Type “Russia collusion” into a Google search, and the search engine will try to guess the next word you’ll type. The first of those is “delusion.”

Accept the suggestion, and you’ll find yourself in a conservative rabbit hole. The first result is a New York Post opinion piece with the headline, “Democrats, get set to lose your ‘collusion’ delusions.” The next result, from the conservative site “American Greatness,” reads “Meowing Media Fuel Mass Delusion of Russian Collusion.” Several other results on the first page — including one from a bodybuilding forum — come from other conservative outlets approvingly quoting an apparent catalyst of the phrase’s popularity: Roger Stone, speaking on Fox News.

Google auto-completes search terms based on popularity, so it’s likely that “Russia collusion delusion” is driven by people turning to Google to learn more about Stone’s phrase.

For Francesca Tripodi, a postdoctoral scholar at Data & Society and assistant professor in sociology at James Madison University, the search results are a powerful tell of a phenomenon she set out to document. The “collusion delusion” results are seeking a conservative audience — which is exactly the demographic that would be more likely to search for the phrase in the first place. “No one in the mainstream media has said ‘collusion delusion,'” she said, “but that phrase has been used [by] Tucker Carlson.”

Her study was published in a Data & Society report last week.

Tripodi’s research began with a question that a lot of non-Trump supporters have asked since 2016: How did Trump win? That question led to another: How do conservatives find the news they read? And that question led her to an ethnographic project tracking how a group of white, conservative, evangelical residents of Virginia found, read and researched the news. The people she interviewed were often educated and middle class, and read a variety of news sources.

Despite a popular idea that the conservative media’s core audience passively absorbs whatever they’re told, Tripodi’s research found that plenty will “Google” something they’ve read or seen, often with the intention of fact-checking or challenging their own beliefs. A liberal and conservative might set out to research the same topic, “but based on the keyword, Google is giving you drastically different information,” Tripodi said.

“Googling it” has become the news equivalent of “do your own research.” But neither Google, nor search terms, are purely neutral. “Even in the face of research and due diligence,” Tripodi wrote in her study, “voters can walk away from Google armed with alternative news and alternative facts.”

And, her research indicates, conservative media — along with bad actors and extreme groups looking to amplify their message to the mainstream right — have gotten really, really good at anticipating what their audience will search for and seeding their information to show up, prominently, as a result. The very searches meant to fact-check one’s own beliefs can end up simply enforcing them — particularly for the audience Tripodi studied, who were unified with their distrust of the mainstream media.

“When you search for something on Google, our goal is to provide you with results that are both authoritative and relevant to the query you have typed. This is why when you change your query and use different words, you may get different Search results. However, irrespective of your query, we continue to be committed to providing you the information you need to form your own opinions by surfacing a diversity of sources on our Search results pages,” Google said in a statement.

Tripodi coined the phrase “scriptural inference” to describe the search tactics she observed in her field study, because the method by which the people she observed picked Google search terms was a lot like the research methods taught in Bible study classes. “The conservatives I observed all hold the belief that certain fundamental truths exist, and they critically interrogate media messages in the same way they approach the Bible, focusing on specific passages and comparing what they read, see, and hear to their lived experiences,” Tripodi wrote in the study.

The conservatives Tripodi interviewed seemed unaware that the search terms they chose might not be giving them the results they were expecting. One person told her they believed Google search results were a “consensus” of opinion on a topic; others described it as a neutral platform that provides access to information. Another interviewee, according to Tripodi, “did not consider how her returns were tied to her own search practices or Google’s algorithmic ordering of information. Rather, she used her Google returns to validate her claims, as though Google failing to return an alternative perspective meant that one did not exist.”

Sometimes, these search terms are exploited to lead would-be Googlers toward more-extreme information. A well-known example of this at work comes from Dylann Roof’s confession to murdering nine people in a black Charleston church. Roof claimed that it was a Google search for “black on white crime” — a phrase popular in many far-right and conservative circles, and less so from mainstream sources — that led him to his first result, a white-supremacist site.

The results for his search have since changed, shifting to more-mainstream sources. But other search terms, drawn from phrases associated with fringe ideas that have gained increased exposure on platforms such as Fox News, can also direct users to misinformation or extremist communities.

Here’s an example Tripodi flagged: Say you were checking out Fox News and saw an interesting interview with Candace Owens, who became famous for being praised by Kanye West. The interview focused on the term “red pill,” which has an explicit association with extreme men’s rights online activism and has since spread into the wider far-right Internet. But in the Fox News piece, “taking the red pill” meant “seeing the lies of mainstream media — and learning the truth.”

A conservative searching “red pill” would immediately see a major subreddit of the same name that is a hub of the misogynistic “manosphere.” The second result is Urban Dictionary, where the top definition for “red pill” is positive: “‘Red pill’ has become a popular phrase among cyberculture and signifies a free-thinking attitude, and a waking up from a ‘normal’ life of sloth and ignorance. Redpills prefer the truth, no matter how gritty and painful it may be.”

Although there are also critical results for “red pill” farther down the first page, including those from mainstream sources, these top results would reaffirm the idea suggested in Fox News that the “red pill” is just another term for an intellectual awakening, with no context of its misogynist origins in online culture and current association with online radicalization.

Google doesn’t dispute that different choices in search terms can lead to dramatically different results — it’s part of the core of how Google is designed to work. But both the company and Tripodi emphasized that her paper is just a starting point for examining the role and extent of Google search term choice in directing would-be researchers down rabbit holes of extremist or inaccurate content.

For Tripodi, her research raises questions about the responsibilities of conservative media outlets, who “are understanding how their audiences are thinking and searching.” These platforms have become amplifiers for ideas that once existed on the fringe but have crept into relevance in the Trump era. And when outlets such as Fox News run positive segments on terms like “red pill” without context, “you’re inviting your audience to Google these phrases,” she said. “I don’t think they’re being careful enough with this blurriness.”

More reading: