One question that we’ve been hearing over and over again since the 2016 election is:
Is Google biased?
There are no shortages of opinions.
Sundar Pichai went before Congress in 2018 and swore under oath, “I’m confident we don’t approach our work with any political bias.”
He also sent an internal memo to staff warning them against letting their personal politics affect their work.
Elon Musk, on the other hand, posted to X, “Google is controlled by far left activists.”
A conservative organization, the Media Research Center, routinely posts articles that show supposed “proof” of Google’s political bias, while left-leaning Vox posted an article mocking conservatives for not understanding how SEO works.
If you’re like me, you’re just reading all the back-and-forth and getting tired of it.
Too many opinions on both sides are based on confirmation bias, sensationalism, or a fundamental misunderstanding of how SEO really works.
And so I thought I’d jump into this hornet’s nest.
Like everyone else, I have my own biases but I’m going to do my best to keep them at bay.
Instead, let’s use SEO tools and techniques to see if we can come to a definitive answer.
‘Google bias’ in the 2024 election?
Throughout the 2024 election, there were many stories about Google’s supposed “bias.”
Let’s take a look at some of the more prevalent ones.
In June, the Media Research Center accused Google of “blacklisting” President Trump’s official campaign website because it wouldn’t rank for [donald trump presidential race 2024] and [republican party presidential campaign websites].
The problem with this is that even a junior SEO could have seen that Donald Trump’s website was pretty horrifically optimized.
Their home page title tag read Home | Donald J. Trump, and most of their substantive content was hidden in a PDF.
In July, many people including Donald Trump, Jr. accused Google of “election interference” because Google autocomplete would not suggest President Trump’s name when someone typed in “assassination attempt on…”
Google’s official explanation was that they have “protections in place against autocomplete predictions associated with political violence.”
To be honest, I didn’t buy that (I could see autocomplete for other contemporary figures), but I just chalked it up to Google autocomplete was embarrassingly slow to update.
The third incident to make waves was on Election Day, when searches for [how to vote harris] spawned a box that told people where to go for their nearest polling place, while [how to vote trump] did not.
Google PR explained that this was because “Harris” is also the name of a county in the U.S., while “Trump” is not.
Again, a perfectly plausible explanation.
Thousands of conservative accounts jumped on these incidents as definitive proof of Google interfering in the election.
The mistake they made was assuming that Google is infallible.
In reality, anyone with a passing understanding of Hanlon’s Razor – which suggests we should not attribute to malice what can be explained by incompetence – would see that it applied in all three cases.
Accusations of Google bias
The problem with focusing on noise like this is that it detracts from the real question.
Do Google search results have bias, and is that bias enough to unduly influence people?
Through the years, a number of whistleblowers and researchers came forward with supposed proof of Google bias. Some highlights:
In November 2016, following the presidential election, an anonymous source within Google sent a leaked video to the conservative outlet Breitbart showing Google executives’ and employees’ negative reaction to the election results.
In subsequent years a number of whistleblowers from within Google came out to provide their reports of bias that they perceived within Google.
In July 2019, senior engineer Greg Coppola came forward to publicly disagree with his CEO’s claim that searches were unbiased.
In August 2019, senior software engineer Zach Vorhies released hundreds of pages of internal Google documents that allegedly showed evidence of manipulation, from intervening in the algorithm to maintaining blocklists.
One of the more interesting presentations was about “algorithmic unfairness,” which discussed the need for search results to reflect a desired state, even if it didn’t reflect current realities.
While not a Google employee, Robert Epstein was a research scientist who went on a number of conservative outlets with research purporting to show Google manipulating public opinion.
The problem with all of this? Because this evidence was mainly hearsay, opinions were split like a Rorschach test.
Conservatives, including members of Congress, pointed to them as definitive proof of Google’s bias, while many in the mainstream media dismissed them as conspiracy theories.
The facts
Let’s take a step back and look at objective facts:
From 1998 to 2018, Google was powered by their original algorithm based mainly on PageRank. That worked really well in the beginning but as more people understood Google’s algorithm, poor quality sites began to rank. Despite their efforts with Panda and Penguin, it became clear that too many legitimately dangerous sites were making their way into Google’s results.
