Toxicity in your game is costing you money. Players aged 13 to 25 spend 54% more each month on games they see as non-toxic than on ones they see as toxic. An ADL survey found that 20% of gamers are spending less in online games because of the hate and harassment they encounter.

Toxicity in online gaming is harmful behaviour between players. It ranges from harassment and hate speech to threats and griefing, across text chat, voice and user-generated content. It drives gamers away and damages brands at launch. It now carries regulatory risk under the EU's Digital Services Act (DSA) and the UK's Online Safety Act. Content moderation spots harm in real time, responds proportionately and gives players clear ways to report and appeal.

 

What is toxicity in online gaming?

Toxicity is any player conduct that makes a game feel hostile, unsafe or not worth returning to. Trash talk has always been part of gaming culture. The line is crossed when players target people instead of playing. Three forms are most common:

  • Harassment and bullying: repeated insults or singling out one player.
  • Hate speech: abuse aimed at someone's race, religion, gender, sexuality or disability.
  • Cheating-related abuse: rage after a loss, griefing teammates and threats to doxx.

 

Toxicity by the numbers

https://checkstep-website.transforms.svdcdn.com/production/assets/global/02-graphic-1-stat-cards.png?w=1600&h=560&auto=compress%2Cformat&fit=crop&dm=1795110044&s=8a2a5ccef3d6e597d02c4be40d120fa5

 

In research for the ADL, a group of students, recent graduates and young adults played Valorant, Counter-Strike 2, Fortnite and Overwatch 2. They used religious, ethnic and national identity-based usernames. Hate and harassment showed up in almost half of the game sessions tested (ADL, 2025).

 

Safety is a business issue

Trust & Safety is often treated as a support function. In practice, it touches every line of the P&L. It decides whether acquisition spend is lost to early churn, and whether parents and platform partners back your game.

Research puts numbers on this. For players aged 13 to 25, average monthly spend was $21.10 on games seen as "non-toxic", compared with $12.09 on "toxic" ones (Steinkuehler, 2023). Take This found that 6 in 10 gamers had decided not to spend money in a game because of how others behaved. Female, new and "casual" players are more likely to pull back from gaming spaces after hateful comments and actions.

https://checkstep-website.transforms.svdcdn.com/production/assets/global/03-graphic-2-business-cost-grid.png?w=1600&h=820&auto=compress%2Cformat&fit=crop&dm=1795110082&s=e879cddd188b4468c0d8414338b4d4cb

 

How moderation helps

Effective moderation means understanding harmful behaviour and acting on it quickly, fairly and at scale. Catching every bad word isn't the aim.

Real-time chat and voice

In-game chat moderation has to match the speed of play. If a decision takes seconds, the damage is done. Low-latency checks run in milliseconds, so abuse is blocked or masked before anyone reads it. For voice, studios can pair live detection with transcription of reported clips.

AI and human review

AI handles the volume. Clear-cut harm is removed or muted instantly. Grey-zone and sensitive cases go to trained moderators with full context. Modern models read intent, not exact spellings, so they catch new slang, swapped letters and coded terms. Your team can update policy with fresh examples in minutes. Player numbers can grow without a matching rise in headcount.

Banter vs abuse

Over-moderation can drain a community as fast as abuse. Context separates the two. A player who crashes and types "I'm such an idiot" is laughing at themselves. "I hope you die" aimed at an opponent is harassment. Contextual moderation reads the full conversation, so friendly trash talk stays while targeted abuse is caught. Some studios mask flagged words. Others swap harsh messages for a playful line in the game's own voice.

https://checkstep-website.transforms.svdcdn.com/production/assets/global/04-banter-vs-abuse.png?w=1600&h=720&auto=compress%2Cformat&fit=crop&dm=1795110114&s=c9835f9323ac3a0b162f888112aecde6

 

What regulators expect

If players can chat or share content in your game, both laws likely apply. The DSA expects easy reporting, a statement of reasons for every restriction, an appeals process, transparency reports and measures to protect minors. The Online Safety Act requires UK services to reduce the risk of illegal content and protect children. Ofcom can fine companies up to £18 million or 10% of qualifying worldwide revenue, whichever is greater.

In 2022, Fortnite maker Epic Games agreed to pay a $275 million penalty over children's privacy and another $245 million to consumers over manipulative design. According to the FTC complaint, Epic employees had raised concerns about chat safety for children as early as 2017. They cited reports of bullying, sexual harassment and exposure to self-harm content. As part of the settlement, the company is barred from turning on voice and text chat for children unless parents give affirmative consent.

Roblox is also facing lawsuits. Attorneys general in Texas, Louisiana, Florida, Kentucky, Iowa, Tennessee, Nebraska and Arkansas have sued the company, alleging child safety failures. Since January 2026, the platform has required all users worldwide to complete a facial age estimation scan to access chat. On 18 November 2026, Arizona's attorney general asked a court to make Roblox hand over information for a child safety investigation.

Compliance is far simpler when policies, reports, decisions and appeals run through one system from launch, not stitched together later.

 

Case study: Star Stable

Star Stable Online is one of the world's largest online horse adventure games. It processes tens of millions of chat messages a month in 14 languages. With Checkstep, it replaced separate per-language models and weeks-long update cycles with a single multilingual approach. Decisions take under 50 milliseconds. Policy updates go live in under a minute, and accuracy rose by around 50% at launch.

As its player base grew, Star Stable's existing moderation struggled to keep up with scale and speed. Harmful content was sometimes missed, only to surface later in screenshots and community forums.

Player language also moved fast. Younger players invent slang and coded terms, often to dodge moderation rules, and Discord communities accelerate that drift. Keyword lists built at launch quickly fell out of date.

Updates were slow. When new harms or cultural references emerged, retraining and deploying models could take weeks, leaving the team reacting rather than shaping its community.

Multilingual support added complexity. Separate models for each of the 14 main languages meant duplicated effort and thousands of messages labelled by hand each week. Many of them were stale by the time they were reviewed.

Traditional LLM-based approaches also added noticeable in-game delays, making them unsuitable for live chat. Star Stable needs decisions in tens of milliseconds so harmful content never reaches the screen.

The new setup is purpose-built for high-volume gaming chat. It lets Star Stable run an AI-first approach behind the scenes, with human-in-the-loop oversight keeping the community healthy.

"Getting started with the Checkstep platform was even easier than expected, a pain-free launch!"

Estella Corbellini, Senior Producer & Localisation Lead, Star Stable

Read the full Star Stable case study

 

FAQs

What causes toxicity in online games?

Competitive pressure, anonymity, frustration after a loss and a lack of visible consequences all fuel toxicity. It thrives in open chat, voice and large public lobbies.

Does content moderation improve player retention?

Yes. Content moderation improves retention because players who feel safe stay longer and invite friends. It removes the hostile experiences behind early churn, especially among new and younger players.

 

Build safer communities from day one

Toxicity isn't a cost of doing business. It's a problem you can measure, manage and reduce. Checkstep helps studios moderate text, voice and live chat in real time while staying compliant with the DSA and Online Safety Act.