• Hi Takis,
    thank you for the great plugin.

    I’m trying to understand if Google assumes that slower full size rather than smaller faster adaptive images are being used on a low traffic site where it does not have real world speed data.

    To explain I see a discrepancy between https://webpagetest.org/result which indicates large full size images being served on mobile and chrome dev tools (mobile device option) which indicates, the true picture I believe, much smaller images served via the adaptive images plugin.

    I was hoping webpagetest.org would show the page serving the AI image but it doesn’t – maybe because it is reading the HTML I’m not sure.

    For this site it seems that there is not enough real world data for Google to use in it’s ranking algorithm i.e. message

    ‘Chrome User Experience Report does not have sufficient real-world speed data for this page’

    in pagespeed/insights

    So I’m now left wondering how Google will determine page speed for the purpose of ranking.

    If Google does not have real world speed data, is it using the pagespeed insights/webpagetest.org approach of reading html which is blind to the use of the adaptive image sized version? Could this account for poor ranking?

    I would be grateful if you could share any insight.
    Thank you
    Pete

    The page I need help with: [log in to see the link]

Viewing 7 replies - 1 through 7 (of 7 total)
  • Plugin Author Takis Bouyouris

    (@nevma)

    Hello, Pete,

    Thank you very much for your message!

    Well, it generally depends on the tool you use to audit your website. If the tool simply analyses your HTML code then id will have no notion of what the plugin does, because the plugin depends on the fact that a real browser will run all the code. Pagespeed does not do that. Webpagetest, on the other hand, does indeed use real machines and real browsers and, if all goes, well it should be able to detect the fact that the plugin serves downscaled images when the test is done in a mobile device with a smaller screen.

    However, I do see in your example that Webpagetest did not correctly detect the downscaled images. But I do see that your website serves them correctly: https://prnt.sc/pyxrd2/. That is not natural and has not happened in any of my tests so far. Perhaps, I need to do some more debugging there.

    The truth is that Google does not actually run any of the above tools to “rank” your website based on its speed. It does detect speed by using Analytics to measure actual timings when users visit a web page. All the above tools are available in order for one to measure and detect their optimisation issues and find ways to solve. They are not used for ranking, not even Pagespeed, as far as I know.

    And that makes sense because, on the one hand, what matters is actual speed and not just best practices and, on the other hand, not all best practices can be used on all websites, so it would be rather unfair for Google to rank this way.

    If you indeed have a generally poor ranking for specific keywords then I suggest you conduct a SEO audit on your website in order to detect all possible problems or areas that you can improve.

    Cheers,
    Takis

    Thread Starter PeteS_UK

    (@petes_uk)

    Hi Takis,

    thank you very much for your reply. I’m glad that you agree that webpagetest is not showing the AI downscaled image – I checked another page and it was the same – perhaps you could check a different site to confirm that it is webpagetest and not something strange with my AI implementation?

    I agree with what you say but I would like to know what Google uses for the speed ranking element when there is not enough data in the Chrome User Experience Report i.e for low traffic sites. I think you are suggesting that it doesn’t use the CUER data at all but collects and uses Analytics data instead, please could you confirm.

    Thank you
    Pete

    Plugin Author Takis Bouyouris

    (@nevma)

    Hello, Peter,

    As far as I know and have researched Google has not mentioned anywhere that it uses any of the known tools (like Pagespeed or Lighthouse) to measure automation. And, as I explained earlier, this does make sense, in my mind!

    What it does use is the browsers’ Javascript perfomance API, which it can invoke through Google Analytics. And there is one more argument to support that: this way it knows a lot more about all of the requests a website serves and not just on or a few.

    Also, what makes even greater impact is the fact that fast web pages create users who are happier and spend more time in one’s website and that is one of the strongest assets Google measures and has nothing to do with on-site or off-site SEO. It is the actual user behaviour towards one’s content. And speed is a major factor to affect it!

    Now, about Webpagetest, I am a little bit surprised because all my earlier tests showed that it did detect downscaled images when I set it to load them from a mobile device. I will need some time to double check this and debug it a little bit deeper though!

    Cheers,
    Takis

    Thread Starter PeteS_UK

    (@petes_uk)

    Thank you for the information Takis, what you say makes sense.

    Let me know if you find out any more.

    For info I have Really Simple SSL and Fastest Cache plugins installed – I wonder if that confuses Webpagetest – don’t know why it should, just a gut feel that something is wrong somewhere.

    Pete

    Plugin Author Takis Bouyouris

    (@nevma)

    Hello, Peter,

    Normally caching plugins do not handle images. However, if you are using a caching server (like NginX) or a CDN, then, yes, these will be handling and delivering images and they can even not let the image requests reach your website at all (for instance a CDN with a subdomain for serving static files).

    Cheers,
    Takis

    Thread Starter PeteS_UK

    (@petes_uk)

    Hi Takis
    It’s Apache server – no CDN involved.

    Let me know the result of any webpagespeed tests you do

    thanks
    Pete

    Plugin Author Takis Bouyouris

    (@nevma)

    Apache standalone should behave as designed and this is what my test in your website show, as I mentioned in my first comment (https://prnt.sc/pyxrd2/).

    I will try to dig into this a little bit more in the next weeks though!

    Cheers,
    Takis

Viewing 7 replies - 1 through 7 (of 7 total)
  • The topic ‘Is discrepancy between webpagetest and AI image a ranking issue?’ is closed to new replies.