I mark my calendar for Steam Next Fest the way some people plan holidays. I brew too-strong coffee, stack my downloads, and disappear into a blur of wild ideas and scrappy prototypes. Two years ago I fell head over heels for a tiny horse-herding adventure during Next Fest and became, against all odds, a complete Horse Guy for a week. That’s the magic of this showcase: human-scale creativity finding its audience. This October? The magic kept tripping over the same uncanny seams. Not because there weren’t good games-there were-but because the event was flooded with generative AI in a way you could see, hear, and feel.
Let’s be blunt: roughly 17% of the demos at October’s Next Fest—507 out of about 2,960—disclosed using generative AI somewhere in the process. That’s not me spitballing; that’s what developers themselves reported. And if 507 told us, I’d bet my fight stick more did not. The disclosure list covered everything from marketing copy and store art to in-game visuals, audio, writing, translation, code, and even live calls to AI during gameplay. It’s not a trickle. It’s a rapid normalization.
I’m not anti-tool. I’ve been playing since Dreamcast-era experiments turned me into a Shenmue lifer, and I’ve spent enough time in the FGC to know a good training dummy when I see one. But what hit me this Next Fest wasn’t careful assistance. It was overreliance. When AI sluices into the heart of your art, audio, or writing without tight human authorship, the result is a fine mist of mediocrity that muddies discovery. You don’t just risk soulless assets—you drown out the thoughtful craft Next Fest is supposed to champion.
Here’s the core of it: tools should amplify human judgment. Instead, I saw too many demos where “AI” read like a shortcut for “good enough.” And at a festival built on tastemaking and first impressions, “good enough” is a slow poison.
I played until my wrist begged for ice. I’m not naming and shaming tiny teams who are probably already stressed—but the patterns were clear. The AI “tell” isn’t always a twelve-fingered hand. Sometimes it’s subtler, and it’s everywhere:
And yes, there were prominent demos on the AI-uses list. Cloudheim cracked the top-50 most played demos of the week, which tells you two things: players will try what looks interesting, and AI involvement alone doesn’t doom a game—strong design can carry. Cloudheim’s own disclosure leaned on “internal communication” as the primary AI use and then left the door open for team members to use the tech as they saw fit. That’s exactly the kind of squishy messaging that makes me twitch. “No people have been replaced by AI,” they added—cool sentiment—but if you don’t define boundaries, players are left to trust vibes after years of broken promises industry-wide. Trust is earned with specifics, not disclaimers.
Then there’s Embark. The studio caught flak previously for AI voice use in The Finals, and Arc Raiders’ page now mentions AI-assisted content creation. When I hear stiff or awkwardly paced VO in an extraction shooter, my alarm bells go off instantly. In competitive games I’ve clocked thousands of hours in (Street Fighter, Tekken, Halo), audio feedback has to be a razor: precise, memorable, motivating. AI voices tend to feel like dull butter knives. Maybe they’ll slice something eventually, but they don’t cut clean.
Worse, we’re seeing studios get called out after the fact. This year alone, games like The Alters and Hotel Barcelona ended up under the microscope when players spotted tells and dug in. If over five hundred Next Fest demos disclosed AI use, how many didn’t? And why are we still playing “gotcha” with disclosures at a festival that should be leaning into transparency as a competitive advantage?
Let’s sit with the numbers. 507 of roughly 2,960 demos flagged generative AI use—about 17%. That’s a chunk, not a quirk. And the disclosures weren’t confined to marketing polish. Developers reported using AI for in-game art, audio (music, SFX, VO), writing, translation, code, ideation, and even live calls at runtime. That spectrum matters. Marketing blurbs aren’t the same as voice talent or narrative tone. Lumping them together under a single “AI used: yes/no” checkbox is part of the problem.
And here’s where discoverability gets ugly. When Next Fest is inundated by demos that look passable at a glance because AI made it easy to conjure a slick image or passable pitch copy, the real standouts—the messy, risky, human-first experiments—get buried. GameDiscoverCo has hammered home for years that early impressions and store page signaling are make-or-break. A thousand AI-polished previews can crowd out the voice that doesn’t have a glossy model sheet because they’re busy building the actually interesting bit: how it plays.
I hear the counterarguments, and some are fair:
Here’s my pushback, from hundreds of hours living with games long after the demo ends. When AI moves from time-saving assistant to aesthetic decision-maker, we don’t just change the process—we flatten the result. The line between a rough draft and a shipped vibe is getting dangerously blurred. And when a studio tells me “no one was replaced” while the dialog sounds like a voice filter and the key art has that “dream with random elbows” feel, I’m sorry, I call BS. If a tool dictates the texture of your world more than your art director does, you replaced something—maybe not a job title, but definitely a layer of human intent.
