Congress cares about TikTok. And they should. The app is not all song and dance. Though TikTok wants you think otherwise, it’s fast becoming a major news source for Americans.
Given the billions of hours Americans are spending consuming news in the app, one might wonder what exactly they are watching.
Or to put it another way: what are the most viewed videos for any given day, week or month?
The answer to that question: no one knows.
Because as far as I know — and I’ve also asked several experts (see Notes below) — there is no regular, public resource showing the most viewed videos on TikTok, for any given time period.
Not TikTok itself, with its bafflingly opaque search feature that doesn’t allow sorting by view count.
Not the official TikTok Research API, which imposes a limit of 1,000 requests a day even to the vetted researchers who have to sign a ‘minefield’ terms of service to access it.
Not any third party resource, such as a TikTok equivalent to AlgoTransparency’s YouTube recommendations data.
What’s particularly bizarre is that this missing data is by definition extremely public. The most viewed videos are very much out there, being watched by tens of millions of people every day.
But the data exists in siloes of different users’ TikTok apps, with no way to see across. I might watch one or two >50m videos on a typical day in the app, but this doesn’t help me see all the videos that are getting to this level.
And to try and pull this data together would be extremely hard, as I’ll explore in a follow up piece.
TikTok themselves have this data, of course. Big tech companies like Google, which crawl the internet, could probably create it. And social monitoring companies like Tubular Labs have databases of billions of TikTok videos which they charge $1000s per month for access to, so they have at least some pieces of the missing data puzzle.
But policy-makers, journalists, civil society? The world at large? Nope.
Why this data matters
Policy-makers around the world are hauling TikTok executives in for questioning about their app. Journalists are writing about the effect of TikTok’s spread on society, and civil society are concerned about how it spreads fake news, hate speech, and content that harms kids.
Insane to think that all of this is happening in the absence of basic data: the most viewed videos.
Why do I describe this as basic data?
We could spend a long time trying to understand every inner working of TikTok’s black box algorithm.
But to short-circuit our way to the largest-scale outputs of the algorithm, and get an 80-20 understanding of how TikTok affects the world, we don’t need to know these inner workings.
We just need to know — at least to start with — what the most viewed videos are, because those, by definition, are the videos that will most shape the information entire societies are consuming via TikTok.
Sure, most of the top videos on TikTok will be entertainment. But mixed in, there’ll be all those things we’re worried about: fake news, hate speech, and content that harms kids.
Once again: by definition, the videos that are most viewed are the ones that spread most widely. So if we had a way to regularly check the most viewed videos, we could stop some truly nasty stuff spreading — or at least, have a debate about how to do that.
But, we don’t.
Regulatory moves
There’s a chance that the data access provisions in the EU’s Digital Services Act force TikTok to provide this data for the EU in a couple of months.
Draft legislation elsewhere, such as the US Platform Accountability and Transparency act, may eventually do the same in a few other places.
But that would still leave a black hole where the rest of the world is concerned.
That’s why I’m considering a Challenge of pulling this data together, and putting it into a nice public-facing dashboard.
But to be honest, I’m intimidated by how huge a Challenge this would be. In my next post, I’m going to talk about how I might do it. The gnarly problems with every method I’m considering reveal a lot about the internet in 2024.
Stay tuned!
Notes
Because this missing data surprised me so much, I asked a number of contacts in the world of tech research about it:
AI Forensics, in particular Salvatore Romano, their Head of Research and in-house TikTok expert, and Marc Faddoul, their Director and Co-Founder
Several fellow members of the Coalition for Independent Technology Research
They all indicated that as far as they are aware, there is no way to see the most viewed videos on TikTok. A few of them very helpfully suggested some techniques I could use to try to get a rough sense of this data, which I’ll dig into in my next post.
If I’ve missed something here, please let me know!