I'm currently hosting a campaign with my friends right now. A while ago I was kind of stuck with writing the plot so I decided to ask ChatGPT for some help since I've seen my friends use it for schoolwork and other stuff. After that I kind of got used to using it in my campaign. Not completely making a campaign, just small things like how I can make something better or coming up with encounters that would fit my campaign. I mentioned this to a few other of my friends who are also DMs and they said I should not do that since "It doesn't teach you to be a good DM" or something in that vein. I was unsure if they were just overreacting or if it really is a bad thing so I decided to ask random people on the internet. What do you think about this question and why? Thanks in advance to anyone who helps me out with this!

It's one tool among many. AI is great for stuff like coming up with names and descriptions for things. Using it to come up with encounters for your campaign is slightly more challenging, though, as it won't be able to properly adapt things to your party. Of course, you can still use it to come up with ideas and then tailor those ideas yourself to better fit the game.

Personally? I think you'd be much better off running your ideas past us. Contrary to the beliefs of some, AI has things it's good at, but this isn't really one of them. AI is good for giving you an answers but since it doesn't know why it's the answer (we'll set aside the issue it has with accuracy for now), it can't reliably explain why it's better, or why you should do it. It also has trouble with long term consistency - it can come up with random encounters or themed ones for you (although I'm not sure I'd trust the CR of it), but encounters should flow from the story, and AI can't handle that at the moment.

So, the problems with AI is that it can't teach you how to be better, at best it might give you good suggestions,but you can't ask it why. It also won't help you with the higher level skills.

Alternatively, you can come here and bounce ideas off of us. We can explain ourselves and why our suggestions are what they are. It also helps us by making us think about our habits...and what helps and what doesn't in our own games.

As someone with ADHD (and am thus a busted fire hydrant of ideas) I find the increasing use of AI to be frankly horrifying; it's basically just assembling the most generic material possible based on the questions put before it and while this can be good in a pinch it undermines your creative vision and individuality.

I'd honestly rather look at some crudely drawn stick figures or try and parse someone's FNAF inspired world building then work with AI if only because it's something that is *real*.

Fair warning, any question of AI in D&D will attract opinions like moths to a flame, and I am one such moth.

I have three core objections to the use of AI in D&D, two ethical and one practical. Let's get the less important practical objections out the way—AI is bad at producing quality results. At the very best it can regurgitate recently ingested data in a sterile and processed fashion, like a factory turning chicken remnants into over salted nuggets via the pink goo process. At worst it will spout out stuff that is objectively (or even harmfully) incorrect. If you rely on AI for your D&D work, everything will take on a weirdly 'off' quality. For example, I recently watched a movie that I'm incredibly certain it was written by an AI. The pacing was bizarre, the plot was technically coherent but disconcertingly disjointed, and overall it only resembled a story from a distance if you squint. That's basically what AI produces, an illusion of the thing you're looking for that looks right if you're not staring directly at it.

Now for the two ethical objections I have for AI as a D&D tool—AI is fueled by two things; inefficient use of power and plagiarism. That's not hyperbole in either count, the computing power to process each prompt is through the roof and the server farms being used to handle those prompts burn through power and water for cooling. They're not even trying to make them renewable at the moment, the closest we've got is microsoft pitching restarting a nuclear power plant to run it's LLM server farm, which I guess is better than nothing. Then there's the plagiarism—the company behind ChatGPT has literally admitted that they plagiarize material to ingest for their training data. There have been leaked lists of artists and creators to prioritize training from as so to produce their specific style—not be inspired by, straight up copy their specific artistic style. Until these two issues are solved—models are being run on sustainable power and are exclusively trained on consensually provided and compensated training data—AI tools are not something I could personally use in good conscious for D&D in any capacity.

Well, that's my two copper pieces. May Bahamut have mercy on your soul for the onslaught that is coming your way.

"Better" (IMO, anyway) ways to come up with encounters for your campaign than AI:

- keep a dream journal. I've had not just encounters but entire side quests sparked by things that came from a dream
- player input. Is there some monster someone casually mentioned during a session that they'd like to fight, or had never run into before? Is there some feature one of the characters hasn't really had a chance to use yet, that you could tailor an encounter around?
- flip to a random page in the Monster Manual, or click on a random monster name on DDB. Seriously -- seeing some interesting creature, or noting some interesting attack or way of moving around, might give you an idea for using them in an encounter -- or a Google search to see if other people have built interesting encounters around them. This has worked for me more than once

AI has some uses for DMing, but not many. I've used it to get feedback on an adventure, with reasonable success. I also really suck at coming up with names for things, so I've used ChatGPT twice for that, but I prefer not to. Just never use it to get ideas for adventurers.

