Search engines that incorporate AI from Google, Microsoft, and Perplexity have been revealing deeply problematic and extensively discredited studies that advocate for race science and suggest that white individuals possess genetic superiority over individuals of other races.
Patrik Hermansson, a researcher affiliated with the UK-based anti-racism organization Hope Not Hate, was engaged in a lengthy investigation pertaining to the resurgence of the race science movement when he sought more information about a widely refuted dataset that purports to demonstrate the superiority of the white race through IQ scores.
His inquiry centered on the Human Diversity Foundation, a race science organization financed by Andrew Conru, a U.S. tech billionaire and founder of Adult Friend Finder. Established in 2022, this group succeeded the Pioneer Fund, which was set up in 1937 by U.S. Nazi sympathizers with the intent of promoting concepts of “race betterment” and “race realism.”
To further his research, Hermansson accessed Google and started searching for the IQ scores of various countries. However, when he entered “Pakistan IQ,” instead of the usual list of links, he encountered Google’s AI-based Overviews tool, which surprisingly was activated by default. The tool provided him with a definitive score of 80.
When he searched for “Sierra Leone IQ,” Google’s AI tool provided a very precise figure: 45.07. A similar specificity was found for “Kenya IQ,” which returned the result of 75.2.
Hermansson instantly recognized the figures being presented to him. They were sourced directly from the very study he sought to challenge, authored by a prominent figure in the movement he aimed to criticize.
The data Google retrieved originated from a study published by Richard Lynn, a professor at the University of Ulster who passed away in 2023 and had held the presidency of the Pioneer Fund for twenty years.
“His impact was tremendous. He was the star and the guiding light of that movement until his passing. Nearly until the end of his life, he remained a key leader of it,” Hermansson states.
A recent investigation by WIRED validated Hermanssons’s findings, revealing that other AI-driven search engines, such as Microsoft’s Copilot and Perplexity, also reference Lynn’s research when discussing IQ scores across different nations. Lynn’s controversial studies have historically been utilized by far-right extremists, white supremacists, and advocates of eugenics to argue that the white race holds genetic and intellectual superiority over nonwhite races. Experts are increasingly concerned that the endorsement of this research by AI could foster further radicalization.
“The uncritical acceptance of these ‘statistics’ poses significant issues,” states Rebecca Sear, director of the Center for Culture and Evolution at Brunel University London, in her comments to WIRED. “The dissemination of such data not only propagates misinformation but also advances the political agenda of scientific racism—the distortion of science to assert that racial hierarchies and disparities are natural and unavoidable.”
To support her assertion, Sear highlighted that Lynn’s findings were cited by the white supremacist responsible for the mass shooting in Buffalo, New York, in 2022.
Earlier this year, Google introduced its AI Overviews as part of an initiative to redesign its influential search tool amid the evolving landscape shaped by artificial intelligence. For select search queries, this tool—currently accessible in only a limited number of countries—provides AI-generated summaries of its findings. It aggregates information from the internet, allowing users to receive answers without the need to follow links.
The AI Overview feature may not always make it clear where the information originates, but after receiving feedback regarding the absence of article references, Google has now included the title of a relevant link adjacent to the AI summary. Since its launch in May, AI Overviews have encountered a variety of challenges, prompting Google to acknowledge it had stumbled during the much-anticipated rollout. The AI Overviews feature is enabled by default for search results, and removing it requires the installation of third-party extensions. (“I haven’t enabled it, but it was enabled,” said Hermansson, the researcher, to WIRED. “I don’t know how that happened.”)
In terms of the IQ results, Google tapped into a range of resources, including posts on X, Facebook, and several lesser-known listicle sites like World Population Review. In most instances, clicking on these sources leads back to Lynn’s well-known dataset. (In certain cases, while the specific numbers published by Lynn are mentioned, the sites do not credit Lynn as the original source.)
When asking Google’s Gemini AI chatbot directly with the same queries, it delivered a far more nuanced reply. “It’s important to approach discussions about national IQ scores with caution,” stated the text generated in response to the inquiry “Pakistan IQ.” The statement elaborated: “IQ tests are designed primarily for Western cultures and can be biased against individuals from different backgrounds.”
Google informed WIRED that its systems did not function as expected in this instance and that it is exploring avenues for enhancement.
