Steam's Release Gate Is the Bottleneck: Valve Concedes Review Takes Up to 12 Business Days
Valve finally put numbers on a problem developers have been posting about for a week. In a Steamworks Development announcement published Tuesday night, the company admitted submission volume has climbed sharply and that reviews now take "up to 12 business days" instead of the 3 to 5 it normally aims for. Its explanation blames rework inside its own pipeline, and its fix is a stricter release checklist that moves labor onto developers. The argument about why the queue is jammed is still doing the rounds; the argument about what Valve actually said is over, because Valve said it in writing.
What Valve actually published
The post is worth reading instead of inferring, because it is specific in places the coverage isn't. Valve says it has seen "a significant increase in submissions for store page and build review in recent months," has spent "the last several weeks" trying to scale its process, and is candid about the current state: "we're running a little behind right now and taking up to 12 business days."
The new required checklist items are not vague guidance. A developer must now launch the game through Steam's own Play button and accumulate at least five minutes of playtime before submitting for review; playtime from any user counts, and it has to fall within the past seven days or since the last build upload. You must declare which common redistributables Steam should install — DirectX, Visual C++, .NET, XNA — or explicitly state that the game needs none. Achievement configuration gets checked for missing names, missing icons and duplicate names. If Steam Cloud is listed as a supported feature, someone in your partner group must actually write a save file to the cloud, and Valve will flag you if cloud sync is still sitting in developer mode. Store pages now need at least five screenshots tagged as gameplay, with new categories for menus, concept art and in-game cinematics. Submission-time reminders are dynamic: promise multiple languages and the checklist asks whether the build contains them; ship multiplayer and there's a field for reviewer credentials.
The line that tells you how the queue really works is Valve's own: "Believe it or not, sometimes we review builds that have not ever been played through Steam." A game that starts fine from a raw executable can still fail for a customer who installs it through the platform — wrong launch options, missing content depots. That is an integration failure, not a content problem, and Valve is now forcing a smoke test before it spends a reviewer's hour.
The queue math nobody published
Valve's diagnosis puts the cost in the cycle rather than the line: "Much of that delay is due to having to re-review games, which isn't great for you or for us." That is how queues misbehave. A reviewer can reject a build in minutes, but the fix-and-resubmit loop costs days on the developer's side and burns a second full slot on Valve's side. Once a meaningful share of arrivals are returning items rather than new ones, effective workload grows faster than raw submission counts suggest.
Here is the part that should annoy anyone who wants a serious answer: Valve has not published the submission volume, the re-review share, or a median review latency. It describes the increase as "significant" with no figure attached. Every number in circulation either comes from Valve's own FAQ — 3 to 5 days normally, up to 12 now — or from an unhappy developer's screenshot. Store page review already carried guidance to submit at least seven business days before a planned go-live. Stack a slower build queue on top of that and a launch date needs weeks of slack where it used to need days.
The developers eating the delay
The Verge reported that indie games are missing planned launch dates because the queue is backed up, naming the developers of Food Devils, Wheel Make It and Cubic Earth as having waited up to 20 days without a response. The Cubic Earth developer told PC Gamer that a trailer pulled more than 60,000 views on YouTube and picked up more than 6,000 wishlists, and that the Steam community page is now "rapidly filling with player complaints" while momentum drains away. Wheel Make It reportedly failed approval because Valve's reviewers couldn't work out how to enable its local co-op — a feature that consists of two players sharing one keyboard. Kotaku's reporting describes developers waiting two to three weeks with no communication, and some rejections costing another month to explain or re-submit.
None of those failures are hard engineering problems. A shared-keyboard co-op mode is a store-page wording problem a reviewer should resolve in one message, not a rejection that costs a launch window. That is what a queue does when it is optimized for throughput and not for the person waiting in it.
Where the sources disagree
Valve did not name generative AI in the announcement. It blamed volume and re-reviews. Outlets covering it are more willing to point: Game Developer framed the story as a "flood of new games" and noted that Steam NextFest recently doubled historical highs for submissions, while linking the pressure to the AI-driven clone flood documented on itch.io. Kotaku's piece led with developers blaming the AI-slop surge. Engadget cited prior reporting that roughly a fifth of Steam Next Fest demos carried a generative-AI disclosure. Those outlets agree the flood is real; they differ on how much of the delay it causes, and Valve's own text won't carry that claim for them.
There is one quiet hook in the checklist that supports the AI reading without Valve ever saying it: Valve's review tips state that testers compare the build against the content survey, "such as mature content and AI usage." Disclosure mismatches become rejections, and rejections feed the very re-review loop Valve is blaming for the wait.
What I Think
The five-minute playtime rule is the right fix attached to the wrong public story. Valve's failures are operational — broken installs, missing depots, half-filled metadata — and forcing a smoke test before review is cheap, fast and genuinely reduces wasted reviewer time. Making developers tag screenshots and declare redistributables is housekeeping that should have existed years ago. Fine. All of it shifts labour onto the smallest teams, which is the trade I'd like more credit for being honest about.
What I'm not fine with is the opacity. Twelve business days is a tax, and it falls hardest on a two-person studio that has no release manager, no buffer and a wishlist curve that decays the day a date slips. Any operation that runs a queue publishes its latency; that's how you know whether a backlog is being worked or merely endured. Valve has told us the maximum and nothing else — no median, no re-review rate, no submission counts. Publish the queue stats, Valve. You're the only party holding them, and until they're on the table, every argument about why the gate is slow is a fight over vibes.