It also can’t avoid false positives (a highly conserved sequence could have recently broken and become nonfunctional, therefore have zero FI). But that’s the least of his problems.
His method straight up doesn’t work. It doesn’t detect design. Even if it could detect high FI (it can’t estimate FI from conservation though as already established), that still wouldn’t allow you to infer design as high FI can evolve. In fact, high FI can’t be designed if it can’t also evolve. The exact problem design is postulated to solve because evolution can’t, design can’t solve either. If there is no navigable fitness surface you are forced to do brute force guessing, which isn’t design it’s just random guessing, and then design and evolution are both hopeless.
Disagree? Then “design” your way to a long random password without getting any feedback whether you’re getting closer to the password. You can’t. The feedback is the very thing that makes design (and evolution) possible. If natural selection or something analogous to it isn’t bringing the guesses closer to the solution, there is no way to design your way out of it.