The first screenshot should be the one frame that makes the genre and the core loop readable at a glance. Across 220 full reports our analyzer wrote between February 10 and September 14, 2026, only 50 first screenshots (22.7%) drew no reservation at all, and 45 of those 50 verdicts described the shot with the word “immediately” or “effectively”. The other 170 got a “but”.
The rest of this post is the tally behind that: what the 170 reservations said, what the same reports asked for further down the gallery, and a five-step order you can apply without making a single new image. Every figure carries its sample size and comes from stored report text or raw Steam facts, never from our scoring model.
How the sample was built
The population is the consented public set: a Steam app id somebody ran through the free analyzer, with the page still flagged public and a resolved game name. One row per app id at its most recent analysis, Steam type game. That is 785 pages, analyzed February 10 to September 14, 2026 (400 unreleased, 385 released).
Only full reports store the screenshot deep dive, so the text tallies run on the 220 of those 785 that carry one. Each holds three free-text fields: a verdict on the first screenshot, a note on the gallery as a narrative, and a list of images the analyzer thought were missing.
Free text cannot be scored, so I did not try. I counted literal phrases with rules written down in scripts/study-screenshots.mjs and stored next to every count in the committed JSON. Every number below is a floor: a verdict that makes the same point in other words is not counted.
One honest wrinkle. Split the 220 at July 1, the date our issue generator changed, and you get 193 earlier reports with 50 unqualified passes (25.9%) and 27 later ones with zero. The later prose is more critical throughout, so read 22.7% as pooled and 25.9% as the bulk of the window, not as a trend in Steam pages.
What the 170 reservations said
Counted across all 220 verdicts:
| What the verdict said about shot 1 | Reports | Share of 220 |
|---|---|---|
| Says it does not communicate or show something | 88 | 40.0% |
| Names the core loop, the hook or the genre as the casualty | 86 | 39.1% |
| Both of the above in the same verdict | 46 | 20.9% |
| Praises the image on looks (atmospheric, cinematic, a render) | 56 | 25.5% |
| Names a better image already sitting on the same page | 26 | 11.8% |
| Calls it cluttered, busy or hard to read | 14 | 6.4% |
| Calls it static, sparse or low energy | 10 | 4.5% |
That fourth row is the one I keep rereading. A quarter of these verdicts open by calling the image beautiful and then spend the second half explaining what it fails to say. Valve is blunt about this in the graphical assets documentation: “Screenshots should exclusively show the gameplay of your game. This means avoiding using concept art, pre-rendered cinematic stills, or images showing awards, marketing copy.”
And 26 reports say the fix is free. The better opener is already uploaded, three slots down, doing nothing.
Our committed benchmark corpus (405 pages, 2026-02-10 to 2026-07-30) counts the same problem through the issue list instead, and lands lower: “first screenshot does not show core gameplay” fired on 16.8% to 32.0% of pages depending on the emitter regime (n=155 and n=250). Expected, that: an issue only fires when something is wrong enough to be an issue, while the deep dive comments on every page.
The five-step order
Grounded in the tallies above and the counts below:
- The core loop, in one frame. Whatever the player spends the most minutes doing, mid-action, legible at thumbnail size. This is the shot 40% of verdicts said was not doing its job.
- The thing only your game does. The hook that 86 verdicts named. If a competitor could use your shot 2 unchanged, it is the wrong image.
- Systems and UI. The single most requested missing image, in 140 of 220 reports.
- Progression and payoff. The end state, the built city, the full skill tree. Asked for in 119 reports.
- Variety. A second biome, a boss, co-op, a different act. Combat or action variety was missing in 102 reports.
Shots 6 and beyond fill the shelf. They do not repair shot 1.
What the galleries were missing
All 220 reports listed at least two missing images. The median was 4, the maximum 6, and none came back empty. That is 844 requested images across 220 galleries.
UI and progression are unglamorous, which is why they are missing. They are also the two questions a buyer with a wallet open is asking: what will I be looking at for thirty hours, and what am I working toward. The benchmark corpus flags the same theme on 20.4% to 61.9% of pages across its two emitter regimes (n=250 and n=155), the widest regime spread in that dataset, so read the direction not the estimate.
