Let me explain what the AI processing on this site does and why it increases the usefulness of the site immensely. There's an automatic process which runs daily, that collects reviews from external sites (currently, only one), analyses them, and if it finds anything useful, it incorporates that "knowledge" into this site.
The way that users provide reviews on this site, are in the form of text comments, and ratings (using our rating widget) for various effects, together with a list of side effects that may have been experienced. Other sites, however, almost always use an extremely simplified approach, where the user gives comments and an overall rating. The utility of our site is really highlighted in the Supplement Explorer, where you can find supplements by the goals that suit you the most, and that's where textual reviews by themselves are not useful.
Our AI processing - built from the ground up - uses the latest techniques in natural language research (including some of our own, unpublished work) to extract the effects and side effects that the reviewer discusses in the text, together with the sentiment - that is, whether it's positive, negative, or neutral.
Now, this process is also rather fallible, and is measured in several ways - precision, recall, and accuracy being common ones. We have achieved pretty good numbers for all these measures, compared to the academic research, but you'll find that the system does come up with incorrect analyses a certain percentage of the time.
One of the areas that really needs improvement is in the analyses of negations, .e.g. where "isn't", "hasn't" or "wasn't" is used in the sentence.
Over time, with improved training and better algorithms, the results can get better. In the meantime however, it has meant that we've allowed our system to grow immensely from just under 2,000 native reviews, to over 63,000 analyzed reviews in total. I can't emphasize enough, the benefits are most apparent when you're using the Supplement Explorer!
Feel free to discuss the AI processing here.