Donald Trump took to Twitter this week to express his frustration with Google News.

He accused the search engine of being “RIGGED” so that people searching for news about the US President would be presented with mainly negative stories from the “Fake News Media”.

So is Google News really “suppressing voices of Conservatives and hiding information and news that is good”?

The claim

Mr Trump tweeted twice on the topic:

Google search results for “Trump News” shows only the viewing/reporting of Fake News Media. In other words, they have it RIGGED, for me & others, so that almost all stories & news is BAD. Fake CNN is prominent. Republican/Conservative & Fair Media is shut out. Illegal? 96% of….

….results on “Trump News” are from National Left-Wing Media, very dangerous. Google & others are suppressing voices of Conservatives and hiding information and news that is good. They are controlling what we can & cannot see. This is a very serious situation-will be addressed!”

Where’s it coming from?

The statistically-minded amongst you might notice the striking figure Mr Trump quotes there: that “96% of results on ‘Trump News’ are from National Left-Wing Media”.

It appears to come from an article on PJ Media, which describes itself as “a leading news site covering culture, politics, faith, homeland security, and more”. The piece, by PJ Media supervising editor Paula Bolyard, suggests “Google is manipulating its algorithm to prioritize news outlets in their coverage of President Trump”.

Bolyard says she searched for “Trump” on Google’s News tab, and “analyzed the results using Sharyl Attkisson’s media bias chart”. (By the way, that chart puts PJ Media as a right-leaning outlet).

By Bolyard’s account, the results showed “blatant prioritization of left-leaning and anti-Trump media outlets”.

Based on her own experiment, “CNN [which Bolyard counts as “left-leaning”] has a disproportionate number of articles returned when searching for ‘Trump’ — nearly 29 percent of the total. In fact, left-leaning sites comprised 96 percent of the total results.”

Her article includes the following chart, which looks at the first 100 Google News hits for the term “Trump”. Sites that Bolyard considers “left-leaning” are shown in blue, while “right-leaning” sites are in red.

According to Bolyard: “other left-leaning sites that appeared on the first page were CBS, The Atlantic, CNBC, The New Yorker, Politico, Reuters, and USA Today (the last two outlets on this list could arguably be considered more centrist than the others)”.

She says “not a single right-leaning site appeared on the first page of search results”.

Bolyard notes that “[her employer] PJ Media did not appear in the first 100 results, nor did National Review, The Weekly Standard, Breitbart, The Blaze, The Daily Wire, Hot Air, Townhall, Red State, or any other conservative-leaning sites except [Fox News and the Wall Street Journal].”

How reliable are the results?

Bolyard “performed the search a multiple times [sic] using different computers (registered to different users) and Google returned similar results”.

She accepts that her test is “not scientific”, although she concludes that “the results suggest a pattern of bias against right-leaning content.”

But we’re not sure that her results are as illuminating as she suggests.

Bolyard doesn’t interrogate the “media bias chart” that she uses to gauge the political persuasions of the news outlets presented by Google. The chart’s own author, Sharyl Attkisson, describes it as “subjective”.

Indeed, Bolyard’s article leans heavily on a study from media analytics company Can I Rank, which seems to disagree with Sharyl Attkisson’s media bias chart.

For example, Sharyl Attkisson lists CNN, the Washington Post, Reuters and Politico as “left-leaning”. But the Can I Rank study puts those same outlets in the category “Bipartisan (conservative, moderate, and liberal)” — rather than in its “liberal and moderate” or “liberal only” categories.

It’s worth saying that we can’t vouch for the validity of the Can I Rank categories.

We were surprised, for example, to find the website Zero Hedge listed as “bipartisan”. A former Zero Hedge writer, Colin Lokey, told Bloomberg that the site’s political stance was: “Russia=good. Obama=idiot. Bashar al-Assad=benevolent leader. John Kerry= dunce. Vladimir Putin=greatest leader in the history of statecraft.”

