Steam Review Score Thresholds: All 9 Rating Labels (2026)

Steam assigns your review label from exactly two numbers: the percentage of positive reviews and the total review count. Overwhelmingly Positive requires 95%+ positive across 500+ reviews, Very Positive requires 80%+ across 50+, Mostly Positive starts at 70%, and Mixed covers 40-69%. If that’s all you came for, the full nine-label table is directly below.

One honesty note before the table: Valve has never published these thresholds. They are reverse-engineered from observing thousands of store pages, and the community-documented version (Lars Doucet did the best early work in his user rating analysis) has held stable for years. Everything else in this guide — how the score is computed, what gets excluded, how the labels feed Steam’s algorithm, and what each label is worth in sales — comes from the official Steamworks review documentation and public data.

Nine labels, two inputs. Here is the complete grid:

Three details in that table trip people up constantly:

• —Below 10 reviews, you have no label at all. Your page shows a raw review count and the line “Need more user reviews to generate a score.” You are invisible to every review-score filter on Steam until review number 10 lands. Getting to 10 fast is a launch-week priority, full stop.
• —“Positive” and “Very Positive” are the same percentage band. The only difference is volume. A game at 88% with 30 reviews shows “Positive.” The same game at 88% with 60 reviews shows “Very Positive.” Nothing about player sentiment changed — the count crossed 50.
• —Overwhelmingly Positive is a volume gate as much as a quality gate. A game at 97% positive with 300 reviews displays Very Positive. It needs 500 total reviews before Steam will print the best label on the store.

The two boundaries worth tattooing on your launch plan are 70% (Mixed becomes Mostly Positive) and 80% (Mostly Positive becomes Very Positive). Those are the thresholds where shopper behavior visibly changes, and the ones our review management guide is built around.

The math is deliberately simple: positive reviews divided by total counted reviews. No stars, no weighting by review length or reviewer reputation. The complexity is entirely in which reviews count.

Per the Steamworks documentation, only reviews from direct Steam purchases feed your score. Reviews from activated Steam keys and free weekends are published and readable on your page, but they are excluded from the percentage. (Family Sharing borrowers don’t get a review box at all — the game has to be on your own account.) This is why handing out thousands of keys never manufactures a score: those reviewers show up in the list, not in the math. Reviews tied to refunded purchases stop counting too, which means a refund costs you the sale and the social proof in one transaction — our refund rate data covers how often that happens by genre.

On the player side, the Steam review system has three eligibility rules: the product on the reviewing account, roughly 5 minutes of recorded playtime, and a non-limited account (at least $5 spent on Steam, ever). The review itself is a binary thumbs up or thumbs down plus free text — there is no 7/10 on Steam. Every lukewarm “it’s fine I guess” gets forced into a yes or a no, and your label is the aggregate of those forced choices.

Once your game has been out long enough, the store page displays two summaries: Recent Reviews (the last 30 days) and All Reviews (lifetime). Valve introduced the split in May 2016 specifically so shoppers could tell whether a game’s reputation reflects its current state. The recent score appears once the game has been available for 45 days and has enough reviews inside the 30-day window; with thin recent volume, Steam just shows the lifetime score alone.

For developers this split cuts both ways:

• —It lets you recover. A rough launch is not a life sentence. Patch the game, and the recent score reflects the patched game while the lifetime score is still digesting launch week. Shoppers see “Recent: Very Positive” above “All: Mixed” and read the story correctly: it got fixed.
• —It exposes decay. A monetization change or a botched update shows up in the recent score within days, long before the lifetime number moves. A game showing “Recent: Mostly Negative” under “All: Very Positive” is wearing its controversy on its storefront.

I’d check the recent score weekly: it reacts within days, while the lifetime number can take months to move.

Since March 2019, Valve runs anomaly detection on review activity. When a game gets hit by what Valve calls an off-topic review bomb — a sudden spike of reviews about something other than the game itself, like a DRM dispute, a developer’s politics, or events in a different game — Valve identifies the date range and by default excludes every review in that window from the score. The reviews stay visible and individual users can opt back into counting them in their own settings, but the default score most shoppers see skips the bombed period. The Steamworks docs state it plainly: “In the case of off-topic bombs, those reviews will not contribute to your overall score.”

Two practical notes. First, the protection is for off-topic bombs. If players are angry about the actual game — a bad patch, broken promises, removed features — those reviews count, and Valve will not save you from them. Second, you can flag an anomalous period to Valve through Steamworks rather than waiting for automated detection. The review histogram on your page is the tool for spotting these windows yourself: a clean cliff on a specific date almost always has a specific cause.

The label on your store page is pure percentage plus count gates. But the moment Steam has to rank games against each other — search results, top-rated lists, recommendation feeds — a raw percentage breaks down. A game with 9 positive reviews out of 9 should not outrank a game with 4,800 positive out of 5,000. Confidence adjustment fixes this, two ways.

The Wilson score interval asks: given this sample of reviews, what’s the lowest true score we can claim with 95% confidence? Small samples get wide intervals and sink. The Bayesian approach instead starts every game at a prior (assume 50% until proven otherwise) and lets review volume pull the estimate toward the observed percentage. SteamDB’s rating is the best-known public implementation — it moved from Wilson to a Bayesian-flavored formula that drags low-volume games toward 50%, which is why a 100%-positive game with 12 reviews ranks below an 89% game with 9,000 reviews on their charts.

