Steam Review Bomb Response: The Playbook Valve Actually Published

What Valve's off-topic system actually does -- measured

This is the part that gets misrepresented most often, so we measured it rather than repeating what other articles say.

Steam's public review endpoint takes a parameter that toggles off-topic exclusion. On 2026-07-31 we queried seven famously bombed games twice each -- once with off-topic periods included and once with them excluded -- using store.steampowered.com/appreviews/<appid>?json=1&language=all&purchase_type=steam&filter_offtopic_activity=0 and then the same URL with the filter set to 1. Anyone can reproduce this.

Game	Off-topic included	Excluded	Difference	Reviews excluded
Slay the Spire 2	56.88% Mixed	60.77% Mixed	+3.89 pts	25,985
Borderlands 2	88.86% Very Positive	89.82% Very Positive	+0.96 pts	5,152
Cyberpunk 2077	86.88% Very Positive	87.28% Very Positive	+0.40 pts	6,405
DOOM Eternal	91.22% Very Positive	91.50% Very Positive	+0.28 pts	2,541
Helldivers 2	76.00% Mostly Positive	identical	0.00	0
Space Marine 2	84.39% Very Positive	identical	0.00	0
Firewatch	90.50% Very Positive	identical	0.00	0

In none of the seven did exclusion change the review bucket. Slay the Spire 2 had roughly 26,000 reviews excluded from its score and still reads Mixed. The largest rescue in the sample was 3.89 percentage points.

And the cases most often cited as proof the system works were never marked at all. Helldivers 2's PSN bomb -- the most famous storefront-policy bomb in Steam's history, around 300,000 negative reviews according to Push Square citing SteamDB -- shows byte-identical filtered and unfiltered scores. Valve never excluded it. Neither Space Marine 2's. Firewatch's 2017 bomb predates the system entirely, and Valve said at the time it "decided not to unilaterally grandfather in what's happened in the past."

Three more things developers routinely get wrong:

• The unit of exclusion is a time window, not a review. Valve draws a box around a date range; it does not evaluate reviews individually.
• Legitimate reviews inside that window are excluded too. Asked directly, Valve confirmed: "Unfortunately, this is correct... it isn't feasible for us to read every single review." Your happy customers from those days stop counting as well.
• Nothing is deleted, and any user can switch the exclusion off in their own Store Preferences.

None of this means skip contacting Valve. It means stop treating Valve as the plan.

Valve's own review bomb checklist, quoted in full

Valve publishes a five-item checklist in the Steamworks user reviews documentation. Almost no ranking article quotes it. Here it is verbatim:

"Reach out to us via email or through our Partner Support channel. We can give you more information, along with adding extra moderators to help you flag abuse and keep an eye on your discussion hub."

"Spend time figuring out where the review bomb came from."

"Don't censor or publicly argue with reviewers - it's a battle you won't win. We've never seen it work at stopping negative reviews, but it's an effective way to keep them coming."

"Don't let someone on your team become a target. If you do communicate, consider doing so from an anonymous developer account. You can create an account specifically for this and add it to your partner group from your Permissions Overview page."

"Prioritize the mental and emotional well-being of your team. This is a potential cost of a review bomb that is hard to measure, but will likely be the biggest challenge faced during an event like this."

The fourth item is the single most valuable and least-known instruction in Valve's documentation, and it is useless if you read it for the first time during a bomb. Create the anonymous partner account now, while nothing is on fire. Add it via Permissions Overview, give it the right to post announcements and developer responses, and agree in advance that all crisis communication goes out through it. That way the voice speaking publicly is a studio, not a named human whose mentions you will be reading at 3am.

Ask for the extra hub moderators early too. It is a moderation problem before it is a score problem.

The line that actually matters is 40%, not 70%

Most developers panic at the wrong threshold. They watch Very Positive slip to Mixed and treat it as an extinction event.

Valve's visibility documentation says the opposite:

"As long as your game's reviews are Mixed or above (40%+), review score is not a factor in algorithmic visibility. If your game dips below 40% into Mostly Negative, your game will be less likely to be featured."

Valve has also described the size of that drop: the boost "is essentially the same for all the Mixed or above review buckets," but "a Mixed game receives over 500% more boost than a game in Mostly Negative."