More screenshots does not fix the first one
If the first screenshot were mostly a supply problem, pages with more images would pass more often. They do not.
Every bucket sits between 18% and 31%, with no direction, and the smallest galleries score highest. Thin cells, so I would not bet on the ordering. The claim I will defend is the flat one: adding a tenth image does nothing for the first. Order and count are separate problems, and only one needs an artist.
How many screenshots pages actually carry
Raw Steam facts, unaffected by any rubric change. Across the 785 pages, the median is 9 screenshots (25th percentile 7, 75th percentile 12, mean 10.3). 91 pages (11.6%) sit at exactly 5, which is Valve’s stated minimum: “You must provide at least 5 screenshots of your product.” 364 (46.4%) carry 10 or more.
Unreleased pages run a median of 8 (n=400) against 10 for released ones (n=385). Our committed percentile corpus, 405 pages over 2026-02-10 to 2026-07-30, gives the same three medians: 9 overall, 8 for coming-soon pages (n=225), 10 for released (n=180).
| Steam genre | Pages | Median screenshots |
|---|---|---|
| RPG | 155 | 10 |
| Action | 342 | 10 |
| Adventure | 307 | 10 |
| Simulation | 230 | 10 |
| Indie | 531 | 9 |
| Casual | 258 | 9 |
| Strategy | 208 | 8.5 |
Genres overlap on Steam, so a page counts in every genre it lists. The spread is one image wide. Count is not where the variance lives.
What this study cannot tell you
- It measures pages, not sales. No wishlists, no click-through. For outcome data see store page conversion benchmarks and Chris Zukowski’s normal click-through rates.
- The sample is self-selected. Pages whose developers went looking for an audit, which biases toward teams already paying attention.
- The verdicts are one reader’s opinion, at scale. A model looked at each gallery once. Consistent, not authoritative.
- The phrase counts are floors. “This reads like a generic fantasy vista” makes the point without a word my rules match.
Frequently asked questions
Should the first Steam screenshot be gameplay or key art?
Gameplay. Valve’s own asset documentation says screenshots “should exclusively show the gameplay of your game” and names pre-rendered cinematic stills and concept art as things to avoid. In our sample, 56 of 220 verdicts (25.5%) praised the first image on its looks while objecting to what it failed to say about the game.
How many screenshots should a Steam page have?
Steam requires at least 5. The median page in our 785-page sample carries 9, and 46.4% carry 10 or more. Released pages run a median of 10 against 8 for unreleased ones. More than 10 is fine; it just will not fix a weak opener.
Does reordering screenshots actually change anything?
We cannot measure the sales effect from this data. What we can say is that 26 of 220 reports (11.8%) explicitly named a better first image already uploaded to the same page, so for those developers the change costs nothing but a drag.
What should screenshot 3 be?
Systems or UI. It was the most requested missing image in the sample, absent from 140 of 220 galleries (63.6%), and it answers the question a buyer asks right after “what is this”: what will I actually be looking at.
Reorder yours in ten minutes
- Open your store page and look only at the first image. Can a stranger name the genre and the core loop from it? If not, 40% of the pages we looked at share your problem.
- Scan the rest of your gallery for a better opener. One in nine reports found one already on the page.
- Check slots 3 and 4 for a UI screen and a progression or end-state screen. Nearly two thirds of galleries were missing the first, more than half the second.
- Count your images. Exactly 5 puts you in the bottom 12% of 785; the median is 9.
- Re-run the page: the screenshot checker for dimension and format rules, the store page checklist for the rest, the free analysis for the gallery read.
Composition advice is in our screenshot guide, asset rules in page asset requirements, and the same corpus counted for demos in the Next Fest demo data post.
Cite this study: “Which Steam Screenshot Goes First? 220 Reports, Counted”, Steam Page Analyzer, September 16, 2026. 785 consented public Steam store pages analyzed February 10 to September 14, 2026, of which 220 carry a full screenshot deep dive. Canonical URL: https://steampageanalyzer.com/blog/steam-screenshot-order