Steam Store Pages in Numbers: September 2026 Edition

This is the first edition of a report we publish after each Steam Next Fest, and it counts what Steam store pages contain rather than what anyone thinks they should contain. It draws on four samples: 405 pages benchmarked February 10 to July 30, 2026; 220 full analyzer reports written February 10 to September 14, 2026; 764 live community tag clouds pulled on September 14, 2026; and 286 unreleased pages captured to August 18, 2026.

Each figure below is one sentence carrying its own sample size and capture date, because the four samples do not overlap cleanly and quoting a number without its n would be useless to you. None of them is a rubric score. Our stored subscores moved three times in 2026 for reasons that had nothing to do with Steam pages, which the rubric bug writeup documents, so this report stays on raw store facts and on text we can count literally.

The percentile corpus behind every benchmark we show is 405 Steam store pages analyzed between February 10 and July 30, 2026, split 180 released and 225 coming soon. It is the smallest of the four samples and the only one with a fixed, frozen window, which is what makes it usable as a baseline.

The median page in that corpus carries 9 screenshots: 10 for the 180 released pages, 8 for the 225 unreleased ones. The median trailer count is 1 overall, and that single number hides the widest gap in the corpus: released pages sit at a median of 2 videos, unreleased pages at 1.

Description length barely moves across release status, which surprised me more than the trailer gap did.

The half character in that last cell is real: these medians interpolate between the two middle pages rather than picking one of them. Steam allows 300 characters in the short description, and the median page spends 232 of them, so roughly a fifth of the field goes unused on a typical page in this sample.

Valve’s asset documentation requires at least 5 screenshots, so a median of 9 is not a target anyone set. It is where pages land when nobody is counting. The corpus benchmarks four facts and no more: screenshot count, video count, and the two description lengths. Those are the only ones that came through every change we made to the analyzer this year untouched, which is why a page you check against them gets a comparison that means something.

Only full reports store the screenshot deep dive, so the text counts run on 220 reports written between February 10 and September 14, 2026. Each holds a free-text verdict on the first screenshot, and the counts below come from literal phrase rules stored next to every number in the committed study file. Every one is a floor: a verdict making the same point in different words is not counted.

Of those 220 first screenshots, 50 (22.7%) drew no reservation at all, and the other 170 (77.3%) carried one. 46 of the 220 verdicts (20.9%) said the first screenshot did not communicate the core of the game or its hook, which is the specific failure that costs a click rather than a general complaint about image quality.

One split has to travel with that 22.7%. Our issue generator changed on July 1, 2026: the 193 reports written before that date contain 50 unqualified passes (25.9%) and the 27 written after it contain zero. The later prose is harsher throughout, so read 22.7% as a pooled figure across the window and 25.9% as the number for the bulk of it. Neither is a trend in Steam pages.

The same 220 reports listed 844 images they thought were missing, a median of 4 per report, and the categories cluster hard.

UI and menu shots were asked for in 140 of the 220 reports (63.6%), progression and payoff in 119 (54.1%), combat and action in 102 (46.4%), and character customisation in 40 (18.2%). Categories overlap, so those shares do not sum to 100. The full tally, including the five smaller categories, sits in the first-screenshot study, and the screenshot checker reads your gallery against the same rules.

Tags came off the live store page on a single day, not out of our database, because captures before July 26, 2026 merged community tags with Steam genres and feature declarations. We requested 785 pages on September 14, 2026 and 764 came back, the rest blocked by an age gate, a region block or a delisting.

599 of those 764 pages (78.4%) sit at the 20-tag display maximum, and only 45 (5.9%) show fewer than 15. Filling the slots is a solved problem in this sample. The gap is between shipped and unshipped games: 84.5% of the 373 released pages sit at 20 against 72.6% of the 391 coming-soon pages.

Two caveats ride along with that split. Twenty is a display ceiling rather than a count of everything applied, so a page sitting at 20 may carry tags the store page is not showing. And community tags get weighted by players after launch, which is part of why released pages sit higher here: some of that order was never chosen by the developer.

Order is where the same pages spend badly. Valve’s tag documentation says the top five tags “should paint a fairly clear picture of your game”. Across the 763 pages carrying at least five tags, 651 of 3,815 first-five slots (17.1%) hold a word the page already prints as a Steam genre, and 61.9% of pages spend at least one slot that way.

The three most common tags in the corpus are exactly the three you would guess, which is the point: Singleplayer appears on 550 of 764 pages (72.0%), Indie on 349 (45.7%) and Action on 332 (43.5%). A tag carried by seven pages in ten separates you from nobody. The tag study has the per-genre tables, and the tag optimizer works on your own five.

These four figures are quoted from our Next Fest demo study, published September 8, 2026, and nothing here recounts them. The sample is 286 unreleased Steam pages captured March 7 to August 18, 2026.

110 of the 286 (38.5%) listed a publicly visible demo, the one Next Fest entry requirement a store page can satisfy on its own. 131 (45.8%) showed no release timing beyond “Coming soon” or “To be announced” as of August 18, 2026. 27 (9.4%) had no trailer or video of any kind. And 70 (24.5%) cleared four basic pre-fest bars at once: a demo, at least one video, at least five screenshots, and a release date more specific than “Coming soon”.

The October 2026 fest runs October 19 to 26, and its deadlines are published per edition on Valve’s Next Fest page rather than as a fixed workback, so the only safe source is the page for the edition you are entering. Our October 2026 dates post tracks them, and the Next Fest prep tool checks your page against the same four bars the 24.5% figure measures.

Cite it as “Steam Store Pages in Numbers: September 2026 Edition”, Steam Page Analyzer, September 15, 2026, and link this page. Every chart above carries an embed control that copies an image tag plus the source line naming the sample and the date; keep that line and its link and you can put any chart on your own site for free. Past and future editions are listed on the reports hub.

How is the sample built?

Every page here was run through our free analyzer by somebody, is still flagged public by its owner, and has a resolved Steam name. One row per Steam App ID, Steam type game only. That consent gate is why the count is in the hundreds rather than the tens of thousands, and why you can see the pages it draws from on the leaderboard.

Why do the corpus and the studies have different sample sizes?

Because each one answers a different question. The 405-page corpus is frozen at July 30, 2026 so percentile comparisons stay stable. The screenshot counts run on the 220 pages that carry a stored deep dive, since only full reports have one. The tag figures come from 764 pages re-pulled live in one day. The 286 pre-launch pages are the unreleased subset from the August capture. Pooling them would produce a number belonging to no sample at all.

When does the next edition come out?

The week of October 27, 2026, after Steam Next Fest closes on October 26. Each edition restates the sample sizes and capture dates from scratch, so figures from two editions are comparable only where the windows say they are.