Christ - OK- I've read it.
It doesn't 'debunk' it - it quibbles with it - eg. is a 'visit' the same as a 'treatment', tries to invoke surveillance bias - when both the treated and un-treated youths were under observation. The criticism continues with the index date being when someone walked into the clinic - not when 'treatment started' (clearly they don't consider treating co-morbid psychiatric symptoms as 'treatment)
He's very miffed that specialised psychiatric treatment is treated as a binary thing (either you did, or you didn't have treatment), rather than graded levels of required contact (ie. the idea that just being referred to a psychiatrist indicates anything if the psychiatrist then goes on to give them the all clear - and note, he doesn't say any psychiatrist did give them the all clear, just that they might have, and that's why he doesn't like the did/didn't have contact as a measurement) - btw this is a limitation freely admitted in the original study, and previous psychiatric treatment was indeed accounted for.
Then does some misdirection by using an analogy comparing people with cancer to the general population (the better - although still imperfect - analogy would be people with a lump, and oncology visits split in to lumps that turned out to be cancer and didn't turn out to be cancer).
Then he decides that it's highly likely that it's the review writer herself who's clearly causing the psychiatric issues in the patients.
He doesn't refute anything in fact - but just waves his hands at possibilities - eg the time from a patient walking into the clinic to getting a prescription - he says can be much longer than the 2 years the study covered - but doesn't say what the average is. He clearly didn't read the study, which clearly said that it wasn't a solid '2 years' but:
Psychiatric treatment, excluding gender identity assessments and appointments with a multi-disciplinary team therein, subsequent to the index contact was defined as continuing two years or more after the index contact, to allow time for the gender identity assessment and potential medical GR initiation
So basically, he (or ChatGPT) read the doc and looked for any possible cracks to exploit, but didn't read it very well, and avoided actually giving any concrete evidence against it.
Oh, and he literally attempts to discredit it by saying the journal it was published in wasn't very good, so I think we're allowed to criticise him as a source of knowledge in return