But whether we think the Can I Rank categories are reliable or not, it’s still strange that Bolyard quotes their study at length to support her overall point, while not adopting their methodology herself.

Bolyard’s article describes how Can I Rank “studied over 1,200 URLs that ranked highly in Google search results for politically-charged keywords like ‘gun control,’ ‘abortion,’ ‘TPP,’ and ‘Black Lives Matter’ and then assessed whether there was a political slant to the articles.” She notes that they “sampled 2,000 results”.

Remember, Bolyard herself acknowledges that her own experiment was “not scientific”. So why did she use her own results in the headline of her piece, rather than those of a much larger study?

One answer is that it would have weakened her case.

If Bolyard had used the Can I Rank categories instead of Sharyl Attkisson’s, her experiment would have have found that 47 per cent of Google News results for “Trump” are “left-leaning” — not 96 per cent as Bolyard and the President claim.

Can I Rank’s own results support Bolyard’s general point, but are less eye-catching than Bolyard’s. Can I Rank “found that searchers are 39 percent more likely to be presented with left-leaning articles”.

Bolyard neglects to mention the overall conclusion of Can I Rank’s study: “No discernible pattern emerged when looking at many of these factors. The evidence, we conclude, is not strong enough to prove an inherent bias within Google’s algorithm.”

Are the results that surprising?

We’re not privy to the precise workings of the Google algorithm, but we do know that sites that are used more will rank higher in the search engine.

Bolyard points out that CNN appears to dominate Google News search results according to her exercise. But is that so surprising when we consider that CNN has the largest digital presence of any US news outlet, with 115 million “unique visitors” to its online platforms in July?

Also high up the list are the New York Times (79 million), the Washington Post (73 million), CBS News (65 million) and the BBC (60 million).

The notable outlier is Fox News, which attracts 90 million users a month, but which according to Bolyard’s experiment gets only two articles in the top 100 Google News items when she searched for the term “Trump”. But again, this is not based on a scientifically-rigorous study.

You might say that the reason that CNN, the New York Times and others dominate the online media market is precisely because they appear high in Google’s search algorithm. But we’re in a chicken-and-egg situation: are these sites top of the list because Google makes it so, or does Google’s algorithm prioritise these sites because they are popular?

We can’t say for certain without a look at the much-guarded secrets of the algorithm.

What does Google say?

Responding to Mr Trump’s tweets, Google said: “When users type queries into the Google Search bar, our goal is to make sure they receive the most relevant answers in a matter of seconds.”

The spokesperson said: “Search is not used to set a political agenda and we don’t bias our results toward any political ideology.” They insisted that “we never rank search results to manipulate political sentiment”.

FactCheck verdict

Donald Trump claimed on Twitter that “96% of results on ‘Trump News’ are from National Left-Wing Media”.

This figure comes from an article by PJ Media that contains a number of holes:

  • The author conducted a small-scale experiment to see what comes up on Google News when you search the word “Trump”. She concedes that her experiment was “not scientific”.
  • The author categorises CNN, the Washington Post and Politico as “left-leaning”. It’s not for us to decide whether that is correct or not. But we note that the author quotes a much larger study by a company called Can I Rank, which categorises those same sites as “bi-partisan”. If the PJ Media experiment had used the same categories as the Can I Rank study, it would have concluded that 47 per cent of Google News results for the term “Trump” are “left-leaning”, rather than 96 per cent.
  • The PJ Media author cites the Can I Rank study — which “found that searchers are 39 percent more likely to be presented with left-leaning articles.” The author does not mention that the study also states: “the evidence, we conclude, is not strong enough to prove an inherent bias within Google’s algorithm.”

Overall, it’s not that surprising that websites like CNN, the Washington Post, NBC and the New York Times are high up the Google rankings: they are some of the most-read media outlets in the US.

Responding to Mr Trump’s tweets, a Google spokesperson said “we never rank search results to manipulate political sentiment”.