And the “players don’t care” take? It’s lazy. Players care when they bounce off without knowing why. They care when a story doesn’t stick, when a sound cue doesn’t spark adrenaline, when a character looks like a mannequin draped in vibes. Human craft imprints in ways AI can’t yet replicate—micro-pauses in VO, painterly imperfections, the specificity of lines that could only come from a lived experience. Shenmue made me care about a cup of tea because all those choices felt human-sized and intentional. Strip that layer away, and you don’t just save time—you remove the reason to care.
The human impact is not hypothetical. This industry has been bleeding talent for two years. Layoffs are the chorus, not the verse. If indies normalize using scraped data models to patch over art, audio, or writing gaps, we accelerate a race to the bottom where fewer people are paid to do the thing we claim to celebrate. Studios like Aggro Crab have been vocal about protections for workers and the value of craft. That’s the energy indie spaces need to amplify, not the weasel words of “we used AI but don’t worry about it.”
Trust is currency. Mixed messaging and defensive disclosures gut it. I saw multiple Next Fest pages where the AI note read like legal boilerplate: “We use AI for a variety of tasks to increase efficiency.” Cool, thanks for saying nothing. Contrast that with a hypothetical disclosure that says: “We used generative AI to translate our store page into German and Japanese; all in-game VO, art, and narrative are human-made.” One earns trust; the other dodges it.
Next Fest is great precisely because it’s curated chaos. You’ve got one week to pitch yourself to people like me who will try thirty or forty demos and evangelize the best to friends and readers. But in a week where 500+ demos are AI-assisted, visual polish and marketing fluff are increasingly decoupled from actual design substance. That devalues the time of the player and the risk-taker who chose to spend their precious window honing a mechanic instead of prompting up a gorgeous splash screen.
And this isn’t a “get off my lawn” rant at algorithms. Cloudheim did well because the core loop caught fire for people. AI didn’t save it; design did. But AI-enabled surface-level gloss makes it harder to identify those design-first successes in the noise. We’re back to the Steam Greenlight vibes, but with prettier thumbnails—a thousand “maybe”s drowning out the handful of “hell yes.”
I don’t want to just moan. I want the festival I love to thrive. Here’s the playbook I used this time—and will refine next round—to surface the human-forward games and hold AI-heavy projects to a standard.
Valve did the bare minimum with disclosures. Now do the obvious next steps:
This isn’t about puritanical gatekeeping. It’s about meaningful transparency and letting the market reward the thing Next Fest was built to showcase: craft. If you bury that under a pile of AI flyers, you’re not neutral—you’re picking winners.
If you’re an indie reading this and fuming because you used AI to bash out a store translation at 3 a.m.—you’re not the problem. Your intent shows in your specificity. The problem is when AI becomes your aesthetic. When your pitch depends on a synthetic voice to carry emotion or when your key art sets expectations your pipeline can’t meet without the model doing the heavy lifting.
I’m asking for restraint and ownership. If you used AI, say where and why. Credit your actors, artists, translators, and writers loudly. Show a process shot. Highlight the weird human thing only your team could have made. Put your soul where the model can’t reach. The teams that do this will build communities that stick around past launch week.
I learned about meaning in games from Shenmue’s gentle insistence that the mundane matters and from a hundred late-night sets where a single frame of invulnerability felt like destiny. Those experiences were designed, tuned, and performed by people whose choices still ripple years later. AI is a fascinating assistive technology; it is not a substitute for that kind of imprint.
Next Fest is where I go to feel surprised by a new imprint. This year, too often, I felt the opposite: sameness, smoothed edges, vibes on tap. It made me mad because I know how much sweat goes into a demo. It made me sad because I saw smaller human-first gems struggle for oxygen in a feed clogged with pretty noise. And it made me determined to push back in my own small way: with my time, my wishlists, my coverage, and my wallet.
Here’s where I landed after a week in the trenches:
And I’m asking fellow players to do some version of the same. We don’t have to be purists. We do have to be principled. If we reward specificity and craft, we’ll get more of it. If we shrug and click on the shiniest AI-polished thumbnail, we’ll get an ocean of shrug back.
Cloudheim’s success inside the top 50 shows that a good game can survive this era, even if the team experiments with AI at the margins. That’s the coexistence scenario I want: human-directed games using tools to lift the boring stuff, not to flatten the soul. But that equilibrium won’t happen by accident. It’ll happen because platforms demand honest disclosures, because devs take pride in human craft, and because players like us refuse to settle for vibe paste.
Steam, do better. Developers, be braver and clearer. Players, be choosier. We can keep Next Fest a place where the next Horse Guy moment happens not because a model guessed what we like, but because a team made something only they could make—and it found the person who needed it.
Until then, I’ll be in the trenches, testing, filtering, and calling it like I see it. Over five hundred demos disclosed AI this round. That’s a warning shot, not an inevitability. Keep the human in indie. That’s the reason I’m here. That’s the reason any of this matters.
Get access to exclusive strategies, hidden tips, and pro-level insights that we don't share publicly.
Ultimate Gaming Strategy Guide + Weekly Pro Tips