Is *Left 4 Dead* about to get AI-generated dialogue? Valve is being pretty bold with this test.

Wow! Valve is even dipping into AI for game writing! The moment this news dropped, alarm bells went off in my head. Not out of fear. Out of excitement. The kind of excitement that says, “This could lead to something really new!”

Recently, Erik Wolpaw, the writer behind classics like Half-Life 2 and Portal, mentioned on a podcast that a small team inside Valve is exploring how generative AI could be used in game writing. Now pay attention, because this is the key point: this isn’t Valve going all-in company-wide, and it’s not a formal project. It’s more like a few people tinkering privately, doing some side research, testing the waters first to see how deep they are.

What makes it even more interesting is Wolpaw’s attitude toward AI. He’s neither blindly hyping it up nor dismissing it outright. He said directly that game writing is probably the “only field worth exploring with AI” in his eyes. But! He’s also not worried that AI is about to take creators’ jobs anytime soon, for one simple reason: right now, it’s still not good enough at writing creative content. He even once joked, “because it’s still pretty bad at it right now.” Honestly, that’s refreshingly real—like someone slamming the table in a livestream and telling it straight.

So why game writing in particular? That’s where it gets interesting. Art and music are often created first and then presented to players to see or hear. But game writing is different—it often has to “react on the spot.” If a player runs left, a character might need to say something. If a specific event is triggered, a teammate has to throw out a line. Left 4 Dead is a classic example: a lot of its dialogue is essentially interacting with player behavior.

So the direction Valve is probing this time is actually pretty clear: it’s not about having AI write an entire Half-Life 2 script, nor is it about replacing the dense, tightly designed writing of something like Portal. It’s about letting AI handle dynamic feedback, real-time responses, and content generated as you play. Put simply, it’s about pushing NPCs from being “fixed-line machines” toward feeling more like real people who can actually respond. That approach gets high marks from me.

Wolpaw also used chess programs as an analogy. No matter how strong the software gets, top human players don’t instantly become obsolete. In game industry terms, that means AI may become a tool, but we’re still a long way from top-tier creative writers being replaced.

My own take is that the first place AI writing will really take hold in games won’t be the main storyline, but side dialogue, environmental feedback, and repeatable variable content. It’s especially well suited for co-op, survival, and open-ended interactive genres as a testing ground. A game like Left 4 Dead, with its emphasis on changing situations and teammate reactions, feels like a natural AI playground.

The player community is definitely going to split into two camps next. One side will say, “This is bad—writers are going to lose their jobs.” The other will say, “As long as it doesn’t strip away the soul, making interactions more natural is a good thing.” I’m with the latter, but with one firm condition: AI can assist, but it can’t take over the core creative work.

One last thing to emphasize: Valve’s research isn’t tied to any specific game right now, and it’s still far from becoming an actual product. But that’s exactly the point—it may still be far off, but the direction is already lit up. As Wolpaw put it, “This technology is so crazy that it would be stupid not to at least look into it.”

Now we’ll just have to see how Valve plays its hand. If they really manage to train AI dialogue well, then the next time you play a game like Left 4 Dead, a teammate’s sarcastic quip might end up being even more entertaining than the chat spam in your livestream.