An AI-scraped AO3 clone sold fanfic audiobooks and drew 2,000 takedowns

Archive of Our Own, the fan-run library where millions of people post and read fan fiction, has been offline since a security incident forced its operator to pull the plug. Into that gap stepped ao3.life, a copycat site built with AI agents that reposted the archive’s stories without asking their authors, then sold machine-narrated audiobooks of them. It lasted about a day before the people who wrote those stories buried it in removal requests.

The ending is almost comic. The mechanism behind it is not. One person with cheap AI tooling copied a community’s output, wrapped it in a business model that community forbids, and put the burden of saying no on the writers themselves, at the exact moment the one place where they could prove authorship was unreachable.

According to 404 Media, the site’s creator said the content was gathered by “a swarm of AI Agents controlling 40 virtual machines” scraping the internet for drives, scrapes and other copies of AO3 works. Readers could download stories as EPUB files or listen to AI-generated audiobooks, which cost money after the first three minutes. A page titled “Authors get paid here” promised writers 91% of the profits, with the remaining 9% covering “narration and processing.” By the fan wiki Fanlore’s count, cited by The Mary Sue, the site held 15,831,236 works before it went dark.

Kotaku reported that writers found old versions of their stories there, along with works they had deleted or made private. Some said AO3’s archive lock, the setting that hides a story from anyone who is not logged in, made no difference. To get paid or to have a story removed, ao3.life asked authors to email from the address tied to their AO3 account or to leave a note on their AO3 profile, which nobody could do while AO3 was down. Many users warned that the process looked like a way to harvest authors’ identities, and Kotaku told readers not to email the site at all.

Money is the part that alarmed fandom most. AO3 bars authors from monetizing their works or even linking to payment pages, and fan fiction has survived for decades in a legal gray zone partly because nobody sells it. As The Mary Sue argued, a stranger charging for audiobooks of fan works gives the companies that own the original characters a reason to pay attention to a space that has stayed safe by staying non-commercial.

The pressure worked faster than the lawyers. The Organization for Transformative Works, the nonprofit behind AO3, told users reporting the clone that it was taking action but asked for patience “because the legal system is much slower than online life,” according to a message relayed by moderators of the AO3 subreddit. Before those steps landed, the creator replaced the site with an apology. “In the time this site has been up, I’ve facilitated over 2,000 takedown requests,” they wrote, adding that the site drew 126,592 unique users in its first 24 hours and terabytes of bandwidth, and that the audiobooks were meant to pay the server bills. “I sold like 11 audiobooks, if it’s any consolation.”

The note is signed by a self-described 17-year-old from India. It says the data came from public scrapes and that its author has no link to AO3, the OTW or whoever breached the archive. An on-site survey, the note adds, split almost evenly: 3,184 visitors said the site made them happy and 3,057 said it did not.

Closing one site does not put the copies back in the box. The Mary Sue counted other archives built from AO3 material, among them fanfic.party, which claims more than 13.4 million works and describes itself as a temporary non-profit backup that will never train AI, and blk archive, which claims over 13 million works from 1.8 million authors. Each offers the same remedy: find your own stories and ask for them to be removed. That reverses the logic of consent and leaves writers policing a growing list of mirrors one email at a time. A takedown also reveals nothing about where scraped files travelled before a site closed.

AO3 itself is still down. The archive went offline for maintenance on October 29, after which the OTW disclosed a security incident. It says no works, bookmarks or kudos were deleted. It does not believe the user database was stolen but cannot rule that out; the database can hold usernames, email addresses, IP addresses and hashed passwords, so the OTW advises anyone who reused their AO3 password elsewhere to change it. AO3 hosts nearly 18.5 million works.

The OTW estimates that restoring the archive will take around two weeks and could take up to four, which points to a return somewhere between mid- and late November. It hopes to reopen sooner in a degraded state, possibly read-only, with more downtime likely while the repair continues. Until then, authors cannot switch on the archive lock that is meant to keep their stories away from scrapers, and every week offline gives the copies more time to spread than their writers have to find them.

Tags: artificial intelligence, web scraping, AO3, ao3.life, Archive of Our Own, fan fiction