Most fantasy castles fail because they are scaled for grandeur, not for war. Real castles were designed to let a garrison of about twenty hold off two thousand attackers for a year, so if your keep needs a small city of defenders to be plausible, it stops working as a castle and starts working as a set piece. The small, mean, awkwardly-placed hold on a defensible rock is almost always more menacing than the mile-wide fortress on the cover.
If you write fantasy long enough, you eventually notice that the castles in your genre keep getting bigger, and that the bigger they get, the less interesting they are on the page. There is nothing universal about that observation, and plenty of writers have got away with enormous fortresses by making the fantasy element pay for the impossibility, but as a default I think the reflex to scale castles up is one of the more quietly damaging habits in modern epic and progression fantasy. This is not a rule against ambitious architecture. It is a plea to run the numbers once before you draw the map, because the numbers are more interesting than most of us give them credit for.
Why does the size of a fantasy castle matter for the story?
A castle is a piece of strategic logic before it is a piece of scenery, and if the logic does not track, informed readers stop trusting the rest of your world. That sounds harsh, but it is only saying what every craft book says about magic systems and economies, applied to stone. Nobody would let you write a magic system where the cost changes chapter to chapter without addressing it, and nobody sensible would let you write a fantasy currency (I have a whole piece on money in fantasy that covers this) that reprices itself every time the plot needs it to. The castle is the same kind of object. It has a job, a garrison, a supply chain, and a specific geometry that lets a few defenders punish many attackers, and when it stops obeying those constraints the reader can feel it even if they cannot articulate why. My feeling, having beta-read a lot of manuscripts over the years, is that castles are actually one of the fastest tells for whether a writer has thought about the physical world at all. Get the castle right and you buy an enormous amount of trust for the rest of the setting.
What did real castles actually do?
Real castles were built to let a small garrison hold out against a much larger attacker, usually by making terrain and geometry do most of the work. That is the entire load-bearing sentence. A twelfth-century motte-and-bailey on a defensible rise might have a peacetime garrison of a dozen or two, and even the more famous stone castles of northern Europe rarely fielded more than a couple of hundred fighting men. Malbork in modern Poland, which is often cited as the largest castle in the world by land area, could push its garrison into the thousands, but Malbork is the exception people bring up precisely because it is the exception, and even Malbork is a monastic order's headquarters as much as it is a working fortress. The English castles most readers can picture, Windsor and Warwick and the like, ran on garrisons in the low hundreds when they were seriously threatened, and in the low tens the rest of the time. The point was never to house an army. The point was to make sure that whoever wanted to move an army through the region could not ignore the little pile of stone squatting on the ridge, because ignoring it meant your supply lines got harvested by the men inside.
That is the crunchy detail most fantasy writers miss. A castle you can ignore is not a castle, it is a manor. The whole reason a keep with twenty defenders is worth the stone and mortar is that the twenty defenders inside can sortie out at night, torch a foraging party, and vanish back behind the walls before dawn. So the twenty men in the keep tie down a besieging force of hundreds, and that math, twenty defenders against hundreds of attackers for a year of stockpiled grain, is where the castle earns its rent. You can (and should) invent freely from that base, but you should know what the base is.
Why do fantasy castles keep ending up gigantic?
Fantasy castles bloat because the author is confusing "impressive" with "large", when in medieval logic those two words are almost opposites. A really impressive castle in the real world is often small, mean, and viciously placed. Think of Château Gaillard perched over the Seine, or Edinburgh Castle on its volcanic plug, or Masada in the desert. None of those are especially big. All of them are terrifying to attack because the terrain does most of the killing for the defenders. The instinct to make the castle enormous almost always comes from wanting to signal the importance of the family that lives in it, which is a perfectly reasonable narrative impulse, but the signal you actually send by describing a mile-long curtain wall is not "these people are powerful", it is "the author has not considered how any of this works". I think this is partly a screen problem. Peter Jackson's Minas Tirith is beautiful, and it is scaled to a level that only makes sense if a god is holding the thing up, and a whole generation of writers absorbed that image as the default fantasy capital without also absorbing that Tolkien did the math on his end and then chose to write a city, not a castle. The distinction matters.
There is also a subtler version of this trap where the author gets the outer scale right and then quietly inflates everything else. The great hall seats a thousand, the barracks house a garrison of five hundred year-round, the wine cellar has been drawing on a hidden river of vintages for a hundred years, and by the time you have added them all up the castle secretly needs a small city inside it to function. The men-at-arms drilled in the outer bailey while a hundred cooks worked the kitchens and three hundred stable hands tended the mounts. That sentence sounds cinematic, but if you stop and ask what the annual grain bill looks like, you have just described a small town. Aim for either a castle or a city and commit to it.
How should you scale your castle to your story?