Generally speaking, I would say AI works best for D&D to augment specific skills the DM is lacking or does not feel are a good use of time.  For example, a DM who can imagine what things (NPCs, monsters, locations, items, etc.) looklike, but lacks artistic skill or time to bring that  visual into reality might want to use an AI art generator to supplement their campaign with visuals that match those in the DM’s head.  Or a DM might want to use AI for tedious tasks that otherwise bog down time, things like “create me a fantasy tavern menu” or “give me a list of a dozen fantasy NPC names.”

Those kinds of tasks tend to be things that vibe well with AI’s skill set - they are simple and do not require much actual “thought” beyond the inputs of the prompt.

Encounter design is a much harder concept - good encounter design takes into account individual character skills, player preferences and habits, the lore and nature of the area, the greater plot, the campaign’s social contract of how deadly it should be, etc.  Those are all factors that really require a human touch - AI is just not at the point, and may never really be at the point, where it can take the sometimes conflicting preferences and emotional reactions of four or more people into account simultaneously.  

Moreover, as the game progresses and players get more and more options to handle abilities, combat becomes more about non-damage ways to challenge your players, with actions designed for their specific skills. That is an area AI is going to struggle with, and might be an area the DM will struggle with if they have not built up a foundation of knowledge on what does or does not work when designing encounters for their specific playgroup. 

I'd point out for people who have trouble coming up with things like names that random name generators have existed since well before generative AI was a thing.

Don't use ChatGPT or AI tools for D&D. Don't use them for anything. They're notoriously unreliable, rely heavily on the stolen works of real artists and writers, and do terrible damage to the environment. 

Just out of interest, do people who oppose the use of AI on environmental grounds also oppose video streaming for the same reasons?

OP, the DnD Beyond forums are always going to have posters strongly against any mention of using AI/chatbots/etc... The commenters are very strong in their opinions and make some valid points. For a different perspective, here is a recent video from the DM Lair Youtuber (I'm no fanelf of his btw, I much prefer generally DnD-positive Youtubers like Dungeon Dudes and Ginny Di) that might be of interest to you as you make your own decisions about how best to enjoy the game.

https://www.youtube.com/watch?v=kkyH7Ro3roE

Don’t do it.You will be a worse DM for it. 

If you are a beginner DM then I would agree with Eric and others that AI is not a good a tool to use, especially to help you become a better DM. On the other hand one of the other DMs in my group is also a long term and expert DM as well as a very proficient programmer and does understand AI’s limitations and abilities. He uses it in his current campaign as an aide and tool not the primary source and I haven’t seen any fall off in his game. AI has a place in the modern world but unless used by experts it’s not ideal for DnD.

Quote from Frachese >>

Just out of interest, do people who oppose the use of AI on environmental grounds also oppose video streaming for the same reasons?

 Not even comparable to the point that I have to question the authenticity of this comment. Video streaming uses less power than traditional TV with estimates of 0.1 kWh per gigabit and netflix estimating 0.077 kW.

Compare this to ChatGPT which is estimated (from the company themselves) to as much as 0.01 kWh per query. This means asking ChatGPT just 10 queries equals an hour of HD streaming. Average users are making 20-30 queries per hour, while API calls are much higher rates.

For some direct figures, in 2019 Netflix used 451 gWh of power, whereas in 2024 ChatGPT used 14,460 gWh, 32 times as much. The point is that AI consumes orders of magnitude more power than streaming, and that's not even factoring that per user (3 people watching netflix consumes the same power as 1) or the fact that the power values for streaming represent the high end whereas for ChatGPT it's the low end (text queries, not image or video)

My advice is not to use AI as a crutch, use it as a tool and know when to put it down. If you write major plotlines with AI, your story will feel more generic and standard. And to be clear, that doesn't mean it will necessarily be bad- I've been in some cringeworthy campaigns without AI that probably wouldn't have been any worse sticking closer to the standard tropes given that their worldbuilding was just painful. But it will be less of you, and that defeats the point in my opinion.

