Mobile phones famously created new challenges for those writing fiction set in the present day. Some authors chose to ignore their existence altogether. They changed how people lived their lives, but do those changes make for good stories?
Ultimately, yes. Or at least, mobile phones are not any less suitable for fiction than other aspects of real life; it has never been possible for a fully realistic account of anyone's daily life to make for a good novel. What was missing, predominantly, were conventions for how they should be integrated. I have no doubt literary conventions we now take for granted had to develop in the past to accommodate prior major ubiquitous sociotechnological changes, like widespread literacy.
I was nostalgically pondering some fiction from the earlier part of the mobile phone era, pre-touch screen, which uses them sparingly but well. A thought—not one I especially liked, instinctively and aesthetically—occurred to me: if LLMs (large language models, ChatGPT and its ilk) are approaching ubiquity, will they, like mobile phones before them, demand a change in how we write stories?
As of 2025—a long time ago, in these matters!—a Pew Research survey suggests that about two-thirds of teenagers report they use LLMs, of which three in ten report daily use. Now, I think there are very good reasons not to overrely on these tools, not to let them replace our own thinking or our own writing. And I personally retain a fondness for the "old-fashioned" search engine, even though Google thinks everyone ought to see a (cheap) LLM's output front and center instead of links. But that's besides the point, isn't it?
I think it might be useful by way of comparison to think about mobile phones and fiction a bit more. Here are some examples of things people often do with phones in real life:
- Communicate with distant people.
- Look information up.
- Engage with some sort of feed of content (social media, news aggregators, &c.).
- Make notes.
- Play mobile games.
My contention is: at any time when a character's actions are significant to the story and receiving active focus, if they interact with a mobile phone, their reason is overwhelmingly likely to be the first item in this list.
If the second item occurs, the focus will be on the information gained; the mechanics of evaluating and choosing between search results will be elided. The third item may serve as a means by which characters receive news, but in a manner no different from the tropes of passing a newsstand or turning on the TV. The last item... probably will not happen much.
In fiction, mobile phones are typically most useful to enable character interaction. And it can absolutely elevate a story for characters to be able to speak when they're not in the same place, although some kinds of plots demand settings where that is not possible.
What, then, are LLMs for, when it comes to end-user interaction?
Data here is self-reported, things change quickly, et cetera, but a report from Rethink Priorities suggests searching for information and seeking explanations of concepts are the most common LLM use cases among US adults. The success of character.ai in a younger demographic definitely suggests some are turning to them as sources of entertainment or even pseudo-social interaction. Some people, at least, use them for emotional support or therapy.
This is all interaction of a sort, but the LLM is not a character. (Outside of some experimental fiction, probably.) You cannot use an LLM to talk to another person, except maybe by using it for machine translation, but that's been around for a while. This is a very significant departure from how mobile phones affected fiction!
So, what of it? Well, I do not think people are interested in seeing a character talking to an LLM in most cases. Admittedly, I could imagine, for instance, a thriller focusing on the messages someone sent to ChatGPT in order to show the change in their state of information and emotion (it's not the AI responses that matter there, though). I think the use of LLMs in lieu of search engines will probably be elided as easily as the mechanics of searching already are, without even needing to distinguish between the different means of looking things up.
When we care about characters figuring things out, we are generally interested in how they figure them out, and would be disappointed by a language model doing it for them. At the same time, if a problem is trivially resolvable by LLMs, it might nag at people for a character not to try using one. Are many problems faced by fictional characters trivially resolvable by LLMs? I don't think that many, but I guess this creates some limitation on possible plots, the same way a setting with phones can't as easily rely on characters being unable to share information quickly over distances. (However, it might not open up many additional interesting plots, unlike the phones.)
I think using LLMs for entertainment is analogous here to playing mobile games and will not be portrayed much for the same reason that isn't. As for emotional support... setting aside whether that use case is advisable, catharsis and character development are, uh, kind of core to fiction, and probably should not involve AI therapy for some of the same reasons we prefer to read about characters seeking comfort in their own minds or from each other, not calming cat videos, and growing through adversity and challenge rather than by reading a self-help book. I can imagine this sort of LLM usage being portrayed as a maladaptive coping mechanism in some kinds of fiction, though, something for a character to grow from.
So, my conclusion: maybe if a TV show portrays a character in high school doing their homework for flavour, an LLM window is incidentally visible. And the fact of their increasing ubiquity may mean writers reference them incidentally. But by and large, the fact that LLMs are not characters and cannot enable character interactions will probably mean their effects on fiction will not be as significant as those of mobile phones.
Now, if they turn out to completely reshape society and automate most work, that might have effects on settings. Who really knows! More on that later maybe.