AI, QA, and the Future of Game Testing

Square Enix’s plan to automate the majority of its game quality assurance (QA) and debugging with generative AI has ignited one of the most heated debates in recent games industry history. When a publishing leader from Baldur’s Gate 3 developer Larian Studios publicly called the idea of large-scale replacement of QA workers “stupid,” it became clear that this is more than a technical experiment: it is a battle over the future of work, creativity, and trust in games.

This article explores what Square Enix is attempting, why Larian and others are pushing back, and what AI-driven QA could mean for players, developers, and the next generation of talent entering the industry.

How Square Enix Plans to Use AI in QA

Square Enix has outlined a strategy to transform its testing pipeline by heavily integrating generative AI into QA and debugging. According to the company’s disclosures and subsequent reporting, the publisher wants AI systems to handle roughly 70% of these tasks by 2027, dramatically reducing its reliance on human testers.¹

Key Components of the Plan

• Automation target: Aim for AI to perform about 70% of QA and debugging work.
• Research partnership: Collaboration with the Matsuo-Iwasawa Laboratory at the University of Tokyo to develop AI-based QA automation.¹
• Efficiency rationale: Square Enix frames the initiative as a way to cut costs and achieve a “competitive advantage” in development.
• Financial angle: Public commentary around the plan has highlighted projected savings in the tens of millions of dollars by reducing headcount in QA.

While automation in testing isn’t new, the scale of this ambition is. Most studios today blend automated tests with manual QA, but Square Enix is explicitly targeting a future where AI performs the majority of the work currently handled by people.

The University of Tokyo Collaboration

The partnership with the Matsuo-Iwasawa Laboratory is focused on “Game QA Automation Technology Using Generative AI,” assembling a joint research team of more than ten members from both the university and Square Enix.¹ Their goal is to build systems that can automatically generate, execute, and interpret tests during game development, reducing the amount of repetitive work for humans.

Why Larian’s Publishing Lead Pushed Back

Larian Studios, the developer behind the critically acclaimed Baldur’s Gate 3, has publicly criticized Square Enix’s approach. Its publishing lead argued that the belief QA staff can be replaced at scale is not just naive but dangerous for the overall health of game development.

Core Concerns Raised by Larian

• Devaluation of human testers: Suggesting that AI can stand in for a majority of QA roles sends a signal that human judgment, creativity, and lived experience are expendable.
• Short-term savings vs. long-term damage: Larian’s representative emphasized that chasing immediate cost reductions could undermine quality, reputation, and future talent pipelines.
• Misunderstanding of QA’s role: The criticism suggests that Square Enix views QA as a mechanical, number-crunching process rather than a creative and analytical discipline in its own right.

The clash is also symbolic. Baldur’s Gate 3 became a poster child for meticulous QA, launching in a surprisingly polished state for such a sprawling RPG. That success strengthened Larian’s credibility when arguing that human-led testing and iteration are key to delivering complex, stable games.

What QA Actually Does in Game Development

Outside the industry, QA is often reduced to the stereotype of “bug hunting.” In reality, the discipline is a multilayered craft combining technical testing, design sensitivity, communication, and documentation.

Key Responsibilities of QA Teams

• Functional testing: Verifying that basic systems work as intended—menus, controls, saving, and loading.
• Systems and balance testing: Checking how mechanics interact in complex games, from combat loops to progression systems.
• Exploratory testing: Deliberately trying to break the game in creative ways, including “playing like a player never would.”
• Regression testing: Ensuring that new changes haven’t reintroduced old bugs.
• Qualitative feedback: Providing impressions on feel, confusion points, difficulty spikes, and UX issues that aren’t strictly “bugs.”
• Documentation and communication: Writing clear bug reports, reproductions, and communicating priorities with designers and engineers.

Human testers also act as early representatives of players, surfacing problems in pacing, accessibility, and clarity that are difficult to quantify. This aspect is especially crucial in story-rich titles like Baldur’s Gate 3, where narrative decisions intersect with gameplay in highly variable ways.

What AI Can and Can’t Handle in QA Today

AI and automation are already part of QA across software industries. Unit tests, automated UI checks, and scripted bots are common. Generative AI extends this by allowing systems to generate test scenarios, simulate players, and analyze logs at scale.

Strengths of AI-Driven Testing

• Scale and speed: AI can rapidly explore permutations—different hardware, configurations, or simple gameplay loops—much faster than humans.
• Repetition tolerance: Machines don’t fatigue; they excel at repetitive regression tests and can continuously run test suites.
• Pattern recognition: AI can mine telemetry data to spot trends and anomalies that might correlate with crashes or softlocks.
• 24/7 availability: Automated testing can run overnight or across global infrastructure without scheduling constraints.

Limitations and Risks

However, current AI techniques have clear limits, which mirror broader concerns about generative AI in other fields.

