SEO Test: How Does Personalized Search Work on Google?
For about three weeks I have performed a very simple SEO test to determine the influence of the newly introduced personalized search on Google.
It allegedly works with cookies only and it personalizes your search results by default even when you’re not logged in.
Google claimed that the sites you visit via their search results most will show up higher up in search results the next time.
How Does Personalized Search Work? How Does it Impact Google Results?
I wanted to test whether and how personalized search works on Google. Yet the results of my SEO test failed to prove that personalized search had a measurable impact.
Default personalized didn’t work for me.
That’s the more than surprising result of my empirical test. What did I test? How exactly did I test? Read on to find out!
To test the impact of personalized search or personal search as I prefer to call it for the sake of simplicity I took the searcher’s perspective. I did not yet want to test the SEO consequences of personalized search.
I just wanted to see how much and how quick the results get personalized and whether the actions I’d perform would lead to the personalization I’d expect.
My assumptions where threefold:
- Clicking a website’s search results repeatedly in a matter of days or weeks will make it climb in my personal Google results.
- Searching for a particular keyphrase will push the matching result of the site I click upwards significantly.
- Once the results get personalized in favor of the site I use to click they will stay the way and the results won’t drop significantly.
All three assumptions have been proven wrong by my test. Personalized search used in the way described above was useless and offered no additional value for the searcher.
How did I test?
- I chose one keyphrase, [seo news] (without brackets) and one site to click the search result of, Search Engine Land.
- I chose SEL for several reasons. This is a highly reputable authority site with stable rankings. This way I wanted to ensure that random movements don’t spoil the results of my test. Also Search Engine Land offered SEO news without really ranking for the term correctly. You’d expect SEL to be at least in the Top 10 in Google.com for seo news but it ranked at #15 on the first day of my test.
- To make sure to be unbiased I tested with a browser I don’t use normally. A browser that I use only for web development to cross check or some private web browsing. This way I wanted to ensure that the sites I click on in search results usually don’t meddle with the results.
- To be certain that Google counts the clicks I did not bounce but stayed on SEL for a while and often read some articles clicking on the respective pages.
- The test took approx. 3 weeks and I took 10 screen shots to document it. Take note that I made a mistake once taking a screen shot of the wrong set of results, page one, where the Search Engine Land results can’t be seen.
- I clicked only the SEL result each time. On the first day I mistakenly clicked Search Engine Watch as well but that didn’t happen anymore. I wanted the test to stay as simple as possible. Thus I ignored all the results.
- Cookies were enabled with no exception.
- I didn’t log in to Google once during the testing time period on the browser I tested.
- I added a date to the file name of the screen shots: It started 20100116 (January 16th, 2010).
On the screen shots I can see how Google highlights different portions of text. On some days articles including the term SEO or news get highlighted while on others it does not or different ones.
My impression is that the on page factors (mentions of the keywords) had a much higher impact on the results than personalization if there was any at all.
The actual results of the personalized Google search of “seo news” (without quotes) and Search Engine Land:
* Screen shot of the wrong page
While on the other hand you see any upwards movement of plus 4 in three weeks there was a seemingly complete random drop of minus 20 in the middle of the test.
So the rankings were quite unstable for no apparent reason other than maybe the keyword density on the front page.
You could argue now hat the personalization works to some extent as the result I’ve clicked ten times moved upwards albeit slowly.
Imagine the average searcher: Does this kind of personalization really help? Do you have to click more than 10 times in three weeks or over a longer period of tie to see any significant personalization?
From the searcher perspective the personalization did not seem to work or have a useful impact on the actual search results.
Does the Personalization Work on Google? For Me it Didn’t!
My goal was to see Search Engine Land in the top 10. I wanted to be able to see SEL on top for the query [seo news]. It didn’t happen. Not even close!
Does Google take more than clicks into account for default personalized search to personalize them a bit? Just clicks didn’t work.
I would be maybe more positive about the results but the huge drop of minus 20 positions rendered all the previous personalization useless if there ever was one.
On that day I really had to check the Google cookies: They were still there and the date of last visited was correct. So they got found and updated. There is no doubt about that.
I’ve proven other Google statements wrong. They’ve indexed “nofollow” links from Twitter for instance before they officially bought the data from Twitter.
Also they claimed to filter #hashtag spam in Google real time results but I was able to place spammy three hashtags in a row for “hot searches”.
Also remember that Google will discontinue some search features without telling you or only telling you afterwards, like they did with PageRank sculpting support. They have announced it months after they had discontinued it.
Personalized Search on Google Since the Test
OK, this test happened more than a decade ago by now. Many things happened ever since. Let me recount a few of them!
Google tried and tested many social and personalized features along the way.
The most important one was probably Google+. Authorship markup was another milestone. Both of them have been discontinued ever since, among many others.
Google+ certainly impacted your search results. Things and sites you added a +1 to aka voted up were showing higher than others.
As long as you were logged in Google also took your search history into account.
