1. Not all, or even necessarily most, fandom is focused on fanfic, or even shipping. I know it may seem that way on Tumblr, but Tumblr is just one website that has its own demographic bias toward the types of fans who tend to write fanfiction. Especially in fandoms where the canon doesn't focus much on romance, and/or where there are a lot of cis dudes (who largely don't write fanfic), the majority of fandom activity may not have anything to do with fanfiction or shipping. And in those cases, there can be vast gaps in which characters or discussion points or relationships are the most popular. (Sometimes it can be the exact reverse: for instance, Julian Bashir is the most popular Star Trek: DS9 character on AO3 by a significant margin, but one of the least popular mains outside of fanfic fandom. Whereas Benjamin Sisko and Quark seem to be most popular in broader DS9 fandom, but are only the 6th and 5th most popular of the 10 main cast members on AO3. There's a ton of discourse about especially Sisko's popularity that doesn't seem to recognize this disparity at all.) Fanfiction has a notable bias toward "which characters have relationship potential" and, in canons that are "complete" (e.g. shows that have finished their runs as opposed to ongoing shows), toward non-canon relationships or those with unsatisfactory endings, since the biggest bias of fanfiction is toward "things that didn't happen in canon that the writers wish had happened." Canon ships tend to be more popular during a series run since fans don't know the ending yet, but once it's over, a subtextual ship, a canonical ship that was underexplored, or a canonical ship where one or both were killed off, etc. is always going to be more popular than an endgame canon ship that was fully explored where fans are largely fine with how it ended (which is another mark against a character like Ben Sisko, especially in contrast to one like Bashir who is both one half of a juggernaut non-canon M/M ship and had an endgame canon relationship that fans largely dislike).

2. AO3 is neither the only fanfiction website nor is it the most popular, and it has some notable biases the other big sites don’t have — particularly toward M/M (which makes up the largest group of fics on AO3), and toward queer content more broadly (there’s also significantly more F/F on AO3 than on the other major fanfic websites), thanks to the fact that it was founded partly in response to censorship of queer fanworks by other websites. When someone crunched the numbers on the big three websites in Anglophone fanfiction fandom (AO3, FF.net, Wattpad) they found that overall, F/M outnumbered both M/M and F/F combined, and gen was about as common as M/M. Even if you're looking specifically at fanfiction fandom, AO3 stats alone are going to miss out on a lot, unless you're looking at the fanfiction writers in a fandom that is predominantly M/M and/or F/F with little in the way of F/M or gen -- but you're still probably going to have to look at FF.net, Wattpad, etc. to know if that's even true in the first place! And even some fandoms that are like that might have their own archives that are more useful: e.g. you'll probably find more useful stats on the F/F-heavy MLP:FIM fandom by looking at FIMFiction than AO3. (Look, I dated a brony a decade ago and y'all know my memory for fandom nonsense is way too good, don't judge me for knowing this)

3. AO3 was founded in 2008 and went into open beta in 2009, which means it is heavily biased toward fandom from the last 15 years. While some older fic has been uploaded to AO3, the vast majority of it hasn't, and so AO3 isn't very representative of what fandom, including fanfiction fandom, was like before that year. For any fandom which existed in any significant numbers before 2009, AO3 just isn't going to work for studying the fandom at all unless you are specifically looking at more recent stuff. (Star Trek again is a great example here: the "Kelvin Timeline"/J.J. Abrams movie trilogy, known to Trekkies as AOS, has about 16k more fics on AO3 than the Original Series from the 1960s, or TOS. But if you know literally anything about the history of fanfiction, you know the idea that there is overall more AOS fic than TOS fic -- or even just more of it on the Internet -- is completely ludicrous. But AO3 premiered the same year as the first AOS movie, a hit mainstream movie that was a lot of younger fans' introduction to Star Trek, whereas TOS fandom existed for 43 years before AO3 was publicly available!)

AO3 stats work fine if you're looking at a specific subset of fandom that uses AO3. It could be very useful if you're studying, say, English-speaking Yuri!!! on Ice fans, or Destiel shippers, or Clexa shippers -- take your pick. It's obviously useful for anything that is specifically focused on AO3 as a community, like the current campaign about racism. But if you're using it to make big, sweeping pronouncements about what broader societal attitudes might be responsible for a character or other thing's overall fan popularity, you cannot just use AO3 stats to gauge that without getting some sense of a) how much whatever you're studying is a Thing in queer fanfiction spaces relative to the rest of fandom and b) if it isn't, if the AO3 stats are representative of the broader fandom outside of it, including both the people who prefer other fanfiction websites and non-fanfiction fandom. Both of which require poking your head out of AO3 into what the rest of fandom is doing. For example, if you want to analyze if racism, misogyny, or homophobia plays a role in the popularity or lack thereof of a particular character: have you looked at what the more het- and gen-heavy corners of fanfic fandom think? Have you looked at what other corners of fandom think? Do they follow those same patterns or are they different? Because queer fanfiction fandom, for all its many issues with racism, probably isn't more racist than more cis-dude-heavy corners of fandom (who are, after all, the corner whose racist tantrums and harassment campaigns are so massive that they make mainstream news and lead to stuff like POC actors being forced to quit social media for their own safety). It certainly doesn't make sense to conclude that "AO3 users" are more misogynistic and/or homophobic than other corners of fandom given that AO3 users are disproportionately likely to identify as women and/or LGBTQ+ compared to the rest of fandom. So if you have a marginalized character who is more popular with non-AO3 fandom, like Sisko (a black man), that should prooooobably be taken into account before assigning "racial/gender/etc. bias" as the only or primary reason for those AO3 stats!* You can still use AO3 to discuss those patterns, of course, but you have to a) be clear that you're specifically looking at AO3 users and/or b) put those stats in context with the rest of fandom, including how AO3 and its users are different from other fandom. So if a character is significantly more or less popular on AO3 compared to other corners of fandom, and you want to understand why, you might want to start by asking questions like "how do female and queer fans feel about their canon romance?" and "do they have a romantic arc and/or a lot of subtext with a character of the same gender?" before you get into all the other stuff.

You would not believe the amount of otherwise-good analysis, including from academics and from professional media critics reporting on fandom for major news sites, that don't do this. (And before the drama-hounds start trying to extrapolate who I'm talking about here, this really isn't inspired by anyone specific; that's how common this particular mistake is.)

Context is all, my friends. And one of the reasons that a lot of fandom studies academics I've talked to discourage studying fandoms where you're deeply invested in the discourse in a particular corner of it is because it's really easy to mistake your personal experiences for being more representative of the whole than they are. If you want to draw conclusions about the whole of a thing, you've gotta actually look at the whole -- or at least, be aware of whether whatever part you're looking at actually represents the whole. This is the same problem social science researchers run into when they assume that studies done on just "university students in the U.S." can be extrapolated to all of humanity when a) the vast majority of humanity doesn't live in the U.S. and b) even within the U.S. itself, most people don't go to university. The stem, the peel and the flesh all make up parts of an apple, but there's significantly more of one of those three things in the apple than the other two. Make sure you know that the chunk you're looking at is the flesh, not the peel or the stem.