Big tech companies have indeed been encouraging the development of algorithms that can detect deepfakes. These algorithms commonly look for the following signs to determine if content is a deepfake: Number of words spoken per sentence, or the speech rate (the average human speech rate is 120-150 words per minute); Facial expressions, based on known co-ordinates of the human eyes, eyebrows, nose, lips, teeth and facial contours; Reflections in the eyes, which tends to be unconvincing (either missing or oversimplified), Image saturation, with AI-generated images being less saturated and containing a lower number of underexposed pixels compared to pictures taken by an HDR camera.