“We have measures and guidelines in place to safeguard against low-quality responses, and when we identify Overviews that don’t comply with our standards, we take prompt action,” stated Ned Adriance, a spokesperson for Google, in an interview with WIRED. “These Overviews were against our policies and have been taken down. Our objective is for AI Overviews to offer links to quality content so that users can click through to learn more, but sometimes there may not be ample high-quality content available for certain queries.”
While tests conducted by WIRED indicate that AI Overviews have been disabled for queries related to national IQs, the results still serve to propagate the erroneous data from Lynn’s work within what is known as a “featured snippet,” which showcases some text from a website prior to the link.
Google has not provided a response regarding this update.
However, it isn’t only Google that is endorsing these harmful theories. When WIRED submitted the same query to various AI-driven online search platforms, similar results were observed.
Perplexity, an AI-focused search platform known for sometimes fabricating information, addressed a question regarding the IQ in Pakistan, noting that “the average IQ in Pakistan has been reported to vary significantly depending on the source.”
This response included a variety of references, among which was a Reddit discussion that drew from Lynn’s research as well as the World Population Review site previously cited by Google’s AI Overview. In response to an inquiry about Sierra Leone’s average IQ, Perplexity directly referenced Lynn’s data, stating: “Sierra Leone’s average IQ is reported to be 45.07, ranking it among the lowest globally.”
Perplexity did not provide a comment when requested.
Microsoft’s Copilot chatbot, which functions within the Bing search engine, presented confidently phrased information, asserting that “The average IQ in Pakistan is reported to be around 80,” with a citation from a site named IQ International that did not specify its sources. When queried about “Sierra Leone IQ,” Copilot indicated a figure of 91, linking to a site known as Brainstats.com, which mentions Lynn’s studies. Copilot additionally referred to work from Brainstats.com when asked about IQ in Kenya.
“Copilot delivers responses by synthesizing data from a variety of online sources into one cohesive answer,” Caitlin Roulston, a representative from Microsoft, explains to WIRED. “Copilot also provides linked citations, allowing users to delve deeper into their inquiries much like traditional search methods.”
Google mentioned that one challenge it encounters when producing AI Overviews is the lack of high-quality information available for certain meticulous queries—and it’s clear that Lynn’s work falls short in quality.
“The scientific foundation of Lynn’s ‘national IQs’ database is so flawed that it raises doubts about the database’s legitimacy,” Sear commented. “Lynn has yet to clarify his approach to selecting samples for inclusion; many countries have their IQs estimated based on exceptionally small and unrepresentative sample sizes.”
Sear highlights that Lynn’s estimation of Angola’s IQ relies on data from only 19 individuals, while Eritrea’s IQ is based on samples taken from children in orphanages.
“The challenge with this data is that the source Lynn utilized to create this dataset is fundamentally flawed, and this issue manifests in several ways,” Rutherford commented, highlighting that the Somali figure in Lynn’s dataset is derived from a single sample of refugees aged 8 to 18 who were tested in a Kenyan refugee camp. He further stated that the Botswana score relies on just one sample of 104 Tswana-speaking high school students aged 7 to 20 who were assessed in English.
Critics of utilizing national IQ tests to suggest concepts of racial superiority argue not only that the quality of the samples gathered is inadequate, but also that the tests themselves are often created for Western audiences, rendering them biased even prior to administration.
“There is evidence that Lynn has systematically skewed the database by favoring samples with lower IQs while neglecting those with higher IQs, particularly for African countries,” Sears remarked, a viewpoint supported by a preprint study from 2020.
Lynn released multiple iterations of his national IQ dataset throughout the decades, with the latest version, titled “The Intelligence of Nations,” being published in 2019. Over the years, Lynn’s erroneous work has been exploited by far-right and racist factions as justification for claims of white superiority. The data has also been transformed into a color-coded world map, indicating sub-Saharan African nations marked in red for allegedly low IQs, in contrast to the Western countries, which are depicted in blue.
“This is a data visualization that you see all over X, formerly known as Twitter, all over social media—and if you spend a lot of time in racist hangouts on the web, you just see this as an argument by racists who say, ‘Look at the data. Look at the map,’” says Rutherford.
However, Rutherford argues that the responsibility does not solely rest on the AI systems, but also on a scientific community that has been uncritically referencing Lynn’s work for many years.
“It’s really not surprising [that AI systems are referencing it] because Lynn’s research on IQ has been accepted without much scrutiny across a large segment of academia, and if you examine the number of times his national IQ databases have been cited in scholarly works, it’s in the hundreds,” Rutherford stated. “Thus, the issue isn’t with AI. The issue lies within academia.”
