A current podcast by Google known as consideration to the truth that web sites are getting bigger than ever earlier than. Google’s Gary Illyes and Martin Splitt defined that the concept web sites are getting “bigger” is a foul factor will not be essentially true. The takeaway for publishers and SEOs is that Web page Weight will not be a reliable metric as a result of the reason for the “extra” weight would possibly very nicely be one thing helpful.
Web page Dimension Relies upon On What ‘s Being Measured
Google’s Martin Splitt defined that what many individuals consider as web page measurement depends upon what’s being measured.
- Is it measured by simply the HTML?
- Or are you speaking about whole web page measurement, together with photos, CSS, and JavaScript?
It’s an essential distinction. For instance, many SEOs have been freaked out once they heard that Googlebot was limiting their web page crawl to only 2 megabytes of HTML per web page. To place that into perspective, two megabytes of HTML equals about two million characters (letters, numbers, and symbols). That’s the equal of 1 HTML web page with the identical variety of letters as two Harry Potter books.
However while you embrace CSS, photos, and JavaScript together with the HTML, now we’re having a unique dialog that’s associated to web page pace for customers, not for the Googlebot crawler.
Martin mentioned an article on HTTPArchive’s Internet Almanac, which is a roundup of web site tendencies. The article gave the impression to be mixing up completely different sorts of web page weight, and that makes it complicated as a result of there are not less than two variations of web page weight.
He famous:
“See that’s the place I’m not so clear about their definition of web page weight.
…they’ve a paragraph the place they’re attempting to love clarify what they imply by web page weight. …I don’t perceive the variations in what these items are. So they are saying web page weight (additionally known as web page measurement) is the overall quantity of information measured in kilobytes or megabytes {that a} consumer should obtain to view a particular web page. In my guide that features photos and whatnot as a result of I’ve to obtain that to see.
And that’s why I used to be stunned to listen to that in 2015 that was 845 kilobytes. That to me was shocking. …As a result of I might have assumed that with photos it might be greater than 800 kilobytes.
… In July 2025, the identical median web page is now 2.3 megabytes.”
Information Will get Compressed
However that is just one solution to perceive web page measurement. One other solution to think about web page measurement is by specializing in what’s transferred over the community, which may be smaller because of compression. Compression is an algorithm on the server facet that minimizes the dimensions of the file that’s despatched from the server and downloaded by the browser. Most servers use a compression algorithm known as Brotli.
Martin Splitt explains:
“I ask this query publicly that completely different folks had very completely different notions of how they understood web page measurement. Relying on the layer you’re looking at, it will get complicated as nicely
as a result of there’s additionally compression.…So some individuals are like, ah, however this web site downloads 10 megabytes onto my disk.
And I’m like, sure. …however perhaps in case you have a look at what really goes over the wire, you would possibly discover that that is 5 – 6 megabytes, not the entire 10 megabytes. As a result of you possibly can compress issues on the community stage and then you definitely decompress them on the shopper facet stage…”
Technically, the web page measurement in Martin’s instance is definitely 5 – 6 megabytes due to compression, and it’s in a position to obtain sooner. However on the consumer’s facet, that 5 – 6 megabytes will get decompressed, and it turns again into ten megabytes, which occupies that a lot area on a consumer’s cellphone, desktop, or wherever.
And that introduces an ambiguity. Is your net web page ten megabytes or 5 megabytes?
That illustrates a wider downside: completely different individuals are speaking about various things once they discuss web page measurement.
Even extensively used definitions don’t totally resolve the anomaly. Web page weight is described as “the overall quantity of information measured in kilobytes or megabytes {that a} consumer should obtain,” however because the dialogue makes clear, there isn’t any one clear definition.
Martin asserts:
“While you ask folks what they assume, if that is massive or not, you begin getting very completely different solutions relying on how they give thought to web page measurement. And there’s no one true definition of it.”
What About Ratio Of Markup To Content material?
Some of the attention-grabbing distinctions made within the podcast is that a big web page will not be essentially inefficient. For instance, a 15 MB HTML doc is taken into account acceptable as a result of “just about most of those 15 megabytes are literally helpful content material.” The scale displays the worth being delivered.
Against this, what if the ratio of content material to markup have been the opposite method round, the place there was just a little little bit of content material however the overwhelming quantity of the web page weight was markup.
Martin mentioned the ratio instance:
“…what if the markup is the one overhead? And I imply like what do you imply? It’s like, nicely, you understand, if it’s like 5 megabytes but it surely’s solely little or no content material, is that unhealthy? Is that worse as on this case, the 15 megabytes.
And I’m like, that’s tough as a result of then we come into this bizarre territory of the ratio between content material and markup. Yeah.
And I stated, nicely, however what if loads of it’s markup that’s metadata for some third social gathering instrument or for some service or for regulatory causes or licensing causes or no matter. Then that’s helpful content material, however not essentially for the top consumer, however you continue to form of must have it.
