As a human D&D Dungeon Master, AI will never replace me — but I wanted to see whether ChatGPT or Gemini were getting close

As an avid Dungeons & Dragons (D&D) enthusiast, the thought of AI replacing my role as a Dungeon Master is intriguing, yet not entirely feasible. The game's essence lies in the collaborative storytelling experience shared among friends around a table. However, the accessibility of AI models such as ChatGPT and Gemini for solo play has led many players to explore this option.

Using these AI tools, I tested their capabilities in creating a D&D adventure for a high-level wizard character, and observed their distinct approaches.

Both Gemini and ChatGPT generated a similar abandoned dwarven stronghold filled with magical traps as the starting point of the adventure. Yet, their unique styles diverged significantly. Gemini's narrative featured a more traditional dungeon crawl, with instances where I used suggested actions, eventually opting to think independently.

The climax involved teleporting away to defeat a Naga using a Cone of Cold spell, accurately matching the monster's hit points from the official rulebook. I asked Gemini about the randomness of its dice rolls, clarifying that it used a pseudorandom number generator and did not manipulate the results.

In contrast, ChatGPT produced a vastly different experience. The adventure centered around a mysterious dimensional prison beneath a mountain, with the AI presenting a more narrative-driven story rather than a standard dungeon crawl. Combat was almost non-existent, and the climax revolved around a perplexing mystery involving ominous portents spoken by the imprisoned guardians.

The AI detected my dissatisfaction and offered suggestions to improve the adventure, but the overall output still fell short of the immersive, collaborative experience one typically enjoys with friends around a tabletop.