The Aug. 1 2018 broad core update (a.k.a., the Medic Update) was Google’s first big attempt to go beyond reactively fighting content and link manipulation and proactively combat this, starting with financial and medical topics (YMYL).
Most of us in the SEO space had our suspicions that Google was putting its finger on the scales for other types of searches. In May 2024, leaked documents from Google confirmed that Google’s organic algorithm indeed treated COVID and election-related searches differently than others through two factors called IsCovidAuthority and IsElectionAuthority, respectively.
None of this is a smoking gun either.
Those who attack Google say this circumstantial evidence is enough to prove Google’s bias.
Defenders of Google will say that all of these steps were necessary to fight the real problem of bona fide misinformation and scams.
The data
So, is Google biased?
Instead of giving you my opinion, I’m going to show you how you can use SEO tools and techniques to figure it out for yourself.
The two tools I use most often for my SEO work are Semrush and Ahrefs. Both of them have a useful feature: the ability to go back in history and see historical SERPs.
For example, these are the top 10 organic results for searches on “donald trump” that Semrush reports from October 2024, one month before Election Day.
And here’s what Ahrefs reports for October 15, 2024.
Both are similar.
The slight variations are due to variations in the way that Semrush and Ahrefs obtain their Google results.
We’re still in Rorschach test territory.
Those who accuse Google of bias will look at the results and cry foul because CNN, AP, Wikipedia, and The Guardian – all known for being left-leaning – are showing up.
Those defending Google will point to Donald Trump’s website and his multiple social media accounts showing up as proof that Google is unbiased.
What if we could take a look at every question that people asked about Donald Trump and Kamala Harris during the election, take the top 10 results for each, and run an analysis of which media outlets are cited most often?
We can. Here’s how.
For this one I’m going to use Ahrefs (which allows me to output 1,000 queries and their top 10 positions and to filter based on date).
I searched for “Questions” that people ask about “donald trump.” I filtered on searches that were seen before Election Day 2024.
Next, I exported the top 1,000 questions with the top 10 positions for each.
I uploaded the CSV file to ChatGPT and asked it to go through the list and tally up how often each news outlet or website appeared.
I repeated the process for questions containing “kamala harris” and tallied everything up. At this point I had a list of all sites that ranked in the top 10 for the top 1,000 questions about Trump and Harris.
Next, I cross referenced this list with media bias charts from AllSides.
My goal here was to see whether Google organic search favored sites of a particular political persuasion. Here’s what the results were.
To double-check, I cross-referenced the list against Ad Fontes Media’s Static Media Bias Chart.
I took every site they listed in this chart that was rated 24.0 and above in News Value and Reliability.
Neither AllSides nor Ad Fontes are perfect.
For example, the left will likely disagree with AllSides’s characterization of AP as “left,” while the right will likely disagree with Ad Fontes’s characterization of RealClearPolitics as “strong right.”
But on the whole, these are the best out there (at least as far as Google is concerned).
Get the newsletter search marketers rely on.
Is this proof of Google’s bias?
I’m going to get half of America mad at me now.
Yes, there is pretty clear evidence that Google’s organic results demonstrate bias when it comes to political searches.
But don’t take my word for it. Repeat the process above for any political phrase you can think of.
Note that Google is not “censoring” conservative and right-leaning outlets – you can still find them in search results if you search on their brand name.
But do any kind of non-branded search and you’ll be hard-pressed to find them ranking.
It wasn’t always this way.
Note in this screenshot how in August 2016 it was fairly common to see centrist outlets like RealClearPolitics and right-leaning outlets like the Washington Times alongside left-leaning outlets like CNN and The Atlantic for searches for “donald trump.”
You can see from this chart of SEO traffic and keywords what happened to RealClearPolitics.
Somewhere around April 2020, their SEO traffic and keywords fell off a cliff.
Semrush chart of SEO performance for RealClearPolitics.com
Today, 92% of their Google traffic comes from branded searches.