1, will Google eat my bandwidth, and show up in analytics (because it has to fetch the pre-loaded data) even if the user has not even clicked on the link?

[indyank]

How many results will be pre-fetched? and it does look like google will artificially boost traffic to some top ranking sites, if the users don't click through those pre-fetched results.

[indyank]

just saw a few tweets that this is only applicable for organic results and adword users needn't fear the pre-fetch.

[ken_b]

What they said was something like ... [most often, just the first result].

[indyank]

Amit says that users take 9 seconds to enter the query and 15 seconds to scan the results but google takes only a second to process the results.Then how is pre-fetch useful and how does it shave off 2-5 seconds?

[indyank]

It can only help the top ranking page by cutting the time it takes for users to see that page, provided users click through the first result.Majority of the top ranking pages will benefit.

[Leosghost]

Posted this on engine's original thread , as soon as G mentioned it ..

Instant ( predictive prefetch ) "is designed not to waste bandwidth"..

But on mobiles, and in fact always .the customer pays the bandwidth..if you don't click on what goog thinks you will ..you just paid for them to load it to your mobile anyway ! ..

Ouch! for all those on mobile bandwidth caps ..

or any non "unlimited" data packages..

edit ::actually the website and the user each pay the bandwidth, so Goog are costing both of them money with this "predictive" service..Amit wouldn't smile so much if he was the one who'd be paying for this.

[trakkerguy]

It can only help the top ranking page by cutting the time it takes for users to see that page, provided users click through the first result.Majority of the top ranking pages will benefit.

Seems like a good idea to me. Good for the top result, good for users. At least in short term.

Also good for Chrome as users will notice faster page loading on search results (but don't really know why). Could lead to a significant in increase in switch to Chrome.

[netmeg]

As with most things Google, I'm walking a thin line between super cool and super creepy on this one.

[Leosghost]

btw..I'm on unlimited bandwith , mobiles and desktops, and hosting..so not affected..

but ..many will be ..

I'm wondering if it isn't a way of getting around the lag and pages loading slowly .. that g analytics and adsense code was causing ..prefetch ..and the problem doesn't go away..but if you are #1 on serps ..or you click on #1 ..it wont be so apparent ..and lest G off the "G code slows sites down " hook..

[Swanny007]

I remember using an "Internet accelerator" program that did prefetching like 10 years ago when I was on dialup... that was cool. Now having a search engine prefetch the first result? Yawn. I've got a 25Mbps connection now :-)

[londrum]

hasn't IE been doing this for years?
i seem to remember that there's a specific user_agent for it, as i added it to my .htaccess list to save on bandwidth

[deadsea]

They got to the part in the presentation where they spelled out the problem:
1) We've worked on getting the search entered faster with voice search and guessing what you are typing.
2) We've worked on our relevance algorithms and Google instant to help you choose the result faster.
3) Google's servers return results in less than a second, its negligible to the search experience.
4) When you click on a 3rd party site it takes on average 5 seconds to load.

I said "Oh no, they are going to host all content and stop referring users to 3rd party sites." I'd have to say that their actual solution to the problem is preferable.

[Leosghost]

yeah .that shot through a few minds at that point :)

[lucy24]

hasn't IE been doing this for years?
i seem to remember that there's a specific user_agent for it, as i added it to my .htaccess list to save on bandwidth

Is the said UA renting a room from google and borrowing its clothes?

Weird pair of entries from yesterday's logs (today, by local time):

119.47.nnn.nnn - - [13/Jun/2011:18:45:50 -0700] "GET /favicon.ico HTTP/1.1" 200 598 "http://www.google.co.jp/search? {snip, snip} &q= {20-word search string here}" "Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.0; {et cetera}" 
  
119.47.{same} - - [13/Jun/2011:19:01:52 -0700] "GET /favicon.ico HTTP/1.1" 200 598 "http://www.google.co.jp/search? {exactly the same stuff again}

That's all. Favicon and nothing else. But just hours earlier (yes, I dug into the raw logs and checked), someone putting on a creditable impersonation of the Googlebot-- the regular one, not the imagebot-- happened to swing by and pick up all files associated with the relevant page* ... except the favicon. Eyes peeled for the ungooglebotic detail:

66.249.71.177 - - [13/Jun/2011:14:24:55 -0700] "GET / {directory} /images/thumbs/smallcaption.jpg HTTP/1.1" 200 3111 "http://www.example.com/ {full name of the page}.html" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"

Even before seeing this discussion, the only explanation I could think of was something involving pre-caching.

btw, and in a nice instance of the right hand not knowing or caring what the left hand is doing, the imagebot drifted by within the hour to pick up the identical images all over again :)


* The search string left absolutely zero room for ambiguity. Besides, I went to google.jp and checked ;) Never been there before, so I was curious.