Very Positive to Mixed costs you nothing algorithmically. Mixed to Mostly Negative costs a lot. That is the line to defend, and it sits far lower than most people assume. Falling to Mixed still costs you conversion and human curation -- press, streamers, curators, and the shopper deciding in three seconds -- which is a real problem and the subject of a separate playbook. But it is not the cliff. Know where you sit relative to 40% before deciding how hard to fight; the tier thresholds are documented here.

Recent versus lifetime, and the 2026 wrinkle nobody has caught up with

Your store page shows two scores: the past 30 days and the lifetime total. During a bomb they behave completely differently, and conflating them causes a lot of unnecessary despair.

Recent is a rolling 30-day window. Bomb reviews stop counting toward it 30 days after they were posted, regardless of anything you do -- arithmetic, not a favour from Valve. It is also why Recent collapses first and hardest: a few thousand reviews against a 30-day denominator is devastating, while the same reviews against a lifetime denominator of 200,000 barely register.

Ready or Not is a clean illustration. After content changes made for console certification in July 2025, PCGamesN reported roughly 4,700 negative reviews, which pushed the recent rating to mostly negative while the lifetime rating stayed Very Positive. One event, two very different numbers on the same page.

Then there is the change most competing articles have not absorbed. Since August 2025, Valve shows shoppers their own language's review score by default: "The Review Score displayed to users will be based on their primary language. This change impacts games that have more than 2,000 publicly visible reviews, and with at least 200 written reviews in at least one language."

Measured on 2026-07-31, Slay the Spire 2 displays as Very Positive at 91.9% to an English visitor while its all-language score is Mixed at 60.8% and its Simplified Chinese score is Mostly Negative at 31.4%. Same game, same day, three different store pages depending on who is looking.

For a single-language bomb in 2026, this matters far more than anything Valve's off-topic team will do. It also means you must measure by language, not in aggregate. An aggregate number can hide both a perfectly healthy English audience and a collapsed regional one.

One observation of our own, offered as analysis rather than fact because Valve has not documented it: on the two live pages we checked, the lifetime "All Reviews" headline was language-specific but "Recent Reviews" appeared to be all-languages. If that holds generally, a foreign-language bomb hits your Recent score at full force while being quarantined out of your headline. Verify it on your own page before relying on it.

The hour-by-hour playbook

Hour 0 -- classify, do not respond

Valve's tool flags anomalous activity in near-real-time, but it notifies humans who have not looked yet. Nothing is happening on Valve's side for a while.

Check sales. Classify product versus context. Check which languages. Check where you sit relative to 40%. Switch external communication to the anonymous partner account. Reply to nothing. Anything you write in the first hour gets written in the worst emotional state you will be in all week.

Hour 1 -- get the facts straight internally

Name the trigger precisely. Not "players are upset about the patch" but "the 0.103.2 change to card X made the infinite loop build unviable." You cannot write a useful announcement until someone in the room can state the grievance in one sentence a critic would agree with. Then decide who owns communication and who owns the fix -- those should be different people.

Hour 6 -- one central acknowledgment

This is peak velocity and roughly when coverage starts. Two actions:

• Email Partner Support. Even for a product bomb, this is how you get extra hub moderators.
• Post one store announcement. One. Not per-review replies. A thousand people making the same point get one central answer; per-review replies there are a thousand invitations to argue.

Vague requests for patience fail. Mega Crit's response to Slay the Spire 2's April 2026 wave was "Trust the process" -- and the bomb continued. Specific commitments with a named owner and a date work. Say what you are changing, who is doing it, and roughly when.

Day 1 -- the decision point

Recent Reviews has already moved buckets. Lifetime has barely twitched if you are a large game.

• If the trigger is reversible, reverse it today. This is the only intervention with documented, large, fast effects. Sony reversed the Helldivers 2 PSN requirement within days and the community organised over 200,000 positive reviews in response. id Software removed Denuvo Anti-Cheat from DOOM Eternal after around 1,200 negative reviews in a day; it reads 91.5% Very Positive today.
• If it is a product defect you cannot reverse, publish the fix with an owner and a timeline.
• If it is neither -- a business decision you are keeping -- acknowledge it, do not relitigate it, redirect to forward commitments, and accept that the score will move slowly.

