Steam Store Page Checklist

A Steam store page will not pass review until nine image assets, five screenshots, both descriptions and the content survey are in place. Everything else -- trailer, tags, system requirements, languages, Steam Deck status -- is technically optional to Valve and effectively mandatory to players. The checklist above walks the whole set in the order most developers build it.

Two things trip up first-time submissions more than anything else. Capsule art is uploaded at double the size it displays at, so a header capsule that looks fine at 920 x 430 in Photoshop can be unreadable at the 460 x 215 it actually renders at, and worse at the 120 x 45 the small capsule shrinks to in some widgets. And Steam rejects screenshots that are not real gameplay: concept art, pre-rendered cinematics, and marketing mockups all come back. Each rejection cycle costs three to five business days, so it is worth being pedantic before you submit rather than after.

For the full spec on every asset, with the notes on what Valve actually checks, see our Steam page asset requirements reference. For the design side of capsules and screenshots, the capsule design guide and screenshot guide go deeper.

Checklists tell you the floor. Our own corpus tells you the distribution. Between 10 February and 30 July 2026 we analysed 405 Steam store pages, and the raw counts -- screenshots, videos, and the two description lengths -- are the part of that data that holds up cleanly. The corpus is majority pre-launch: 225 coming-soon pages against 180 released ones. That skew matters for every number below, which is why the tool matches you to a cohort before it compares anything.

The pre-launch trailer gap

Released pages average 2.39 videos; coming-soon pages average 1.20. The difference is +1.20 videos, with a 95% confidence interval of +0.86 to +1.57, so it is not noise. 81.8% of coming-soon pages have one video or none, against 49.4% of released pages. Pages accumulate video after launch -- update trailers, gameplay cuts, accolades -- which is exactly the wrong order if the point of a coming-soon page is to collect wishlists.

A quarter of pages sit at Steam's screenshot floor

24.7% of all pages carry only 5-6 screenshots, and 14.6% carry exactly 5 -- the minimum Steam accepts. Coming-soon pages do it more: 31.6% versus 16.1% of released pages, a gap of +15.4 percentage points (CI +7.2 to +23.4). Hitting the minimum is a decision, and it is worth making it on purpose rather than because five was what you had ready the day you submitted.

Descriptions: one distribution, one myth

The median page runs a 4,404-character detailed description, with the 10th percentile at 1,850 characters and the 90th at 7,801. Short descriptions average 224.5 of the 300 characters Steam allows. That last number gets misread constantly, so to be explicit: in our data, short-description length does not predict a better page. Pages using 250+ characters are indistinguishable from pages using fewer, and the null replicates across two independent halves of the corpus. Use the space if you have something to say. Do not pad to 300 because a checklist told you to.

A percentile here answers one question: how many of the pages we analysed carry fewer than your number. That is all. It is not a score, and it is not a ranking factor. Steam does not rank, sort, or surface your page based on how many screenshots you uploaded, how many videos sit in your gallery, or how long your About This Game section runs. Discovery on Steam runs on tags, wishlists, sales velocity and player behaviour, not on asset counts.

What a percentile is useful for is noticing where you are unusual. Sitting at the bottom of your cohort on videos while your competitors ship three is a fact worth knowing before launch week. Sitting at the top on screenshots is not a win by itself. Neither number is a target to chase; both are context for a decision you make about your own page.

We only benchmark the four raw facts Steam itself records, because those are the numbers our instrument cannot distort. We deliberately do not publish our own rubric scores as a benchmark: the analyzer changed three times during the collection window, so a score from February and a score from July are not the same measurement. Every percentile you see carries the cohort it came from and the number of pages behind it.