Can an AI actually run D&D 5e?

Yes, and the useful answer is a specific one. On Rolegend, code enforces most of D&D 5e: dice and modifiers, hit points, conditions, spell slots, concentration, rests, inventory and leveling. Movement, cover and the other rules of position are handled on the Virtual Tabletop, and the Game Master narrates whatever is left.

The rest of this is what those three numbers mean, because the split between them is not arbitrary. There is a clean line running through fifth edition, and the line is the battle map.

What is solved

The bookkeeping half, which is the half most people assume is the hard part.

Dice, modifiers and advantage. Hit points and damage. Conditions. Spell slots. Short and long rests. Inventory and carried weight. Experience and leveling. On Rolegend these belong to Keystone Engine, a game engine built apart from the AI. It handles the dice and the game state, and it is the only thing on the platform permitted to change a number. It settles every number before the narration starts. When your character acts, you roll your own die, and the engine adds the modifiers from your sheet. ROWAN, our narrator, is handed the result and writes what happened, and it cannot ask for a different one.

Conditions are the example worth dwelling on, because they are the case where "the AI knows the rules" and "the software enforces the rules" come apart most visibly. A condition on Rolegend is a stored row, and the dice code reads it: saves that should auto-fail do auto-fail, an attacker who should have advantage gets it, a character who should be rolling at disadvantage rolls at disadvantage, and one who cannot act cannot act. Nobody has to remember. Nothing depends on the condition being mentioned in the last few hundred words of narration.

That is the general shape of the solved half. It is arithmetic over stored state: a database problem wearing a rulebook. In a plain chat window it goes wrong because nothing there stores anything, so a value lasts only as long as somebody keeps repeating it.

What the Virtual Tabletop adds

A chat window is worst at positioning, and that is the job of the battle map. On Rolegend's Virtual Tabletop the geometry is enforced, with a token for every hero and every enemy.

Movement and distance. On a battle map the budget is real: every square is charged, difficult terrain costs double, Dash doubles it, and a move you cannot afford is refused rather than narrated.

Cover. On a map, cover is traced from the walls, corner to corner: half cover adds 2 to the AC that the attack is checked against, three-quarters adds 5, and total cover refuses the attack outright.

Opportunity attacks. On a map, any move reports exactly which enemies it provokes, worked out from reach and the path taken, gated on sight and the reaction and aware of Disengage. Your character's reaction stays your choice: the Game Master asks before you swing.

Flanking. On a map the engine grants the advantage itself, as the Dungeon Master's Guide variant: a melee attack lands with advantage while a living ally stands on the opposite side or corner of the enemy's space.

Grapple and shove. The contest is an ordinary roll, and what the map adds is the part that matters: a grappled creature cannot move at all, and a shove resolves as a real push, clipped by walls, with the exact fall derived over a ledge.

Encumbrance. Carried weight against Strength is enforced. Encumbered costs 10 feet of speed, heavily encumbered costs 20 feet and imposes disadvantage, and the battle map reads the reduced speed.

Light, vision and fog of war

On a battle map, a square your character has not seen stays hidden, and fog of war follows what your character can really see. Torchlight, darkness and darkvision change what each creature can see, and how well it fights.

Mounted and underwater combat are enforced too: a mount moves with its rider, and a weapon attack in water takes the disadvantage the rules give it.

Why positioning was the hard half

Each of these rules comes down to where things stand on the grid, and that is hard to track in text.

A grid answers "how far" by subtraction. Prose has no coordinates. Write "I back away from the ogre and shoot" and you have described an action perfectly well without ever fixing a distance, because no reader needed one. But an opportunity attack is entirely a question about a distance: did you leave its reach? To enforce that, the software has to know where you were, as a number, before the sentence exists. To write the sentence, it only has to be plausible.

Cover is worse, because it needs three positions rather than one: the attacker, the target, and the thing in between, plus whether the line from the first to the second passes through the third. None of this can be derived from a paragraph, because the paragraph was never carrying that information.

The tempting move is to ask a language model to hold the battlefield anyway. It will happily do it. That is precisely the problem: it always answers. It will tell you the ogre is twenty feet away, fluently, and thirty feet away four turns later, with no event in between that moved anybody. A rule enforced inconsistently is worse for a player than a rule that is openly not enforced, because you cannot plan around it, and planning around it is the entire point of tactical combat.

Which is why the fix was not a smarter model but an actual map. On Rolegend's Virtual Tabletop, positions are coordinates in encounter state, the geometry rules read them, and the map shows you how far your token can really go.

What a human Game Master still brings

Everything above is solvable, and the battle map is the proof: drawing one moved movement, cover, range and opportunity attacks a column to the left together, because they were always one problem.

One thing works differently.

A good human Game Master does something no coverage table can measure: they plant something in your second session that pays off in your twentieth. They notice that you keep mentioning your character's brother and quietly build an arc out of it. They throw away the dungeon they spent a week on because the party got interested in the innkeeper instead. The surprise in that is not improvisation; a language model improvises fluently. It is intention held across months, by someone who wants the payoff to land on you specifically.

Structured state gets you the mechanical version of this. Rolegend stores plot flags, locations, NPC relationships and a rolling recap, so the world stays consistent and things you did forty turns ago are still true. Consistency is not the same as intention. A world that never forgets is a real achievement and it is not the same thing as a storyteller who has been setting you up for a month.

So an AI Game Master is not a lesser version of a great table. It is a different thing: available at two in the morning, it never cancels, it runs a fight by the actual rules without arguing, and your friends can join it.

Why we publish the whole table

It is not normal, in this category, to publish what the engine does with every rule. An AI Game Master demos extremely well for ten minutes, and the incentive is to let the demo be the answer.

Hour forty is where these products are decided, because that is where a number held in a conversation starts to drift and a number held in a game engine does not.

So, can an AI run it?

Yes. It can handle the dice and keep every number, apply every condition, track your slots and your rests and your pack, run a fight on a battle map with fog of war, seat your friends at the same table, and play a coherent world at any hour you like.