- Do we even have a way of measuring serotonin levels in a live human (as opposed to a dead lab rat[1])? Remember, we're trying to measure serotonin levels in the synapses between neurons in the brain (serotonin is a neurotransmitter, so its levels anywhere else in the body probably wouldn't matter), so you can't just draw someone's blood and test it. If the measurement involves drilling a hole through someone's skull, you can be sure it hasn't been done on lots of people (no ethics board would allow that experiment, even if you could find non-depressed people to volunteer for it). If it requires injecting people with radioactively tagged serotonin precursors and scanning their brains, it probably hasn't been done on a lot of asymptomatic people (risky and expensive).

- Assuming we can measure serotonin levels in humans, how do we determine whether someone's level is "low"? Does "low" mean "low compared to the average of non-depressed people"? Does it mean "low compared to the level the patient had before they became depressed" (if so, how would we know their baseline levels)? Levels of many chemicals (hormones, glucose, sodium, etc.) vary significantly between "normal" individuals.

[1] We can't talk to a rat to determine whether it's depressed. We can only observe certain aspects of its behavior, like appetite and activity levels, which we assume are proxies for depression, but could be due to unknown factors. We don't really know if rats have evolved to become depressed the way people do.

IIRC, the current consensus among the psychiatric community is that SSRIs do work but no one actually knows what mechanism could explain their effectiveness. The one thing that is known for sure is that it doesn't come down to simply high vs low levels of serotonin because while SSRI treatment immediately raise serotonin levels the therapeutic effect takes some time to "kick in".

I also recall reading that SSRIs have been found to be more helpful against anxiety than depression.