Bottom Line: Creative Assembly shipped a mediocre campaign and accidentally built the greatest strategy sandbox ever made. Buy it for Immortal Empires, tolerate the DLC economics, and expect to lose a year of your life.
The Gameplay Loop
The loop is two games wearing one coat, and the seam is the whole appeal.
The turn-based layer is settlement management, building chains, technology research, diplomacy, and army recruitment — competent, occasionally clever, rarely the reason you're here. Then two armies touch on the map and you drop into a real-time battle where a badly angled charge undoes fifteen turns of careful economy. That whiplash between abstraction and carnage is what the series has always sold, and Warhammer III sells it harder than any predecessor because the fantasy license lets the battles get genuinely weird. A unit of halberdiers holding a hill against a Bloodthirster is a real tactical problem, not a scripted setpiece.
The faction differentiation deserves specific credit. Historical Total War games struggled to make Rome feel different from Carthage. Here, switching from Kislev to Khorne isn't a reskin — it's a different resource economy, a different victory pressure, a different reason to declare war. Kislev manages two competing religious factions and an ice-court economy. Ogres need constant meat and take contract work as mercenaries. Slaanesh seduces enemy armies into defecting. That's design labor, and it shows.
Where It Still Falls Down
The campaign AI is the game's oldest unresolved debt. On the map it remains erratic — it declares wars it cannot prosecute, wanders full-strength armies in circles, stacks doomstacks without a coherent plan, and mostly relies on difficulty-scaled economic cheats rather than competence. On the battlefield it's better but still exploitable: bait a charge, flank with cavalry, watch a numerically superior force fold. Veterans don't lose to the AI. They set themselves handicaps so the AI has a chance.
Late-game turn times are the second chronic complaint, and Immortal Empires made them worse by definition. A map with ninety factions has to resolve ninety factions' decisions, and by turn 150 on a mid-range CPU you will be watching the end-turn animation long enough to check your phone. Patches have trimmed it. They haven't fixed it, and the architecture suggests they can't fully.
Sieges have improved substantially from launch — the minor settlement battles that shipped as beige, indistinguishable box mazes have been reworked and diversified — but they're still the weakest battle type in the game. Attacking a well-garrisoned city is more a matter of grinding chokepoints than executing a plan.
The DLC Problem
This is where I stop being generous. The trilogy's business model asks you to buy three games and a shelf of expansions to see the sandbox at full strength, and the pricing has periodically been indefensible. The Shadows of Change pack in 2023 shipped thin at a raised price and triggered a player revolt loud enough to force a public apology and a substantial free content top-up. Creative Assembly did course-correct — Thrones of Decay was the model of what a paid pack should be, three factions deep and broadly praised — but the pattern is established: value per pack is a coin flip, and the game's own store page is a maze for newcomers.
If you're arriving fresh in 2026, the honest advice is to buy the base game on discount, play a hundred hours of Immortal Empires with what you own, and add races only when you specifically want them. Do not attempt to complete the set. That's a four-figure hobby.
Interface and Onboarding
The UI is dense, information-heavy, and assumes you've played a Total War before. Tooltips nest inside tooltips. Critical campaign mechanics — corruption, supply lines, the specific rules of your faction's unique resource — are explained once, in a panel you will close and never find again. Onboarding friction is real, and the community's answer is a wiki tab and a mod that reformats the recruitment panel. That's a design failure being patched by volunteers.