Games have used AI since the first enemy that knew how to chase you. That is not the interesting part anymore.
The interesting part is that the technology has climbed out of the game and into the hands of the people playing it. Coaching platforms that pick apart your last hundred matches. Chatbots used as strategy advisors. Bots that play for you, which is where things get messy.
Here is what is actually happening on each side of the screen.
What It Does Inside the Game
The traditional role has not gone anywhere. Non-player characters still need behavior, difficulty still needs balancing, and worlds still need populating.
What has changed is scale. No Man’s Sky generates billions of planets, each with its own ecosystems and terrain, because no studio could hand-build that. Procedural generation is the reason a small team can ship something enormous.
Developers also use it to tune difficulty on the fly, build events that respond to how you play, and simulate physics convincingly enough that movement stops feeling like animation.
Coaching Tools That Watch You Play
This is the most practical use for anyone trying to get better.
Mobalytics and Aim Lab both analyze your play and hand back specifics. Reaction times, accuracy, the decisions you made under pressure and what stronger players did in the same spot.
The value is not the data. It is that the data is comparative. Knowing your accuracy is 41 percent means nothing on its own. Knowing that players two ranks above you sit at 52 percent in the same situations tells you exactly what to practice.
Using a Chatbot as a Strategy Partner
This one surprised us, because it works better than it has any right to in games where thinking matters more than reflexes.
One player ran their Civilization VI campaign past ChatGPT, asking it to play out different diplomatic paths. What happens if I ally here instead of declaring war there. The model walked through the likely consequences of each, and the player made better decisions than they would have alone.
Tabletop players have gone further. Character backstories, branching dialogue, plot twists mid-session when a Dungeons & Dragons party does something nobody planned for. The game master who can improvise a coherent subplot in thirty seconds now has help.
Neither case is the machine playing for you. It is closer to having a patient friend who has read the manual.
eSports Has Gone All In
At the competitive level this stopped being optional a while ago.
Platforms like SenpAI and Gosu.ai chew through thousands of hours of footage from League of Legends, Dota 2, and Counter-Strike, looking for patterns a coach would need a season to spot. Team tendencies, individual weaknesses, the exact timings an opponent keeps repeating.
When matches are decided by fractions, a tool that finds one exploitable habit pays for itself.
And Then There Is Cheating
The same capability that powers a coaching tool powers an aimbot. That is the uncomfortable part.
Some players run bots that automate play outright. Others use tools that sit closer to the line, nudging aim or reading information the game did not intend to expose. The first gets you banned. The second is a genuine argument about where assistance ends and cheating begins.
The honest position is that the line is blurry and getting blurrier. A tool that reviews your replays afterward is clearly fine. A tool that whispers during a ranked match is clearly not. Plenty of what is shipping now sits between those two, and no platform has a convincing answer yet.
Moderation and Community
Automated moderation now handles a large share of the work in big online games, flagging abuse and detecting cheating at a volume no human team could match.
It is better than nothing and worse than advertised. False positives punish people who did nothing wrong, and the appeal process is usually another automated system. Anyone banned by mistake will tell you the technology is not ready to be the only judge.
On the stranger end, virtual influencers and synthetic streamers have started appearing, which raises questions the industry has not really engaged with.
The Questions Nobody Has Answered
Four worth sitting with:
- If a level was generated rather than designed, who gets the credit, and does the answer change when it is sold?
- Coaching platforms need your gameplay data to work. Where does it go, and who else sees it?
- Automated moderation makes mistakes. What does a fair appeal look like when the first reviewer is also a machine?
- Procedural tools do work that junior designers used to do. That is a career ladder with a rung missing.
None of these have settled answers. They are worth arguing about now rather than after the norms harden.
Where This Goes
The near-term direction is opponents that actually learn. Not difficulty sliders, but enemies that notice you always flank left and stop falling for it.
Storylines that respond to how you play rather than which dialogue option you picked. Combine any of it with VR and you get a game that reshapes itself around one specific player, which is either the most exciting idea in the medium or an expensive way to be lonely, depending on your mood.
What is certain is that the tools are already in players’ hands, and the rules for using them are being written well after the fact.
Have you used any of this in your own games? Tell us what worked. For more on where AI is showing up across other fields, check out our AI Blog.
About the Author: Gavin Stacey