[bw100]

Well ... don't most users click on the first result, anyway?
So, clicking on the first result provides a great user experience (instant page: no wait).
Was there a significant problem with page load times to prompt the pre-load for the #1 result (which might involve more than one site, considering the amount of user-specific results filtering that is occurring)?
And, if the user continues down the page, clicking on results #2, #3, #4 ... which, not being pre-loaded, may have longer load times: is this creating a subtle, subliminal preference for the #1 ranked result presented?
Or, is this more smoke and mirrors?
When you consider the page 1 real estate is primarily text ads, together with that annoying cloudlet (which most users will never turn off), an increasingly limited number of sites are viewable above the fold. Google is masterfully controlling traffic and driving ad clicks.

[dstiles]

I would only expect google to use a genuine UA/IP if they were worried that webmasters may block a prefetch - which many of us do.

IF that were the case then google would be a) gaming the system and b) thieving.

If, on the other hand, they were to use proper prefetch headers that would be legitimate and blockable (they did this before with google accelerator and yes, we did block it).

However you look at it, though, they are proposing to cost site owners more - again! And if it's only for the first result, it's a gimmick. How many people actually click on the first result anyway? Especially post-panda. And as for only enabling it for their own browser - that is anti-competitive: ask MS about that one.

[austtr]

The unwritten mantra of the IT world..

Can do = must do.

Then create a rationale for why it was needed.

[AG4Life]

How would this affect bounce rate and time on site stats, especially if the users don't end up clicking on the first result? I'm hoping Google hasn't forgotten to remove pre-fetch results from these stats, but these days, you never know.

Also, what's the point of shaving off a few seconds (and only if the user click on the first result), when users have to spend ages trying to find the right result on the messed up SERPs right now (was searching for "interest rates" the other day, and the top 6 results were interest rate pages for other countries).

[Onders]

From Google blog:

"Search traffic will be measured in Webmaster Tools just like before this feature, with only results the user visited counted."

Ok - this is great.. sounds like GA has been considered.. but what about other analytics programmes? Will they be able to differentiate between an actual visit and a pre-loaded background thing which the user may not visit?

[limoshawn]

Pre-loads the top result in most searches... I bet wikipedia's servers just wet their pants!
pretty soon google's search results are going to consist of 5 ads at the top of the page and a screen shot of the wikipedia page relating to the search. And then when Knol takes off *snicker* google will be able to replace wikip.

[anand84]

Ok - this is great.. sounds like GA has been considered.. but what about other analytics programmes?

Makes me wonder if Instant Pages could actually get Google one another antitrust lawsuits. The pre-loading of first results can actually mess up not only web analytics results, but can also potentially damage revenues made from Adsense alternatives since these networks could be seeing many more impressions with more or less similar clicks (If other ad networks had a Smartpricing-like system in place, the publishers could stand to lose a lot of money thereby pushing them to choose Adsense over these rival ad networks).

Effectively, while Google says they will keep their analytics and Adsense metrics accurate, the same cannot be expected of other rival products. Definitely a candidate for antitrust.

Probably, Google should let webmasters decide whether or not they would like their pages to be 'instantized'. I would so readily take the offer.

[Onders]

Surely they would have thought this through.... do I sound like I'm convincing myself?!

[aak1962]

Gave the issue of impact of this on all analytic programs. The only ones that should be impacted are ones that still process traditional access logs.

If you're using GA, Ominture, Webtrends OnDemand, etc. where you gather data via a java script means no impact. For now at least the Google bot doesn't execute the java script nor do other search bots.

If you still process server log files, yes it will register. So Google please tell us the new bots name.

[ember]

Why does everything have to be so fast?

[Leosghost]

Because you can have "cheap or fast or quality"..

fast is easier to deliver ..especially when Google isn't paying for it ..