I've been tracking the number of restricted works on AO3 since about 24 hours after the recent Reddit post on ao3 probably being scraped for use in commercial writing AIs. I think it could be interesting to make some meta with that.
Do you, or does one of your followers happen to have data on the number of restricted works on ao3 prior to December 2nd? Or know where I could find it?
Oh, that sounds really interesting! I don't have any recent data about restricted works on AO3 prior to that date (it's vaguely possible I looked at this number at some point in the past, but if so, it was years ago). Any of my followers have relevant data?
In December 2022, I made a post with statistics on the number and percentage of restricted works on ao3, because quite a few people wanted to archive-lock their work due to the reddit post calling attention to the fact that ao3 has been scraped to train AI text bots.
Since then I have been tracking the percentage of restricted works on the archive as a whole once per week.
On May 13th, the OTW made an announcement on 'AI and Data Scraping', which renewed interest in the matter. So, here are a couple of stats looking at the change since last year!
First, a chart that shows the percentage of restricted works on the whole archive, including some individual historic values and weekly values since the beginning of December.
There were 3.38% of restricted works a day after the reddit post (Dec 2nd) and 5.79% at the time of my last post (Dec 16th). The percentage of restricted works dropped slightly and steadily since since then until Mid-May.
The decrease is mostly due to an increase in overall works and the number of restricted works staying relatively constant at around 580,000 works.
The week before the OTW announcement, there were 5.27% of restricted works (May 7th) and since then the percentage has risen to 8.25% of restricted works (June 25th).
Other things (like comment bots) may influence the number of restricted works as well.
Do specific categories drive this increase?
I looked at three ways of categorizing works that I had already made stats for last December and compared the changes since then: Rating, Language and Year of Last Update.
All three charts show the percentage of works restricted in Dec 2022, the percentage of works restricted in June 2023 (exact value on June 25th in chart), the rate of increase in percent between these two values and finally the rate of increase in percent for the total number of works in the category. The category names include the number of total works the categories have in June 2023.
For example, for works with No Rating: There were 1,014,291 works with No Rating on the archive in Dec, of which 3.6% were restricted. Yesterday (June 25th, 2023), there were 1,129,002 works with No Rating on the archive, of which 4.89% were restricted. The percentage of restricted works increased by 35.85% and the number of total works by 11.31%.
Rating:
The percentage of restricted works increased most for Teen and Gen (~50%) and least for Explicit (~30%).
In the last post it was brought up how creators of Explicit works may be more hesitant to make their works archive-locked, because people may only want to interact anonymously.
Language:
The percentage of restricted works increased most for Indonesian and Filipino (~80%) and least for Chinese (decrease by ~10%). Indonesian also has the highest percentage of restricted works at 11.35%. The number of restricted Chinese works has actually increased by 600, but since the total number of works increased by about 100,000, the percentage of restricted works fell.
I suspect the drastic variation between languages could be because the reddit post and ao3 announcement spread differently in these language communities. And obviously many other things might influence language communities differently, like political events in the countries the language is spoken in. We also speculated a bit on the reasons for differences between languages in the notes of the old post.
Year of Last Update:
The number of total works decreases in the later years, this will be mostly because people update their works and that pulls them into being updated in 2023. Interestingly, though, we can probably see the archiving efforts in the increase of total works in the years before 2010.
We can also see that even though older years had and have a higher percentage of restricted works, the more recent years experienced a larger increase of restricted works - likely because those users are more active on ao3 still.
A reddit post on ao3 being scraped as learning material for AIs recently made waves in fandom spaces. I've been tracking the number of restricted works on ao3 since December 2nd (about 24 hours after the post) until now, December 16th.
The total share of restricted works since Dec 2nd has increased by 70%. The plot below also includes some historical values thanks to the data points we gathered here with @destinationtoast.
We can also look at this as the daily increase/decrease in the total percentage. The slight bump on December 7th may be due to an increase of spam bots on ao3 at that time.
Unsurprisingly, rating correlates with the restricted percentage. Numbers in brackets behind categories are total works for that category.
I also looked into the current number of restricted works by fandom, both the largest fandom overall and the fandoms with highest number of new works in the last 7 days (excluding largest fandoms, thanks to @ao3-anonymous' Fandom Dashboard).
We can see the Men's Football RPF fandom is especially restricted. The Hockey RPF fandom is also known for having a high percentage of restricted works.
I saw quite a few Russian works when doing this, so I did a split by language, but Russian is actually not that restricted. A recent ban on 'gay propaganda' may be influencing Russian activity on ao3 right now.
Rating, fandom and language numbers are all from December 16th.
[Edit: Correction of inverted percentages in Hockey RPF diagram]