• Lack of human context: AI doesn’t understand story, tone, or emotion in the way players do. It may flag or miss issues that hinge on narrative coherence or character motivation.
• Blind spots in novelty: AI tends to perform best in domains similar to its training data. Highly novel game mechanics or emergent systemic interactions can confuse or mislead it.
• Over-optimization risk: If AI is used to hit easy-to-measure metrics (like crash rates) without human oversight, subtle quality issues can proliferate.
• Bias and gaps: Automated systems might fail to represent diverse player experiences, including accessibility needs or different play styles.

Major technology firms researching AI in software testing acknowledge these trade-offs. For example, Microsoft’s work on AI-assisted testing emphasizes AI as a complement rather than a replacement, supporting engineers and testers rather than eliminating them.²

The Broader Context: Automation, Jobs, and Ethics

Square Enix’s initiative sits within a much larger conversation about AI and work. International organizations such as the International Labour Organization (ILO) and OECD have examined how generative AI is likely to impact employment. Their findings suggest a nuanced picture: many tasks may be automated, but outright job replacement is far from inevitable in the short term.^3,4

What Research Says About AI and Jobs

• Task-level disruption: The ILO has found that generative AI is more likely to change the way jobs are performed than eliminate them entirely, especially in knowledge and service work.³
• Augmentation vs. substitution: OECD research highlights that AI has strong potential to augment workers—taking on repetitive tasks while humans handle complex judgment—but risks grow when firms pursue pure cost-cutting.⁴
• Uneven impact: Lower-wage, entry-level roles are often more vulnerable, which is exactly where QA sits in many studios.

From this perspective, Larian’s criticism aligns with broader policy concerns: if companies use AI primarily as a tool to reduce headcount and labor costs, the social and creative consequences may be severe.

Why QA Jobs Matter for the Industry’s Talent Pipeline

In game development, QA has historically functioned as an entry point. Many designers, producers, and even programmers began their careers in QA, learning how games are built by testing them day after day.

QA as a Training Ground

Working in QA gives aspiring developers:

• Hands-on exposure: Daily interaction with builds, tools, and development branches.
• Cross-team contact: Regular communication with designers, engineers, and artists over bug reports and feature changes.
• Systemic understanding: A high-level view of how different components of a game connect and break.
• Soft skills: Practice in documentation, communication, and prioritization—skills that carry over into production and leadership roles.

Replacing large portions of QA with AI doesn’t just affect current testers; it shrinks the ladder into the industry. Larian and other critics argue that this will make it harder for new voices to enter development, leading to a narrower and possibly less innovative future.

Potential Player Impact: What Could Change On-Screen?

For players, the debate can feel abstract—until it affects the games they play. If AI-driven QA becomes the norm, several possible outcomes emerge.

Possible Benefits

• Fewer technical crashes: Automated systems may be excellent at catching repeatable, low-level bugs across countless hardware combos.
• Faster patches: With AI running continuous tests, post-launch fixes might roll out more quickly.
• More stable live services: For online games, AI monitoring could detect server-side issues before they escalate.

Possible Downsides

• More “soulless” issues: Games might technically “work” while still feeling unpolished in pacing, storytelling, or UX.
• Accessibility oversights: Subtle accessibility barriers may be harder for AI to detect without focused human testing and feedback.
• Homogenized design: If AI testing shapes how systems are built and tuned, developers might gravitate to designs that are easier for automated tools to handle.

Recent high-profile releases show how fragile trust in AAA launches can be; buggy or poorly optimized games are harshly criticized by both media and players. Publishers experimenting with aggressive automation risk exacerbating this trust deficit if quality slips.

Toward a Hybrid Future: AI as a Tool, Not a Replacement

Few experts argue that AI should be excluded from QA. The real fault line is between using AI to augment human teams and using it to justify large-scale layoffs. A more sustainable approach likely lies in a hybrid model.

What a Healthy Hybrid Model Could Look Like

• AI for repetitive checks: Use automation for regression, performance sweeps, and basic compatibility testing.
• Humans for high-level judgment: Keep human testers responsible for exploratory, narrative, and UX testing.
• Transparent governance: Involve QA leads in decisions about how AI tools are deployed, including guardrails and escalation paths.
• Continuous training: Upskill QA staff to work alongside AI—designing tests, interpreting results, and validating automated findings.

Policy bodies like the OECD and ILO recommend such augmentation-focused strategies, noting that they preserve human agency and can raise productivity without hollowing out entire professions.^3,4

FAQ: AI and Game QA

Conclusion: A Test the Industry Can’t Afford to Fail

Square Enix’s aggressive AI QA strategy and Larian’s outspoken rebuttal have crystallized a pivotal moment for the games industry. On one side is the promise of faster, cheaper, and more scalable testing; on the other is the risk of eroding human expertise, damaging talent pipelines, and releasing games that technically “work” but feel less alive.

Research from labor and economic bodies suggests that the healthiest path forward treats AI as a powerful tool that extends human capabilities rather than displacing them wholesale.^3,4 For game developers, that means embracing automation where it truly helps, while defending the irreplaceable value of human testers who understand not just how games function, but why they matter to players.