Google is the king of the internet. When they push out a major algorithm update, webmasters race to optimize their websites. Whether it was adopting SSL, or making websites faster — Google decides.

They don’t publish their exact ranking factors, which leads people to wonder whether or not Google takes into account political affiliation. (They shouldn’t.)

And so, I decided to set up an experiment.

To begin, I turned off search personalization and searched Google for the terms “trump news” and “politics” and extracted the first 20 URLs for the first query, and first 10 for the second, shown below.

“trump news”

https://www.nbcnews.com/politics/donald-trump/democrats-aren-t-smart-enough-read-trump-s-tax-returns-n994291
https://www.cnn.com/2019/04/14/politics/rick-scott-trump-sanctuary-cities/index.html
https://thehill.com/homenews/sunday-talk-shows/438809-sanders-congress-not-smart-enough-to-look-through-trumps-taxes
https://www.cnn.com/videos/politics/2019/04/14/rick-scott-trump-immigration-sot-sotu-vpx.cnn
https://www.usatoday.com/story/news/nation/2019/04/14/trump-newsom-california-battles-changes-first-100-days/3353383002/
https://www.cnn.com/videos/politics/2019/04/14/april-ryan-stephen-miller-warning-upstage-trump-ebof-vpx.cnn
https://www.cnn.com/videos/politics/2019/04/14/rick-scott-trump-immigration-sot-sotu-vpx.cnn
https://www.cnn.com/videos/politics/2019/04/14/jerry-nadler-trump-contempt-for-law-sot-sotu-vpx.cnn
https://www.msnbc.com/am-joy/watch/white-house-reportedly-proposed-releasing-migrants-in-pelosi-s-district-1491068995800
https://www.nbcnews.com/nightly-news/video/trump-s-9-11-tweet-on-rep-ilhan-omar-draws-condemnation-from-democrats-1491120707941
https://www.msnbc.com/11th-hour/watch/trump-and-washington-brace-for-the-release-of-mueller-s-report-next-week-1490755651659
https://www.cnn.com/videos/politics/2019/04/12/federal-judge-rebukes-donald-trump-segregationist-era-de-vogue-sot-nr-vpx.cnn
https://www.cnn.com/videos/politics/2019/04/13/ilhan-omar-trump-violence-response-steve-rogers-ndwknd-vpx.cnn
https://www.youtube.com/watch?v=rAjkUuTTUd0
https://www.youtube.com/watch?v=Lrryep_qUzs
https://www.nbcnews.com/politics/donald-trump/democrats-aren-t-smart-enough-read-trump-s-tax-returns-n994291
https://www.cnn.com/2019/04/14/politics/jerry-nadler-cnntv/index.html
https://www.cnn.com/2019/04/14/politics/rick-scott-trump-sanctuary-cities/index.html
https://abcnews.go.com/Politics/sanctuary-cities-plan-choice-option-table-sarah-sanders/story?id=62383549
https://www.foxnews.com/politics/trump-duels-over-twitter-with-oakland-mayor-over-immigration-policies

“politics”

https://www.nbcnews.com/politics/meet-the-press/biden-bernie-lead-early-polls-don-t-place-bets-yet-n996761
https://www.nytimes.com/2019/04/20/us/politics/biggest-stories-of-the-week.html
https://www.nbcnews.com/politics/meetthe-press/biden-bernie-lead-early-polls-don-t-place-bets-yet-n99676
https://www.staradvertiser.com/2019/04/21/editorial/on-politics/pro-development-forces-flex-muscle-in-hawaii-politics/
https://www.staradvertiser.com/2019/04/21/editorial/on-politics/pro-development-forces-flex-muscle-in-hawaii-politics/
https://www.cnbc.com/2019/04/20/avengers-endgame-the-not-so-hidden-marvel-environmental-politics.html
https://www.cnbc.com/2019/04/20/avengers-endgame-the-not-so-hidden-marvel-environmental-politics.html
https://www.nytimes.com/2019/04/21/world/asia/sri-lanka-religion-christians.html
https://www.nytimes.com/2019/04/21/world/asia/sri-lanka-religion-christians.html
https://www.foxnews.com/politics/trump-pelosi-and-other-us-lawmakers-express-condolences-condemn-sri-lanka-bombings

For the first search, the following domains appeared in the following order:

Next, I added each news outlet’s bias, according to Media Bias Fact Check, along with each domain’s Moz authority (higher means more likely to rank).

