Abstract: In one example, a method for evaluating screen content for accessibility with a screen reader device is disclosed. The method provides a baseline document including a script of expected screen content that conforms accessibility requirements. The method may generate an audio file based on screen content elements. For some implementations, the method uses a machine learning model to transcribe the audio file into an output transcription file. The method may determine whether output transcription file matches the baseline document and a corresponding output report is generated.