A new study highlights how the deluge of online images may be subtly shaping our perceptions, revealing that search engine imagery is reinforcing gender stereotypes in a way that text-based searches do not. As more than 6 hours a day is spent online on average, visual input from social media feeds, websites, and digital ads contributes to a cycle that may be increasing implicit biases in society.

The recent research analyzed image search results for various occupations on platforms like Google, Wikipedia, and IMDb. It found that images were overwhelmingly gendered, particularly in fields historically viewed as male- or female-dominated. For example, searches for “heart surgeon,” “investment banker,” or “developer” predominantly returned images of men, while terms like “housekeeper” and “nurse practitioner” were associated almost exclusively with women.

The study went beyond just measuring bias in search results. Researchers conducted an experiment where 423 U.S. participants used Google to search for occupations, with some participants receiving visual representations through Google Images while others used text-based Google News searches. Those exposed to image results displayed a marked increase in implicit gender biases, as measured by a standard association test, even days after the experiment. The findings highlight the impact of image-heavy platforms like Instagram and TikTok in normalizing biased visuals, raising concerns that the sheer volume of gender-stereotyped images might be entrenching outdated perceptions.

Vicious Cycle of AI and Bias

The problem extends to AI models, which are trained on vast repositories of online content, including stereotyped images. When users ask AI platforms like ChatGPT to visualize various professions, they often receive outputs that reflect existing biases. A request for images of “doctor” or “scientist,” for example, may yield predominantly white male figures, reinforcing societal stereotypes. Similarly, descriptors such as “successful” or “smart” also skew heavily towards images of white men, reflecting the biases embedded in the data used to train these systems.

The study’s authors warn that this cycle could worsen as AI tools continue to rely on biased online content. “The rise of images in popular internet culture may come at a critical social cost,” they write, noting that these biases not only influence AI outputs but also shape user perceptions. The more biased imagery we encounter, the more normalized these stereotypes become, perpetuating a feedback loop of implicit bias.

Seeking Solutions and Reclaiming Visual Space

Amid growing concerns, experts point to several solutions to mitigate the spread of biased visual content. Technology companies bear much responsibility, though attempts to address the issue have sometimes led to overcorrections. Google’s AI tool Gemini, for instance, has been criticized for inserting diversity where it historically wouldn’t exist, sometimes producing historically inaccurate imagery. Yet, even with the best intentions, fixing ingrained biases remains a challenge for tech firms.

One approach recommended for individuals is to curate their social media feeds to follow diverse creators and photographers from around the world. Another option is the “digital detox,” as outlined in art entrepreneur Marine Tanguy’s The Visual Detox: How to Consume Media Without Letting It Consume You, which suggests limiting screen time and reclaiming time away from devices. Tanguy advocates setting daily app timers, deleting unused apps, and spending time outdoors to reduce reliance on screens.

Perhaps most importantly, experts stress the value of self-awareness in understanding how digital imagery influences our beliefs and perceptions. Unlike previous generations, modern users encounter a constant stream of images that subtly shape their worldviews. For much of human history, art and visual media were limited, yet today’s image-saturated environment is altering how we see others and ourselves, often without conscious realization.

As visual culture continues to expand online, recognizing these subtle influences may be crucial in building a more balanced and less biased digital world.

Recent reports reveal Google’s influence on public opinion through its search algorithms, which may reinforce users’ existing beliefs and contribute to societal divides. A comparison of Google searches related to Vice President Kamala Harris illustrates this phenomenon. When users searched for “Is Kamala Harris a good Democratic candidate,” they were met with articles highlighting positive aspects, including a Pew Research Center poll that found Harris “energizes Democrats” and an Associated Press article noting widespread support among Democrats for Harris as a potential president. However, when searching for “Is Kamala Harris a bad Democratic candidate,” users encountered more critical results, including a top article from Reason Magazine and several opinion pieces with negative views on her candidacy.

This divergence in search outcomes underscores a wider issue: Google’s algorithm often reflects user inclinations back to them, creating a “feedback loop” that reinforces their original queries. According to Varol Kayhan, assistant professor of information systems at the University of South Florida, search engines like Google heavily shape the information people see and the beliefs they form. “We’re at the mercy of Google when it comes to what information we’re able to find,” he said.

Sarah Presch, digital marketing director at Dragon Metrics, an SEO-focused platform, noticed the impact of Google’s approach in searches beyond politics, such as healthcare. Presch found that when searching “link between coffee and hypertension,” Google’s Featured Snippet quoted from the Mayo Clinic, suggesting caffeine could cause a temporary spike in blood pressure. However, searching “no link between coffee and hypertension” produced a snippet from the same article, stating that caffeine does not have long-term blood pressure effects. Presch notes this as evidence that Google surfaces content tailored to confirm user queries, regardless of the nuances in source material.

Google’s spokesperson defended the search giant’s practices, stating that the goal is to present high-quality, relevant information and allow users access to a range of perspectives. Google also pointed to research suggesting that user choices drive exposure to partisan information rather than algorithmic design alone.

Despite these assertions, some experts argue that Google’s algorithms play a central role in perpetuating echo chambers, particularly in politically charged queries. Silvia Knobloch-Westerwick, professor at Technische Universität Berlin, notes that even if users have some control over the information they engage with, the algorithms determine the options that appear before them.

Another issue lies in Google’s approach to search queries themselves. Mark Williams-Cook, founder of the SEO tool AlsoAsked, explained that Google’s algorithms prioritize user reactions to content rather than deep document analysis. A 2016 internal Google presentation stated that “we hardly look at documents. We look at people.” Williams-Cook argues that this reliance on user engagement to rank content creates a “feedback loop” that feeds users content matching their interests—potentially at the cost of objectivity.

Although Google says its algorithms have evolved significantly since 2016, Williams-Cook believes the underlying model persists. He likens Google’s system to “letting a kid pick out their diet based on what they like,” adding, “they’ll just end up with junk food.” As search engines continue to influence how people find and consume information, experts warn that algorithm-driven bias could amplify confirmation bias and limit exposure to diverse viewpoints.