Notice the absence of conservative sites, such as New York Post, RealClearPolitics, and Washington Times, despite the NYPost having a higher D.A. than NBC and also reporting on Trump’s taxes.

ABSENT DOMAINS

Other very reputable domains, such as Reuters (DA: 95, Bias: LEAST) that also published news stories about Trump didn’t appear either.

For the second query, I did the same.

Based on each page’s DA, Fox News (Right) should have easily outranked Star-Advertiser (LEFT-CENTER).

Next, to factor Moz DA into the data, I gave each search position and bias a score, which I then multiplied together and divided by Moz DA to determine the Bias Offset. Numbers rounded to nearest hundredth.

Position Rankings:

Rank on Google: 1 — Score: 20

Rank on Google: 2 — Score: 19

Rank on Google: 3 — Score: 18

etc…

Bias Multiplier

RIGHT: +2

RIGHT-CENTER: +1

LEAST: 0

LEFT-CENTER: -1

LEFT: -2T

Results

“trump news”

-.22

-.4

-.19

-.34

-.16

-.29

-.27

-.25

-.21

-.4

-.09

-.15

-.05

-.08

-.06

-.2

+.1

Total: -3.26

“politics”

-.22

-.20

-.20

-.21

-.19

-.15

-.14

-.13

+.21

Total: -1.23

( – = Liberal, + Conservative)

As you can see, liberal news websites are significantly more dominant in Google search results, and, based on the data, tend to outrank similar conservative news websites.

Google Vs. DuckDuckGo

Being privacy-focused and completely non-personalized, DuckDuckGo is generally considered the most “unbiased” search engine. Here’s how it stacks up for “trump news.”

“trump news”

https://www.cbsnews.com/news/trump-climate-change-trump-tower-and-other-buildings-could-go-green-under-new-york-city-climate-mobilization-act/
https://www.marketwatch.com/(S(rnrsydaynixa5x55oiibxm45))/story/trumps-easter-morning-a-swarm-of-tweets-followed-by-church-and-brunch-2019-04-21?link=MW_home_latest_news
https://www.washingtonpost.com/politics/trump-closes-florida-vacation-with-easter-service/2019/04/21/36ae2ed8-6442-11e9-a1b6-b29b90efa879_story.html
https://www.politico.com/story/2019/04/21/rudy-giuliani-mueller-report-rebuttal-1284140
https://nypost.com/2019/04/21/trump-touts-new-york-times-op-ed-saying-media-owes-him-an-apology/
https://www.dailymail.co.uk/news/article-6944865/Trump-attends-Easter-service-Melania-Tiffany.html?ns_mchannel=rss&ns_campaign=1490&ito=1490
https://www.dailymail.co.uk/news/article-6944951/Trump-claims-Sri-Lanka-terrorist-attacks-killed-138-MILLION-people-botched-tweet.html?ns_mchannel=rss&ito=1490&ns_campaign=1490
https://www.rgj.com/story/news/world/2019/04/21/sri-lanka-explosions-donald-trump-says-millions-dead-easter-blasts/3533292002/
https://ktla.com/2019/04/21/as-russia-probe-began-trump-called-on-spy-chiefs-for-help/
https://www.bostonglobe.com/2019/04/21/muellertrump/LYrjoYSZhcTMJgdmNqRkHO/story.html
https://gazette.com/elections/trump-and-colorado-republicans-it-s-complicated/article_91f1fc5e-6161-11e9-b6ae-0701b6e686df.amp.html
http://feeds.latimes.com/~r/MexicoStories/~3/Hb7Tgmy7p_k/la-oe-kristian-yemen-trump-saudi-arabia-veto-20190418-story.html
http://www.chicagotribune.com/ct-trump-giuliani-russia-mueller-20190421-story.html
https://www.politico.com/story/2019/04/20/mueller-report-trump-romney-1283781
https://thehill.com/homenews/administration/439910-trump-attends-easter-church-services
https://www.oaoa.com/news/us_news/article_b29f5a2a-f604-59a5-920f-dead3271941e.html?mode=image&photo=10
https://www.salon.com/2019/04/20/trump-muscles-in-on-independent-government-agencies_partner/
https://www.dailymail.co.uk/news/article-6942439/Trump-lashes-nasty-Mueller-report-again.html?ns_mchannel=rss&ito=1490&ns_campaign=1490
https://m.huffpost.com/us/entry/us_5cbac302e4b06605e3eed15c
https://thehill.com/homenews/administration/439876-trump-hits-romney-for-mueller-criticism