Day 3 -- velocity breaks

Valve notes bombs generally "calm down after a few days." Now, and not before, start selectively replying to individual reviews -- specifically to reviews containing a factual error you can correct or a bug you have already fixed. Not to opinions. Not to politics. The standard response framework applies, and tone matters more than usual because every reply is being read by people deciding whether to buy.

Flag genuinely abusive reviews -- and remember Off-Topic flagging does nothing to your score. If Valve marked a period, measure the delta yourself with the filter parameter above and expect a small number.

Week 1 -- ship

Recent still carries the full bomb. Your lifetime score may look far healthier than the raw data because of language-specific scoring -- do not let that lull you.

Ship the fix. Nothing else meaningfully changes the trajectory. Shipping also unlocks a free platform mechanic: "If a user leaves a negative review for a game and subsequently continues to log substantial playtime, Steam may prompt them to update their review." That only fires for people who keep playing, which is another way of saying that fixing the game beats managing the storefront.

Do not run a "please leave a positive review" campaign. Valve's rules prohibit soliciting reviews for rewards and prohibit in-application review prompts outright, and there is precedent for studios being removed from Steam over review manipulation.

Week 4 -- the first honest read

The bomb has aged out of the 30-day window. This is your first clean signal. Measure per language. Compare against your pre-bomb baseline, not against zero.

Account for normal decay before you diagnose a lingering problem. Valve observes that "most games slowly trend downwards over time, even if they haven't changed in any way." Some of what you see at week four is gravity, not grievance.

Read the bomb as a spec

The most useful reframe available: a bomb is usually a feedback channel operating at maximum volume because the quieter ones did not work.

Casey Yano of Mega Crit, on why Chinese players used Steam reviews during Slay the Spire 2's bomb: "I think it's difficult for players to feel like they're heard. So I understand why they choose to use Steam, doing something that would obviously impact things like visibility in the store and things like that to try to portray the feedback."

Extracting the one change players actually want is usually the whole fix. Compare two 2025-2026 cases that ran the same play in opposite directions.

Hollow Knight: Silksong (September 2025). Trigger: a Simplified Chinese translation players described as tonally wrong. Team Cherry's response was maximally specific -- they stated that "our first priority remains to improve Silksong's Simplified Chinese translation," brought in an external team, gave a three-to-four-week timeline, and publicly recommended a fan translation mod as an interim fix, then shipped the fan team's translation into the game. Measured on 2026-07-31: Very Positive in every language view including all-languages. Full recovery.

Slay the Spire 2 (2026). Trigger: balance changes from a beta branch pushed to main, and later a second, unrelated controversy. The public response was "Trust the process," then silence through a second wave. Valve did mark a period and excluded around 26,000 reviews. Measured 3.5 months later: still Mixed at 60.8%. What is protecting the store page is not Valve's exclusion -- worth 3.89 points and no bucket change -- but the language split showing English buyers 91%.

Two cases with many confounders is suggestive, not proof. But the difference between them was not effort or budget. It was whether anyone named a specific fix with a date.

The uncomfortable finding to sit with

Here is the counter-evidence, because leaving it out would make everything above less trustworthy.

Valve looked at what happens to review scores after bombs and reported that the score "generally recovers, in some cases fully back to where it was beforehand" -- a "temporary distortion." Good news. But in the same analysis:

"In some review bomb cases, the developers made changes in response to the community dissatisfaction, and in others they didn't - but there didn't seem to be much correlation between whether they did and what happened to their Review Score afterwards."

Read that honestly: Valve's own longitudinal look found that developer action did not obviously predict score recovery. Everything in this playbook is still worth doing -- it is how you keep players, keep your team, and avoid the self-inflicted second bomb -- but nobody should promise you a number.

What does reliably move a score back is new reviews, not recovered ones. Independent analysis of roughly 100 million Steam reviews found that just 12.7% of reviews are ever updated at all. Your recovery is mostly people who have not bought the game yet.

What making it worse looks like

The fastest way to double a bomb is to give it a second story.

