It's well within the capability of ai right now to read a thread and determine what the topic is about, and within our capability to create the system that does that and tags it correctly. With 500k threads, it's a wonderful application of the technology and so far we've found it to be very accurate.
Where we are at right now is we went through a "first pass" to catch the low-hanging fruits of identifying threads for models by the subject line (thread title) content. The next level will be to read the first post as well and identify if a thread title is "problem with steering" and the thread content identifies the make and model. As others have suggested, there's a bit more to the story when you factor in threads that apply to multiple models, such as with attachments or maintenance topics. To answer the question of where comparing tire types go, that would be, for now, in the forums they are in.
We are more likely going to organize the topics by tag with advancements and improvements to our tag pages. For example, tagging all threads about "grading a driveway" and giving users the ability to sort those tag pages by date posted, number of replies, etc.