Relying on AI for anything is going to hurt your skills as a writer, storyteller, and participant in the D&D experience. If you get to the point where you're in the middle of a session and you're unable to answer a question without going to ChatGPT you're going to be a worse DM (though to be fair that's the same if you rely on Google forever, or opinions on reddit, or whatever- but AI makes it easy because there's nothing too "niche" for AI to hallucinate a halfway-plausible wrong answer to). You should build your skills independent of AI, and try to make sure you at least *can* run a session without running to AI for help.

That said, I don't think AI is bad a priori. Obviously there are going to be problems with AI training and resource consumption and all that, but in moderation I think you can probably find a middle ground where you use *some* AI to help with your weaknesses. I suck at coming up with fantasy names, and a lot of name generators feel very generic or pull from existing names that might not fit a setting. You could use ChatGPT to make a list of names following specific rules that fit your setting without having to rack your brain on how things work. If you need to make a picture for a NPC, you might use an AI to at least draft something on the fly to immerse your players more (though I think it's generally more useful to get actual art from an artist if a NPC will be important, I don't think Random Commoner 493 needs that level of attention). I have aphantasia, so when a DM describes something to me without a picture reference I see absolutely nothing. Judge me for it if you want, but sometimes I slap those descriptions into AI and it makes games feel a lot more alive to me. It helps me as a DM too, making sure that things make good physical sense with regards to scale and getting a sense for how to describe a NPC.

That's obviously related to a specific medical condition, so your mileage may vary, but really that's the point of AI- use it to help where you struggle, but don't rely on it to replace your story telling. When I make a NPC portrait with AI, I feed the AI a description, but then I tell the players the significance of an amulet the character is wearing or describe how the magical aura around an object shimmers with force using the AI as a springboard for storytelling. I never found myself able to do that well before AI, because other than just referencing existing art often pulled from a source where those interpretations were already made for me, my imagination for visual things comes down to artificially constructing language around a scene that I myself cannot see.

What I'm trying to say is don't let AI do storytelling for you, use AI to help you tell better stories. Be critical about how you use AI, and do it in a way that respects your players by not shoveling slop into their face and in a way that respects you by giving you room and authority to tell the story you want to tell, not the best calculated response to a prompt you wrote and then handed off to a number crunching model. I don't think it's fair to give a hard no, especially since there are a lot of minor things that AI does well that a DM would have to practice for years to do, but when you find yourself handing off those tasks to AI analyze what it does well (and importantly what it does not) so you can hopefully improve as a DM to the point where you aren't reliant on it. You never know when you'll find yourself sitting at a table with friends and the internet goes out in the middle of your session, or just using those improvisational skills in conversation. Use AI to find where you can grow as a DM, and take meaningful steps to try to do so when you can. And when you're done, do what you can to support the storytellers, artists, and communities that make great works of art and fiction so that human creativity can thrive. 

Quote from Davyd >>

Quote from Frachese >>

Just out of interest, do people who oppose the use of AI on environmental grounds also oppose video streaming for the same reasons?

 Not even comparable to the point that I have to question the authenticity of this comment. Video streaming uses less power than traditional TV with estimates of 0.1 kWh per gigabit and netflix estimating 0.077 kW.

Compare this to ChatGPT which is estimated (from the company themselves) to as much as 0.01 kWh per query. This means asking ChatGPT just 10 queries equals an hour of HD streaming. Average users are making 20-30 queries per hour, while API calls are much higher rates.

For some direct figures, in 2019 Netflix used 451 gWh of power, whereas in 2024 ChatGPT used 14,460 gWh, 32 times as much. The point is that AI consumes orders of magnitude more power than streaming, and that's not even factoring that per user (3 people watching netflix consumes the same power as 1) or the fact that the power values for streaming represent the high end whereas for ChatGPT it's the low end (text queries, not image or video)

That 451 gWh figure apparently was just the energy consumption of their servers, while the vast majority of the total consumption comes from data transfer. It's difficult to find exact figures, but the number you quoted (and was provided by the company itself) is definitely too low.

Regardless of what the numbers actually are, responding to someone that says “Thing A is bad” by saying “Unrelated Thing B is just as bad” is a very silly argument.