Over the years Google increasingly got criticized for creating a filter bubble due to the personalization. Especially privacy-focused search engine DuckDuckGo (which is based on Bing results mostly) was adamant on how Google promoted bias during elections e.g.
Due to scrutiny by lawmakers Google backtracked publicly and claimed that search results do not get personalized significantly.
Google claimed that search results remain neutral and the impact of personalization is negligible. At that time it really wasn’t. By now (2022) it seems that personalization is really low key again.
What Google results were customized for and still are is your location though! It’s called localization then.
Depending on the language you use and the country you search from you get very different results even when searching on Google.com (not even a local Google version like Google.de).
It’s very hard to view the original unbiased search results on Google. They differ from place to place.
Even when you use a VPN (Virtual Private Network) to hide your true location you just get the localized results for the fake location the VPN server or IP suggests.
Great test and write up! I’ve been experimenting a bit too and noticed that keywords that bring up forum posts (think long tail searches) tend to be influenced by personalised search. The indented posts themselves are switched around within the site’s listing. Leading me to think that keywords that are more likely to be part of repeat search behaviour patterns are more likely to trigger personalised search. Searches for niche help topics, for example. It certainly doesn’t seem to be on by default for all searches as you rightly point out.
“my empirical test” – You’re joking, right? Where’s your control? Where’s your large dataset? Where’s your elimination of outlier results?
“I’ve proven other Google statements wrong” – no, all you’ve done is shown what an exceptionally poor grasp you have on how to conduct proper testing and derive true empirical results from them.
I believe you can make inferences, but as you noted with news related sites there could be too many on-site factors changing regularly to use as a test.
Joanna: Yes indeed the impact on long tail queries seems to be more visible. That’s what many people have observed. Thank you for your feedback.
Hey Brent: Make sure you don’t mix up the keyphrase I’ve used with the topic of the site in question. I’ve meant that the keywords that appeared on the site had more impact than the actual personalization.
Barry: You don’t use the term “empirical” right. It means “information gained by means of observation, experience, or experiment.” (Wikipedia). It doesn’t mean computed by thousands of drones and overseen by a team of scientists. Also I’ve written right at the beginning that it was just a “simple SEO test”.
Do you want to say that just because the test was based solely on my own subjective measurement it proves me wrong? So in other words you suggest that Google personalization works fine? Or what’s the point of your comment beyond to show off how condescending you are?
Last but not least I’ve said literally “I’ve proven other Google statements wrong.”
Make sure you read closely before you start dismissing work by others.
I thought “personalized” in this context meant that you can influence your personal serps by (re)moving results. Nothing to do with clicking on them.
I comefrom a country and live in a remote part at that where there are no specialist doctors.
So, my son who is 9 years now and my wife who is around 35 years always go to the Homeo Doctor and don’t try Alopathy unless it is too serious or urgent.
I am sorry – Homeopathy is a real gem of medicinal discipline.
@onreact – I amno SEO expert or involved in the profession – but frequent on most SEO blogs frequently – read posts – like a journeyman.
I have been hearing about this Google’s claim for almost 2 years – some call it personalized search and some like to call it behavioral search!
Although I am not an SEO – I do try to work on my own sites/blogs and have seen success more than failures. Towards, my working, I do a lot of work on SERP results of my targeted keywords – but never noticed any improvement (noticed decline only) on my results.
I see SEL @ 11 also – so it was not your personalized search that it improved your result – but it has been a natural fluctuation instead.
I don’t see the reasons for attacking @onreact regading post nider it’s title. On contrary, I see those testing as valuable in understanding Google personalized more. Now we see it isn’t so simple and that is good. In some of previous posts here on blog it was a presuption that Google personalized can make search process boring and this test give us reason to belive it will not be so boring in that way. And that SEO will take it’s place in future, too.
Great illustrations and research.
Great hypothesis, method, and documentation for your final conclusion.
Most probably Google guys know what they are doing with this new personalized search updates. Hope that your test is true because for my “new” sites wouldn’t be so good for strong implementation of new type of personalized search.Nice post!
Thanks for sharing your testing idea! I sure learned a couple of new things to apply on my own site.
I think each person will experience differently, some may experience it, but some others may not experience
I am new at this online business but I am sure of one thing…in order to have a valid or appropriate conclusion one should do series of tests and different test subjects.
It’s just a thought. Who am I to tell you, you are the pro and I am just browsing.
I’ve done something similar to your test and found Personalized search has no influence in ranking links! But I’ve not taken this as confirmation as I believe there should be something I should have missed! And now it’s really great to know that there is no worry with Google’s personalized search! I always ask my team guys to remove personalized search in their browsers. But your research has cleared my confusions. Thanks for sharing!
What seemed to be an interesting article has turned into a personal feud about the term empirical and the the scientific relevance of it.
I personally found it quite interesting in what onreact has posted here and even though his findings might not be 100% waterproof it gave me food for thought and I will definitely look into it more.
Thanks for sharing your testing idea! Great illustrations and research.Great hypothesis, method, and documentation for your final conclusion.Good work.Keep it up.