Just because it's approximated by a bell curve doesn't make it a bell curve. There are quite obvious separate phenomena shaping the curve at different times.
Are you discarding the utility of Gaussian functions in analysis simply because the independent variable is time? A Gaussian curve can be used as a descriptive model without claiming that the observations themselves are a probability distribution
The fit does not prove causation, but it does show that the decline was already well described by a trend that began years before generative AI. If the claim is that 2023 created a separate structural break, it's different claim then the title describes