The Role of Search Engines in Contesting or Reinforcing Election Fraud Narratives

Bret Schafer

“Look up SMARTMATIC and tweet me what you think?” tweeted Rudy Guiliani, then-US President Donald Trump’s personal lawyer, in the weeks after the 2020 US presidential election. It was part of a broader campaign to delegitimize the election by casting doubt on the companies that manufacture voting machines. It also borrowed a tactic—a call to “do your own research”—that has become so ubiquitous in the chat rooms, online forums, podcasts, and traveling roadshows dedicated to proselytizing mistrust, not just of election results but of academic and scientific consensus more broadly, that it has earned its own acronym: “DYOR”.

Much of the research on the spread and adoption of false and misleading narratives about election administration and security has focused on the role of social media platforms. Though social media platforms are also increasingly being used—particularly by younger demographics—to conduct user-generated searches, those who seek out more information on a topic are still more likely to “do their own research” on search engines. In the days after Guiliani’s tweet, for example, searches on Google for Smartmatic surged to their highest level in the past five years. The results generated by search engines can therefore serve as correctives to false information encountered elsewhere—or they can act, as certain studies have shown, as vicious feedback loops that reinforce and perpetuate those falsehoods.

To better understand the role of search services in the spread of both accurate and misleading election information, ASD at GMF has been tracking daily search results since last December that appear on the first four pages of Google Search, Google News, Microsoft Bing, and Bing News on roughly 30 search terms related to the administration and security of US elections. These terms range from very generic queries—for example, “Is mail-in voting safe?” and “same-day voter registration”—to more loaded queries—for example, “ballot drop box tampering” and “same-day voter registration fraud”. We geolocated searches to the states hosting the first three primaries or caucuses (Iowa, New Hampshire, and South Carolina) and the states that, in our estimation, currently serve as exemplars in the public imagination of conservative and progressive governance (Florida and California). Throughout the primaries and general election, we will continue to monitor a static list of search terms, while also incorporating additional terms connected to new or emerging narratives that could potentially become data voids that are more easily exploited by bad actors. We will communicate our findings in a series of blogposts, short reports, and, perhaps most importantly, via regular engagements with state and local election officials.

This project has several objectives, including:

  • gaining a better understanding of the sources and narratives generated by major search engines related to queries that suggest an interest in or concern about the security and reliability of US elections
  • testing whether differences in the framing and construction of similar search queries (for example, the use of language that indicates a more conservative or liberal point of view) leads users to different information environments
  • monitoring results for the existence of “pink slime” news sites or sites that misrepresent their purpose or ownership
  • detecting foreign information manipulation or interference (though we rate this risk to be low)
  • providing stakeholders, particularly those charged with counter-messaging, with a better sense of the enduring or emerging narratives that are appearing on search platforms
     

This work is a continuation of a limited number of previous studies that have investigated the role of search engines in promoting or debunking election-related rumors or conspiracy theories. It also builds upon six years of research conducted by ASD looking at the ways in which search engines can be exploited by state-backed actors seeking to advance preferred narratives around a range of topics—from the People’s Republic of China’s (PRC) efforts to promote conspiracy theories about the origins of the coronavirus to Russia’s attempts to discredit investigations into the poisoning of Sergei and Yulia Skripal, the former Russian intelligence agent and his daughter who were poisoned in Salisbury, United Kingdom, in 2017.

Unlike our past research, however, the focus here is not—at least primarily—on identifying domain-level risks. We do not seek to evaluate the relative merits of individual websites that appear in search results, except for sites that misrepresent their purpose, ownership, or are otherwise engaged in behavior that violates search engines’ terms of service (for example, link schemes, sneaky redirects, or automatically generated content). This research is also not meant to evaluate the veracity of claims that appear in search results—unless those claims could conceivably pose a risk to the safety of election workers or mislead voters about the time and place of elections, among other potentially illegal conduct.

We recognize that competing viewpoints about election security and integrity are a healthy part of democratic debates. It is therefore not our intention—nor do we think it is defensible—to rate search results as problematic merely because they contain links to websites that promote skepticism about the security and integrity of elections, even if we find those narratives to have merit. Instead, this project is meant to increase our understanding of—and our ability to respond to—the narratives being shown to voters who may have good faith questions about the mechanics of elections in the United States.

Preliminary Findings

After nearly two months of data collection, we have now collected almost 1.5 million search results across the four studied search products. In the coming months, this data will be more rigorously coded and examined. A few, if initial, high-level observations are, however, possible:

  • State-backed information sources—particularly from countries that US intelligence agencies consider to be most active in election interference activities—are almost entirely absent from search results. In total, there were fewer than 4000 results, far less than 0.5% of all results, from outlets that the state media monitor considers to be controlled by foreign governments (as opposed to independent publicly funded media, like the BBC), with more than 90% of those observations coming from the Qatari-backed Al Jazeera English. As of February 16, 2024, there were only 66 observations from PRC state media outlets and only five from Russian state media outlets, all coming from TASS and unrelated to elections in the United States. While this suggests that the threat of foreign manipulation of election integrity debates on search services is relatively low, we caution that these findings only include overt, state media outlets. They also do not include the potential reposting of foreign state media articles on unaffiliated websites, something we have repeatedly observed in our previous research.
  • Of the top 100 domains observed in our results, nearly all came from well-known media outlets, US government websites, or established think tanks. This, of course, does not mean that the content posted by those sites is inherently more credible, but it does suggest that it may be harder for less established, and potentially more problematic, information sources to break through in search environments, at least on queries where there is ample material for search engines to pull from.