Start from how many soldiers can plausibly hold it for a season without starving, then subtract, and only then add just enough grandeur to serve the scene. That is the actual craft move, and it is worth being ruthless about. If your story needs a court scene with fifty nobles around a table, fine, but the fifty nobles are visitors, not residents, and the castle that hosts them does not need to house them permanently. If your story needs a siege that lasts a winter, decide how much grain and salt fish the cellars can hold, decide how many mouths that grain feeds, and let that number tell you the garrison. Robin Hobb does this well with Buckkeep across the Farseer books, where the castle is grand but the grandness is always in service of the political scene we are in, and the fabric of the place always feels like it could actually be swept and lit and provisioned. Joe Abercrombie is even better at it, because his forts and holds in the North read as small, cold, and functional, and every plot beat inside them respects the walls as walls.
A useful exercise here, if you have not tried it, is to name every space in your castle before you name the castle itself. Kitchen, well, chapel, armory, sally port, stables, hall, solar. When you know where each space is, and how it connects to the others through corridors and courtyards, the size problem tends to solve itself, because you will find that a castle of about the size of a large modern high school is enough for almost any scene fantasy asks it to hold. The keep does not need to be visible from three counties over.
What about the fantastic fantasy castle?
Big castles work when the fantasy element pays for the scale. That is the loophole, and it is a real one. If a dragon lives in the topmost tower and the tower is that tall because the dragon needs a landing pad, the reader will forgive the math because the math is the story. If a god's bones make up the foundation, or if a floating rune-anchor holds up the outer wall, or if the whole thing was carved out of a mountain by an earth mage in a single week seven hundred years ago, the size becomes evidence of the world's magic rather than evidence against the world's plausibility. My advice, then, is that if you want the impossible castle, earn it explicitly. Name the thing that makes it possible on page one of that setting, and then let the size do work in the plot, so that when the rune-anchor is threatened or the earth mage dies or the dragon wakes up, the castle is in danger in a way a normal castle could never be. You have paid for the scale in setup, so now the scale can pay you back in stakes.
The castles I remember from fantasy are almost always the ones the author has clearly thought about as physical objects. Guy Gavriel Kay's palaces in the Lions of Al-Rassan sequence, Steven Erikson's brutal keeps in Malazan, the Bloody-Nine's grim, awkward strongholds in the First Law, the fortified cities in the Codex Alera books. In every case the author has done the small work of making the space feel used before they let it be dramatic. And in every case a smaller, purpose-built castle would work as well or better than a sprawling one, because purpose is what makes a castle threatening. If you want a starting point for this style of thinking, I would honestly point you at the worldbuilding-comes-last piece I wrote earlier, because the reflex to build the castle before you know what the castle is defending against is the same reflex that produces overbuilt worlds in general. Figure out what happens inside the walls first, then draw the walls around it.
When you get this right, you buy a lot of credit with the reader. They will forgive a great many other worldbuilding decisions because they know from the castle that you have thought about how the world holds together. That is worth more than any single cool location, because it is the compounding kind of trust that keeps someone reading past chapter three. If you write web fiction, where the reader is always two clicks from bouncing, that trust is the whole game.
Common questions about writing fantasy castles
How big should a fantasy castle be?
Most historical castles held garrisons of twenty to a couple hundred, not thousands, because the point of the walls was to let a few defenders hold off many attackers for as long as possible. Scale your castle from that number outward, decide how long the garrison can be fed and watered from inside, and only then think about ceremonial halls and bedrooms. If the castle needs a thousand soldiers to feel plausible on the map, it is not really a castle in the medieval sense, it is a fortified city.
Where should I put my castle on the map?
Put it somewhere the terrain does most of the defensive work for you, which is what real castles did without exception. A crag, a river bend, a narrow mountain pass, a coastal cliff. If your castle sits in the middle of a flat plain because the map looked lonely there, you have to invent a mountain of magical or economic reasons for it, and readers who know a little history will feel the gap before they can name it.
Does the castle really need to be defensible if magic exists?
It usually still has to look defensible, because your reader parses the castle before they parse your magic. If a wizard can flatten any wall then no one would build walls at all, so the walls have to signal a rule about magic in the world, maybe that raw magic cannot cross worked stone, or that siege mages are rare enough to matter. Skipping that step is what makes fantasy castles feel like theme parks.
Can I write a fantasy castle without doing research?
You can, but only if you keep the camera close and never pull back to the exterior shot. Real trouble starts when a reader who has been to Carcassonne or Krak des Chevaliers gets a wide-angle description that violates basic siege geometry, because that is the moment they stop trusting your narrator. My advice, then, is to read the Wikipedia pages for three real castles, look at their floor plans, and then invent freely from that grounding.
What is the biggest mistake fantasy authors make with castles?
The most common mistake is treating the castle as scenery instead of as a machine with a purpose. A castle is a piece of strategic logic before it is a piece of atmosphere, and when you write it that way, describing what it is defending against and how it does the defending, the atmosphere shows up on its own. Big fantasy castles that read as menacing are almost always ones where the author has clearly done the math and then chosen to be grand within the constraints, not in spite of them.
by Jacob Tam · July 1, 2026
I run IlorisNovel, a platform for web fiction writers. If this kind of craft writing is your thing, the rest of the blog lives here.