It will be bizarre to say that that’s worse than the web page the place the load is generally content material.”
What Martin is doing right here is shifting the concept of web page weight away from uncooked measurement towards what the information really represents.
Why Pages Embrace Information Customers By no means See
A serious contributor to web page weight is content material that customers by no means see.
Gary Illyes factors to structured knowledge for example of content material that’s particularly meant for machines and never for customers. Whereas it may be helpful for engines like google, it additionally provides to the general measurement of the web page. If a writer provides loads of structured knowledge to their web page with a view to make the most of all of the completely different choices which can be obtainable, that’s going so as to add to the web page measurement although the consumer won’t ever see it.
This calls consideration to a structural actuality of the online: pages usually are not simply constructed for human readers. They’re additionally constructed for engines like google, instruments, AI brokers, and different techniques, all of which add their very own necessities to the load of an internet web page.
When Overhead Is Justified
Not all non-user-facing content material is pointless.
Martin talked about how markup might embrace “metadata” or a instrument, regulatory, or licensing objective, making a form of grey space. Even when the extra knowledge doesn’t enhance the consumer expertise straight, it does serve a objective, together with serving to the consumer discover the web page by way of a search engine.
The purpose that Martin was getting at is that these concerns of web page weight complicate makes an attempt to label web page weight pretty much as good whether it is below this threshold or unhealthy if the web page weight exceeds it.
Why Separating Content material and Metadata Doesn’t Work
One doable answer that Gary Illyes mentioned is separating human-facing content material from machine-facing knowledge. Whereas Gary didn’t particularly point out the LLMs.txt proposal, what he mentioned form of resembles it in that it serves content material to a machine minus all the opposite overhead that goes with the user-facing content material.
What he really mentioned was a solution to separate the entire machine-facing knowledge from what the consumer will obtain, thus, in principle, making the consumer’s model of an internet web page smaller.
Gary rapidly dismisses that concept as “utopic” as a result of there’ll at all times be hordes of spammers who will discover a solution to make the most of that.
He defined:
“However then sadly that is an utopic factor. As a result of not everybody on the web is enjoying good.
We all know how a lot spam now we have to take care of. On our weblog we are saying someplace that we catch like 40 billion URLs per day that’s spam or some insane quantity, I don’t bear in mind precisely, but it surely’s some insane quantity and positively billions. That can simply exacerbate the quantity of spam that engines like google obtain and different machines obtain perhaps like I might wager $1 and 5 cents that may really enhance the quantity of spam that engines like google and LLMs and others ingest.”
Gary additionally stated that Google’s expertise is that, traditionally, when you will have separate sorts of content material, there’ll at all times be variations between the 2 sorts. He used the instance of when web sites had cellular and desktop pages, the place the 2 variations of content material have been typically completely different, which in flip brought about points for search and in addition for usability when a web site ranks an internet web page for content material on one model of a web page, then sends the consumer to a unique model of the web page the place that content material doesn’t exist.
Though he didn’t explicitly point out it, that clarification of Google’s expertise might shed extra mild on why Google is not going to undertake LLMS.txt.
Because of this, engines like google have largely settled on a single-document mannequin, even whether it is inefficient.
Web site Dimension vs Web page Dimension Is the Actual World
The dialogue finally challenges the unique idea of the issue, that heavy net pages are unhealthy.
Gary observes:
“The primary query is, are web sites getting fats? I believe this query will not be even significant.
As a result of it doesn’t matter within the context of an internet site if it’s fats. Within the context of a single web page, sure.
However within the context of an internet site, it actually doesn’t matter.”
So now Gary and Martin change the main target to net pages which can be getting heavier, a extra significant method to take a look at the problem of how net pages and web sites are evolving.
This strikes the dialogue from an summary concept to one thing extra measurable and actionable.
Heavier Pages Nonetheless Carry Actual Prices
Even with sooner connections and higher infrastructure, bigger pages nonetheless have penalties, and smaller weighted pages have optimistic advantages.
Martin explains:
“I believe we’re losing loads of assets. And I imply we, we had that in one other episode the place we stated that we all know that there are research that present that web sites which can be sooner have higher retention and higher conversion charges. Yeah. And pace is partially additionally primarily based on measurement. As a result of the extra knowledge I ship, the longer it takes for the community to really switch that knowledge and the longer it takes for the processor of no matter machine you’re on to really course of it and show it to you.”
From a broader perspective, the problem is not only efficiency however effectivity. As Illyes places it, “we’re losing loads of assets.”
The net could also be getting heavier, however the extra essential takeaway is why. Pages are carrying extra than simply user-facing content material, and that design selection shapes each their measurement and their affect.
Featured Picture by Shutterstock/May_Chanikran