In May 2026, Square Glade Games replied to negative reviews of Outbound -- a game that arrived with roughly 1.5 million wishlists -- with this:

"Sorry to hear that Outbound is not your cup of tea. But that is of course totally understandable. No hard feelings. Feel free to send a support request to the Steam support to get a full refund on your purchase. If you do so, we would appreciate if you would update or remove your negative review. Thanks a lot."

The request became the story. The developer then had to apologise publicly: "First and foremost, we truly apologize for how our recent replies have come across. We understand that our communication has felt wrong to many of you, and for that, we are genuinely sorry." The community response generated additional negative reviews about the conduct itself.

The failure was not malice. It was a tired team writing replies without a policy, which is the argument for writing one before launch rather than during a crisis.

The other reliable amplifiers: arguing publicly, gatekeeping on playtime, calling it a review bomb in your own announcement, and replying to every review individually. Valve's line on the first one -- "we've never seen it work at stopping negative reviews, but it's an effective way to keep them coming" -- is worth taping to a monitor.

The one documented recovery worth copying

DON'T SCREAM TOGETHER launched in December 2025 to a mostly negative Steam rating, largely from audio and microphone-calibration problems. Digital Cybercherries, a five-person studio, did four things and climbed to Mixed while selling 100,000 copies in two days.

Their framing is the best sentence in this entire subject:

"People do not lose trust because a game has problems. They lose trust when developers pretend those problems are not there. So we decided to face everything head on."

The four things:

• An honest public post that named the actual cause rather than apologising in the abstract -- in their case, that "much of the instability and distortion was caused by the Windows 11 Audio Enhancements system."
• Same-day hotfixes. "Within the first day, we released several hotfixes and our first major update."
• Individual replies to affected reviewers once fixes shipped -- "going through each review from players affected and letting you know that the fix is now live." Note the sequencing: replies came after the fix, carrying news rather than promises.
• A follow-up post closing the loop instead of going quiet after the first announcement.

They were also candid about the cost: "Seeing the negative reviews has been really hard. It's a bitter pill to swallow, and I'm not going to pretend it hasn't hit us emotionally."

Your team is the cost nobody measures

Valve puts team wellbeing in a five-item emergency checklist and calls it the item that "will likely be the biggest challenge." That is not garnish -- it is the most-cited consequence in developers' own accounts of bombs.

Casey Yano, on how Slay the Spire 2's bomb landed across Mega Crit: "I think it depends on who in the studio. Because I'm used to negativity and harsh feedback, and I'm pretty immune to it. But some of our other folks on the team, they weren't with us with the first game. And I think it was much more extreme than any previous types of negative feedback."

That is the practical finding. Veterans absorb it. People on their first shipped game do not, and they are usually the ones with the review page open. Three steps:

• Assign review-reading to one or two named people and tell everyone else not to look. Undirected exposure is where the damage happens.
• Route public communication through the anonymous partner account so no individual accumulates a pile-on that outlasts the bomb.
• Set an end time for each day. The bomb will still be there tomorrow; your ability to write a good announcement will not survive a 20-hour day.

The longer-term version of this problem -- what repeated exposure to negative reviews does to a small team over months -- is covered in our piece on review-driven burnout.

What to have in place before it happens

You cannot prevent bombs. You can shorten the gap between the first review and your first informed decision, which is the only variable you fully control.

• The anonymous partner account, created and permissioned. Ten minutes now; unavailable when you need it.
• A named comms owner and a named fix owner, so nobody negotiates roles at 2am.
• Velocity alerting that fires outside business hours. Steam does not tell you when your score starts moving; developers routinely learn about it from a player DM.
• Per-language score tracking. The aggregate number is not what your buyers see, and a regional collapse can hide inside a healthy global average.
• Your distance to 40%, calculated in advance.
• A written reply policy, including the sentence you will never send. Square Glade Games did not have one.

Run a free review audit to see your current tier, your distance to the next threshold in both directions, and how your sentiment is trending by language.

The bomb is loud and it is temporary. The thing that determines where you land four weeks out is whether you named a real fix, shipped it, and kept your team intact while doing it. Everything else is storefront management.