Racial bias in video games shows up in more than one place: in which characters appear and how they are portrayed, in how players treat one another online, and in the policies and design choices that shape those experiences. The evidence on each is different. A historical study of game characters cannot tell us how often players face harassment today, and a small study of selected online sessions cannot establish a rate for all games.
That distinction matters when deciding what to change. Better representation, identity-aware rules and moderation, reliable reporting and support, transparency, and prosocial design address related problems—but none is a proven universal fix.
How does racial bias appear in game characters?
Bias in representation can involve who is present, who is missing, and whether characters are given varied, meaningful roles rather than reduced to stereotypes. One frequently cited study offers a historical baseline, not a picture of today’s releases.
What the historical content study found
In “The Virtual Census: Representations of Gender, Race and Age in Video Games,” published in 2009, Williams, Martins, Consalvo, Ivory, and colleagues analyzed a sales-weighted sample of 150 games across nine platforms. Comparing game characters with the U.S. population, they found white characters overrepresented and Hispanic and Native American characters underrepresented, among other findings. The sample was drawn from one year and was not an audit of every game; its results should be read as evidence about representation at that time, not as a current census.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →#1 Best Overall
What players say about diverse characters
A 2025 report from the Geena Davis Institute with Nielsen asked U.S. and U.K. respondents about character diversity, using Nielsen’s Video Game Tracker in June 2024. Sixty-four percent agreed they were more likely to play a game with characters more diverse than those they usually see, and 62% agreed they were more likely to play a game if its playable characters were diverse. These are reported attitudes, not observed purchases or proof that diversity by itself causes a game to succeed.
What does racial bias look like in online play?
In online spaces, bias can take the form of slurs, identity-based insults, threats, or exclusion. Whether a particular incident counts as racial harassment depends on what was said or done and how a study defines and records it; ordinary rudeness should not automatically be labeled racial abuse.
An Anti-Defamation League (ADL) study published in 2025 recorded harassment during controlled play sessions in Valorant, Counter-Strike 2, Overwatch 2, and Fortnite. Fifteen student researchers, recent graduates, and young adults played using usernames expressing Jewish, Muslim, Chinese, Israeli, or Mexican identities. From late 2023 through summer 2024, they logged almost 250 cumulative gameplay hours. Nearly half of the experimental sessions included some form of harassment, and one-third included identity-based harassment. Those percentages describe the study’s sessions—not the share of all players or all online play that involves harassment.
Rank #2
The ADL reported some form of harassment in about two-thirds of its tested Valorant and Counter-Strike 2 sessions. Overwatch 2 and Fortnite had lower identity-based harassment rates in the tested sessions. The participant group was small and non-random, and the results do not establish a definitive ranking of whole game communities or show that any of the four games is broadly safe.
The ADL’s 2022 report also describes how harassment research can ask about discrimination by a stranger based on ethnicity, targeting tied to an avatar’s appearance or username, and exclusion based on identity. These categories help distinguish identity-related abuse from other unpleasant interactions, but they do not make every rude encounter evidence of racial bias.
How should these findings be compared?
Representation studies, player surveys, live-play observations, and educational-game evaluations measure different things. Keeping their units and populations separate prevents a statistic about one kind of experience from being used to imply another.
Rank #3
- The Game Console 2.0: A Photographic History from Atari to Xbox
- No Starch Press
- ABIS BOOK
| Evidence | What it measures | Who or what it covers | What it can support |
|---|---|---|---|
| “The Virtual Census” (2009) | Character representation in game content | A sales-weighted sample of 150 games across nine platforms, compared with the U.S. population | A historical finding of overrepresentation and underrepresentation in that sample; not a current industry-wide count |
| Geena Davis Institute with Nielsen (2025 report; June 2024 survey) | Self-reported likelihood to play based on character diversity | U.S. and U.K. respondents in Nielsen’s Video Game Tracker | Respondent sentiment; not actual sales or a causal effect |
| ADL (2025) | Harassment observed during experimental play sessions | 15 participants, four named competitive games, almost 250 cumulative hours, late 2023 through summer 2024 | Evidence that harassment occurred in those sessions; not a population prevalence estimate |
| Kaatz and colleagues (2017) | Learning outcomes from an educational role-playing game | 47 scientific workforce trainers at an NIH training-program directors’ meeting | Suggestive findings about perspective-taking and bias literacy in that context; not proof of lasting change among general gamers |
What can game makers do?
The ADL recommends a set of measures aimed at prevention, response, and accountability. They are recommendations, not guarantees that any single change will eliminate abuse.
- Make policies identity-aware. Rules and design should address identity-based abuse explicitly, rather than leaving players unsure whether it is covered.
- Strengthen voice-chat moderation. Improve how abusive conduct is detected and addressed in live communication.
- Make reporting and support usable. Give players clear ways to report incidents and support targets after they do.
- Publish regular transparency reports. Explain how reports are handled and what enforcement actions are taken so players can assess accountability.
- Encourage prosocial behavior. Design systems and community practices that reward constructive interaction, not only punish harmful conduct.
Representation is a separate design responsibility: teams can examine which characters are included and how they are written, while recognizing that a more varied cast alone does not resolve hostile player interactions.
Can games help people recognize bias?
Games can also be used as educational tools, though evidence from a specific training context should not be generalized beyond it. Fair Play is an educational role-playing game in which players take the perspective of Jamal, a Black male graduate student facing discrimination in a science PhD program.
Rank #4
A 2017 mixed-methods evaluation by Kaatz and colleagues involved 47 scientific workforce trainers attending an NIH training-program directors’ meeting. The authors’ findings suggest Fair Play can promote perspective-taking and increase bias literacy among that group. The evaluation does not establish durable behavior change, or show that the same outcomes would occur among the general gaming population.
As ADL put it in its 2025 report, “Hate and harassment in games are not inevitable and should never be normalized.” The evidence supports treating representation and player safety as connected but distinct responsibilities: measure each with the right methods, then match responses to the problem being addressed.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minute
