The privacy-focused search engine DuckDuckGo conducted its own study about Google Search’s filter bubble being real.
DuckDuckGo says Google Search is still using user data even when users are logged out of their Google accounts, and when searching via “incognito mode”, which most people believe means “searching anonymously”.
The results of DuckDuckGo’s study showed that Google Search is giving most people unique search results even if those people search for identical search terms (e.g. “gun control”, “immigration”, “vaccinations”) at the same time while in “incognito mode”.
The results from Google Search on incognito mode for the term “vaccinations” had the most variants with 73 for 87 participants. The term “gun control” followed that with 62 variants, and “immigration” with 57 variants.
The term “gun control” resulted in 19 domain names ordered in 31 different ways. There were also 22 domain names that showed up on Google search result for the term “vaccinations”, while there were 15 for the term “immigration”.
According to DuckDuckGo, the order of those links is very important because it influences how people would choose what link to click on. Also, each link on the search result gets twice as many clicks as the link after it.
Search results for Google’s news and video carousels showed different sources for different people as well.
DuckDuckGo said that the variations in search results mean that Google is using personal information, like user’s search and purchase history. According to DuckDuckGo, Google does this to show personalized search results to people which Google think they would most likely click on.
Google’s public liaison of search, Danny Sullivan, reponded to DuckDuckGo’s study on Twitter saying:
[Start of response]
Over the years, a myth has developed that Google Search personalizes so much that for the same query, different people might get significantly different results from each other. This isn’t the case. Results can differ, but usually for non-personalized reasons. Let’s explore.
The assumption is that results have been customized in some way based on information unique to an individual, such as search history. FYI: we do not personalize search results based on demographic profiles nor create such profiles for use in Google Search.
Personalization doesn’t happen often and generally doesn’t dramatically change search results from one person to another. It is usually so lightly applied that the results are very similar to what someone would see without personalization.
Anyone who wants to test how lightly personalization is applied can easily check this themselves. Do a search in a fresh “Incognito” or similar private browsing window and you’ll have no account-based activity that is used. You can compare to a regular search.
Anyone who doesn’t want personalization using account-based activity can disable it using the Web and App Activity setting. You can also choose to keep your search history stored but exclude Chrome and app activity (https://myaccount.google.com/activitycontrols/search?pli=1)
As said, personalization doesn’t dramatically change results. So why might two different people searching for the same thing see results that are different? That’s often due to non-personalized reasons: location, language settings, platform and the dynamic nature of search.
One of the most common reasons results may differ between people involves localized results, when listings are customized to be relevant for anyone in a particular area. Localization isn’t personalization because everyone in the same location gets the same results.
Localization is extremely useful in making results more relevant. For example, people in the US searching for “football” do not generally want UK football results, and vice versa. People searching for “zoos” in one area often want locally-relevant listing.
Another reason results may differ is when they are customized to be relevant for anyone searching in a particular language. This also isn’t personalization because everyone gets the same language results.
As with localization, language customization helps make results more relevant. Someone who searches in French generally wants information written in French. Similarly, those searching in English generally want information written in English.
Results might also differ slightly because of how dynamic and distributed our search system is. Our various data centers are constantly being updated with the latest information we have gathered from trillions of pages, along with our latest ranking improvements. But while we make changes as simultaneous as possible, some new or changed pages might not appear in all data centers immediately, or our latest ranking improvements may not have fully rolled out. These factors, not personalization, can create slight or brief differences in results.
Time is also a factor. Those who search a few hours or even minutes apart may see variations as new and updated material is added to our search engine. In particular, our “Top stories” section, which is never personalized, can change significantly in a short period of time.
Results may also differ slightly because of platform. On mobile, we prefer fast-loading, mobile-friendly pages, if content on those is as good or better than desktop versions. If you’re using Android or iOS, we may list apps that are appropriate for the respective devices.
[End of response]
According to DuckDuckGo, if search results are totally anonymous when using “incognito mode”, queries for the “same terms” at the “same time” should be similar.
The different results of Google Search could not be explained by changes in time, location, by being logged in to Google, or by Google testing algorithm to a small subset of users.
Because of that, DuckDuckGo said Google filters search results no matter what browsing mode you’re using. And they said this proves Google’s filter bubble is real.
People should not be lulled into a false sense of security that so-called ‘incognito’ mode makes them anonymous. Unfortunately, this is a common misconception as websites use IP addresses and browser fingerprinting to identify people that are logged out or in private browsing mode.
Tailored search results are alarming especially when people are searching about politics, according to DuckDuckGo because undecided voters use search engines to know more about the candidates. DuckDuckGo said:
If they’re getting information that is swayed to one side because of their personal filter bubbles, then this can have a significant effect on political outcomes in aggregate.
You can read the full details of DuckDuckGo’s study about how Google is influencing what you click on its search results, here: spreadprivacy.com/google-filter-bubble-study