Reading a comparison written by a direct competitor

We publish comparisons against rivals. So do they. Neither is neutral, and both can still be useful if you extract the right layer.

The layer worth taking

The test design. What face was used, how many source photographs, how many headshots generated, how many were usable. That is portable — run it yourself and the incentive problem disappears.

The layer to discard

The conclusion. A company comparing itself to a competitor reaches a predictable result. That is structure, not dishonesty, and treating it as either revelation or scandal misses the point.

Three questions that expose a weak comparison

Does it name the source photographs? Does it show a run that failed? Does it carry a date?

Most do none of the three, which tells you they curated rather than tested.

What a portrait comparison should control

The face used, the number of source photographs, and how many headshots came out usable. Anything without those three is an opinion with pictures.

The version nobody publishes

Take the face you are least confident about — yours, or a colleague with strong prescription glasses — and run it through two free tiers. Twenty minutes, and no affiliate revenue for anybody.