Besides having more variety, DuckDuckGo’s results are a lot more evenly spread out.

While their ranking factors are probably different, based on one search results, DuckDuckGo is a lot more unbiased than Google — and Google is skewed towards liberal websites.

Conclusion

From the data, we can conclude that the Google search results were favoring liberal websites.

It is also possible to conclude that DuckDuckGo results have a greater political variety compared to Google.

A WIRED article provides a case for Google not being politically biased, but it doesn’t explain the absence of conservative websites appearing.

Our advice? Don’t believe everything you see. When in doubt, cross check the information and consider what a viewpoint is missing.

Methodology Explained

During the study, I considered a variety of factors, including bias, ranking position, DA, and number of appearances. As outlined below, there are some potential flaws and inaccuracies.

Importance:

Rank Position: Users are more likely to click on higher ranking links, making higher rankings more valuable and contested.

Bias: The purpose of the study was to determine the political bias of results in Google. I used data from MediaBiasFactCheck.com.

Domain Authority: To account for other ranking factor, I used Moz’s Domain Authority Metric. In general, when the same content is published at the same time, the website with a higher D.A. will rank higher.

Bias Offset

The Bias Offset factored in each website’s Moz DA, Bias, and Search Position. Higher ranking websites were given higher scores since their impact is greater (users are more likely to click on it). The Bias Multiplier adjusts the initial rank value to include bias.

A #1 ranking website with Left bias (20*-2 = -40) will have a higher impact on the Bias Offset than a #2 ranking website with Left bias (19*-2 = -38).

The resulting number is divided by Moz DA to adjust for the fact that there are other ranking factors.

IE: #1 position Left bias website with Moz DA of 23, #2 position Right bias website with DA of 98.

20*-2 / 23 = -1.74 (Unreasonable)

19*2 / 98 = 0.39 (Somewhat Reasonable)

Higher number = greater impact so a website with a Bias Offset of -1.74 would show extreme favoritism of website, most likely more political reasons. As ranking positions go down, each individual number would also go down since lower rankings have less impact.

In the end, the numbers are added together. An unbiased search ranking would be along the lines of this:

20*1 / 98 = 0.2

19*-2 / 92 = -0.41

18*1 / 93 = 0.19

etc..

Total: -0.3

A website with a very low DA comparatively would throw the results off and would be a signal of unfair rankings.

Comparison

Using the extractor, results from Google and DuckDuckGo were compared for the same search term and analyzed the same, except the Bias Offset of omitted, as it was already very clear that the results were a lot more evenly spread out.

Potential Flaws

1. MediaBiasFactCheck’s method for determining bias has been criticized before. While usually accurate, its method is not very scientific and has room for human error
2. Google also considers other ranking factors besides backlinks (Moz DA’s main calculation factor), mainly actual content.
3. Google’s results are constantly changing, especially for news — maybe just these batches were biased?
4. Moz’s link database isn’t as large as Google’s — their could be more undiscovered links/ranking factors
5. Human error