Valve has never published its internal ranking formula, so the honest claim is narrower: in every observable Steam reviews ranking surface, volume increases the weight of your score. Our search ranking breakdown goes through the evidence, but the working model is simple — 500 Very Positive reviews outrank 15 Very Positive reviews for the same query, every time. This is an observed pattern, not a Valve guarantee. Either way, the percentage sets your label, and the review count decides how much weight that label carries when Steam ranks you.

Labels are the first social proof a shopper processes, often before reading a single review. In our store page conversion benchmarks, pages with Overwhelmingly Positive reviews convert first-time visitors at 2-3x the rate of pages showing Mixed. The gradient between those endpoints steps at the label boundaries instead of sloping smoothly:

The cliff is at Mixed. A 71% game and a 69% game are nearly identical products, but the 69% game wears a yellow warning label that a meaningful share of shoppers treat as a hard no. The same step exists in Steam’s algorithm: Mostly Positive is roughly neutral for visibility, Mixed is a negative signal, and below 40% comes real suppression. Crossing 70% in either direction changes both how many people see your page and what fraction of them buy — the two multiply. That’s why a two-point move around a threshold can shift revenue more than a ten-point move mid-band, and why your wishlist conversion rate at launch is so sensitive to where the first hundred reviews land.

Review counts are a proxy for sales, and the exchange rate is documented: the current review to sales multiplier is about 30x for recent releases, with a 20-40x working range. Run the label gates through that multiplier and the thresholds turn into sales targets:

Read that chart as a reality check. Your first score label arrives around 300 sales. Very Positive becomes possible (not guaranteed — you still need the 80%) around 1,500 sales. And Overwhelmingly Positive is gated behind roughly 15,000 units at typical review rates, which is why the label is rare: by indie revenue data standards, most Steam games never sell 15,000 copies and are mathematically locked out of the top label. If you want to see what those unit counts mean in dollars for your price point, run them through the revenue calculator — it applies the multiplier, refund rates, and Steam’s cut in one pass.

Here’s the math nobody does before promising their team “we’ll patch our way out.” Say you’re at 65% positive with 200 reviews: 130 positive, 70 negative. To reach 70%, you need (130 + x) / (200 + x) to hit 0.70. Solve it and x is 34 consecutive positive reviews. Realistically your post-patch reviews won’t run 100% positive — at a healthy 80% positive on new reviews, you need about 100 new reviews to cross the line, which at the ~30x multiplier implies roughly 3,000 additional sales. Climbing out of Mixed is a sales problem wearing a reputation costume.

That math should shape your tactics:

• —Fix the things the negative reviews actually name. Sort your negative reviews by helpfulness and count complaint categories. The top two usually account for most of the damage, and patching them is the only move that changes the sentiment of future reviews — the only ones you control.
• —Tell players you fixed it. Patch notes, a pinned forum post, and a visibility round update that says “we heard you, here’s what changed.” Players who left a negative review over a fixed bug will sometimes update it — but only if they find out about the fix.
• —Ask for reviews the legal way. You may ask players to leave a review. You may not pay for reviews, gate rewards behind them, or ask specifically for positive reviews — Valve treats all of those as manipulation, and the penalty is a store page warning banner or removal. A neutral in-game prompt at a high-engagement moment (“Enjoying the game? A Steam review helps a lot”) is standard practice and fine.
• —Respond to reviews sparingly and fix-first. Developer responses on top negative reviews work when they say “this is fixed in patch 1.2,” and backfire when they argue. Our review management guide covers the full playbook, including what to do in the first 48 hours of a review crisis.

How do Steam reviews work?

Players who have a game on their account with at least about 5 minutes of playtime, on a non-limited account, can post a thumbs up or thumbs down with written text. Steam divides positive by total counted reviews to get your percentage, and the percentage plus the review count determines which of nine labels your page displays. Only direct Steam purchases count toward the score; key activations and free-weekend reviews are visible but excluded. The mechanics feed directly into visibility — see our Steam algorithm guide.

How many reviews do you need to get a score on Steam?

Ten. Below 10 reviews, your page shows “Need more user reviews to generate a score” and no label. At the ~30x review to sales multiplier, that first label typically arrives around 300 units sold. The next gates are 50 reviews (unlocks Very Positive) and 500 reviews (unlocks Overwhelmingly Positive).

What format do Steam reviews use?

A binary recommendation — thumbs up or thumbs down — plus free text. There are no star ratings or numeric scores. Review text supports the same BBCode-style formatting used in Steam community posts (bold, spoiler tags, headers, lists). The binary format is why the threshold labels carry so much weight: the label is the only graded signal shoppers get, and it moves conversion benchmarks at every boundary.

How do you review games on Steam?

Open the game’s store page or your library entry, find the “Write a review” box, pick thumbs up or down, and write your text. You need the game on your account, roughly 5 minutes of playtime, and a non-limited account ($5 lifetime spend on Steam). Reviews can be edited or deleted anytime.

Does review bombing permanently damage your score?

Usually not. Off-topic review bombs are detected by Valve and excluded from your score by default, and on-topic anger fades from your Recent score 30 days after it stops. The lifetime score absorbs the hit but recovers as positive volume accumulates — the same climb-out math from our Mixed-to-Mostly-Positive section applies. The review management guide covers crisis response step by step.

Ready to see how your review score interacts with the rest of your store page? Run your page through the free Steam Page Analyzer — it reads your label, your review velocity, and every conversion element around them, then tells you which fixes matter most. Then model what crossing your next threshold is worth in revenue with the Revenue Calculator and check how the games at the top handle social proof on the leaderboard.

For the surrounding strategy, read our review management playbook, the review to sales multiplier lookup table, and the Steam algorithm explainer. Your score is the most visible number on your page, and you now have the same table Steam uses to print it.