A/B testing empirically measures the value of any technique to improve the value being optimized; if you the value you are optimizing for (based on what it is you are measuring in the A/B test) is not "effectively communicates the value of your product", but, e.g., "increases conversions", then dark patterns which optimize the value sought (increasing conversions) by "confusing the user into spending money" will be favored by A/B testing.

There is nothing intrinsically favoring honest communication or opposing dark patterns about A/B testing.