Nostalgia Meets AI: Why Interactive Fiction Apps Are Having a Moment

Cover image by pch.vector via Magnific
Remember those choose-your-own-adventure books? The ones where you’d hold your finger on page 47 for the door, or flip to 112 if you were too scared to stick around? They sold like 250 million copies from the late 70s through the 2000s. That number is just wild. That format never really died. It just went dormant for a while, lying in wait until the technology was finally ready for it. And now, funnily enough, it's made a comeback. Only now, you don't just flip to page 112 and call it a day. It keeps going, because something on the other end is actually writing new pages for you.
That's the strange, slightly uncanny appeal driving a fresh wave of interactive fiction apps. They're not just digitized gamebooks. They're something closer to a living conversation, and a lot of people are finding that hard to put down.
The old format never actually left
Text-based games have always had a longer memory than anyone ever gave them credit for. Colossal Cave Adventure in '76, then Zork a couple years later. Both text-only, both banking entirely on your imagination to paint the picture. Infocom, the Zork studio, was actually topping the sales charts in the early '80s. With nothing but text. Meanwhile, graphical games were already out there, fighting for the same players.
So the appetite for narrative you actively shape, rather than just watch, has been there for decades. What's changed is the ceiling. A branching-path book — even a really ambitious one — has maybe a few dozen endings. Tops.
Why AI removed the wall
Large language models don't need a pre-written branch for every possible choice. They figure it out as it goes. Every next moment is built from whatever you just said. Ask a character something the writers never thought of, and there's no error message. You just get an in-character response — and the plot finds a new path.
A handful of things made this shift possible almost all at once:
- Inference got cheap enough to run in real time without lag that breaks immersion
- Fine-tuning let developers give models distinct personalities instead of one generic voice
- Context windows grew large enough to remember a conversation from ten minutes — or ten sessions — ago
- Mobile hardware finally caught up. No gaming PC required anymore
So now we've got apps like an AI chat with characters where you're basically talking to a character that listens — not just to your words, but to your tone, your history, your vibe. The 'book' is more like a conversation partner now.
Somebody talking to a detective character might notice the detective gets sharper and more suspicious if they've been evasive in earlier messages. That's not scripted. It emerges from the model tracking context, which is a genuinely new kind of storytelling mechanic — one no printed gamebook could ever pull off, no matter how many footnotes it had.
The data behind the hype
None of this would matter much if people weren't actually using it, but the numbers suggest they are, in a fairly big way. Character.AI, probably the most talked-about name in this space, reportedly handled tens of millions of monthly active users at its peak, with sessions that ran noticeably longer than typical social app usage. At one point, reports put active users’ average time on the platform at around two hours a day — closer to binge-watching a show than casually scrolling a feed.
A large share of the audience skews Gen Z. Similarweb says almost 60% are between 18 and 24. And they're not just checking in for a minute. Apparently, active users are logging well over an hour a day.
Fan fiction communities, the ones built around forums and Archive of Our Own, started overlapping with these apps instead of competing with them. Writers who spent years crafting alternate storylines for existing characters found a new outlet where they could actually talk to a version of that character instead of just writing about them. It didn't replace fan fiction. It became a companion habit, almost like sketching versus writing — different muscle, same underlying love of a story world.
This isn't a chatbot
You could easily throw this in the same bucket as customer-service bots or voice assistants, but that would miss the point entirely. A support chatbot is there to resolve a ticket ASAP. These narrative apps are engineered for the opposite goal — they want the conversation to last, to feel like it's going somewhere emotionally or dramatically.
A few structural choices separate the two:
- Persistent memory across sessions, so a character "remembers" what you told it last week
- Personality consistency, meaning the same character shouldn't suddenly talk like a different one
- Narrative pacing tools, where the system nudges toward tension or resolution rather than just answering questions
- User-created characters, letting people build and publish their own personas for others to interact with
It turns users into co-authors of an ecosystem rather than passive customers, which is basically the same energy that made fan fiction communities thrive in the first place, just running on different infrastructure.
A practical note for anyone curious
If any of this sounds appealing, it's worth treating it the way people treat any hobby that can eat time without announcing itself — set a rough boundary before diving in, not after. These systems are genuinely good at keeping a conversation going, and that's exactly the feature that can turn a fifteen-minute break into a two-hour one without anybody noticing the clock. There's nothing wrong with getting absorbed in a story; that's the whole point of storytelling, always has been. Just go in with eyes open about how the format is designed, and maybe check the time once in a while. The old gamebooks made you stop at "the end." This one doesn't come with that page.