How Meta Chose Workers for Layoffs: Sworn Court Filing Reveals the Decision Process

When Meta announced another round of layoffs in May, many employees were left wondering how the company had decided who would lose their jobs. Some former workers claimed artificial intelligence played a role. Others questioned whether employees on parental leave or those holding company-sponsored visas had been unfairly affected.

A new sworn court filing has now provided the clearest account yet of how Meta says those decisions were made.

The 13-page declaration, submitted under oath by Linh Doan, Director of HR Business Partner Enablement at Meta, forms part of a lawsuit brought by 26 former employees. The plaintiffs allege the company relied on internal AI tools, productivity data, and AI usage metrics when selecting workers for layoffs. Meta denies those allegations and says every redundancy decision was made by human managers.

Judge Sought Answers Over Four Visa Holders

The declaration was filed after a federal judge ordered Meta to explain why four plaintiffs, whose work visas were sponsored by the company, had been selected for redundancy.

The judge declined to temporarily block the layoffs, but required Meta to provide a sworn explanation because those employees faced a limited period to secure another sponsoring employer or risk losing their legal right to remain in the US.

The filing explains Meta's decision-making process for those employees, but does not discuss the immigration deadline they faced.

Meta Says AI Played No Role in Layoff Decisions

One of the strongest statements in the declaration addresses the central claim in the lawsuit. Meta states that AI was not used to determine who would be laid off.

According to Doan, business leaders first identified which parts of the organisation would be affected. They then grouped employees into specific cohorts based on job level and role before establishing objective selection criteria linked to business needs.

Doan stated that the criteria had to be finalised before any individual employee was assessed. She also said managers were not allowed to depart from those criteria once they had been approved. The declaration further states that Meta's employee activity monitoring tool launched on 22 April, after the layoff selection process was already under way. Redundancy notices were issued on 20 May.

Performance Ratings Became a Key Selection Factor

The filing also provides a rare insight into Meta's internal performance review system. According to Doan, Meta uses seven performance ratings.

One plaintiff received a Consistently Met Expectations rating in the 2025 review cycle. The declaration describes this as the lowest rating that still meets expectations. Despite meeting expectations, the employee was selected because Meta had decided to reduce headcount at that level while retaining employees with stronger historical performance.

The declaration also explains that Met Most Expectations is regarded internally as a below-expectations rating. Those definitions have attracted attention because their wording differs from how many employees might interpret them.

Organisational Structure Also Influenced Selections

Meta says performance ratings were only one part of the process. According to the declaration, selection criteria could also include job level, tenure, location, specialised skills, and organisational structure.

The filing explains that managers with relatively small teams could be selected if the business wanted to reduce management layers. One example referred to employees working six reporting levels below Chief Executive Mark Zuckerberg. Managers overseeing fewer than seven direct reports could also be considered where larger teams averaged around 10 employees. Meta says these criteria were established before employees were individually evaluated.

Leave Status Remains a Disputed Issue

The lawsuit alleges that workers who had taken parental leave or disability leave were disproportionately affected by the layoffs. It also claims internal AI usage dashboards failed to account for periods when employees were away on authorised leave. Meta disputes those allegations.

According to Doan, managers making redundancy decisions did not know whether an employee had taken leave. She stated that leave status, disability status, accommodation requests, and other protected characteristics were not included in the selection criteria.

One plaintiff was dismissed less than two weeks after returning from maternity leave. Meta stated that nearly 40 per cent of that employee's team had also been selected for redundancy, arguing that she was not singled out.

Questions Around AI Adoption Remain Unanswered

While Meta firmly denies that AI selected employees for redundancy, another issue remains unresolved. The declaration does not state whether employees' use of AI tools influenced their performance ratings, even though historical performance formed part of the layoff criteria for some workers.

The lawsuit alleges that AI adoption became an important workplace metric through Meta's Checkpoint programme. It also cites claims from current and former employees who said AI usage dashboards did not reflect periods when staff were on authorised leave, causing their recorded activity to fall. Meta has denied that AI made layoff decisions but has not publicly addressed those specific allegations within the sworn declaration.