This study investigates whether employers in an online hiring experiment exhibit discrimination based on workers’ accents that indicate English is not their primary language. To assess accent bias, we implement a randomized treatment design in which participants acting as employers are assigned to one of two conditions: a treatment where the worker’s accent is revealed (“Accent Revealed”) or a control where it is not (“Accent Blind”). Using incentive-compatible methods, we elicit employers’ beliefs about the productivity of randomly assigned workers, providing brief demographic information and audio clips that either reveal or mask accent characteristics. We evaluate worker productivity in two skills: Mathematics and Verbal reasoning. We find that employers rate accented workers as less capable than their non-accented counterparts in both skills, and this gap persists after providing employers with a signal of a worker’s test score. Employers also display lower willingness to pay, particularly in Verbal skills tasks, even when provided with performance signals.