The Steam Workshop's Disappearing Mods (2026): 17,717 Mods Vanished in Six Weeks — and Took 10.8 Million Subscribers With Them

Every number anyone quotes about the Steam Workshop is a snapshot of what is there now. Six million subscribed mods. Twelve and a half billion subscriptions. Thirty-nine thousand item-tags. The catalogue, measured.

Nobody measures the other column. What left.

ModSignal tracks the Workshop continuously, and when a mod stops resolving — Steam reports it as no longer available — we keep the row and mark it removed rather than deleting it. That is how we can now say something the platform itself never does: in six weeks of tracking, 17,717 mods that were live disappeared.

They are not all dead weight. Among them is a Counter-Strike 2 map with 669,836 subscribers, a Project Zomboid stability mod with 460,467, and a 2022 Squad bot pack with 64,600. Together, the removed mods strand 10,832,905 subscriptions — eleven million installs pointing at things that are no longer there.

That is 0.087% of every subscription on the Workshop. Small, as a share. But it is not evenly spread, and where it lands, it lands hard. One game has lost 7.86% of its entire subscriber base to removals. Another lost a single mod worth 669,836 installs.

The quiet ledger

Start with the arithmetic, because it is smaller than you would guess and stranger than you would hope.

The removed mods are, in aggregate, less successful than the live catalogue — 6.98% cleared a thousand subscribers against 11.60% of everything still standing. Most of what disappears is what never got traction. That is the boring half.

The interesting half is the tail. Ten removed mods had more than 100,000 subscribers. Twenty-two had more than 50,000. And unlike a mod with forty subscribers, a mod with 669,836 does not vanish without someone noticing.

The giants that are gone

These are the largest removals ModSignal recorded in the window, ordered by the subscriber count each one held the last time Steam answered for it.

Two of these are worth reading closely, because they are opposite stories wearing the same red number.

eris_food_expiry and eris_nightvision_goggles are a pair of Project Zomboid quality-of-life mods from 2019, each with north of 450,000 subscribers and more than 10,000 favourites. Together, one game — Project Zomboid — accounts for 1,517,447 stranded subscriptions, the largest single-game total in the dataset. When mods this deeply embedded to a community are pulled, the loss is not the mod's; it is every player who now boots a save with a missing dependency.

SquidBots — Bots for Seeding [NOW PART OF BASE GAME] — 64,600 subscribers, 1,444 favourites, removed from Squad. Read the title. This mod did not die. It graduated: its function was absorbed into Squad itself, and the standalone version was retired. That is one of the genuinely good reasons a mod leaves the Workshop, and it is indistinguishable in the data from a takedown unless you read the name.

That ambiguity is the honest limit of what a removal count can tell you. Steam's API does not say why a mod stopped resolving. It says it stopped. Everything past that is inference.

What disappears is young

Here is the pattern that surprised us, and the one we trust least — for reasons we will get into.

Bucket the removed mods by how old they were, and compare that to the live catalogue:

Nearly 43% of everything removed was published in the last six months. A full 76.4% — more than three in four — was published within the last two years. The removed cohort's median age is 274 days. The live catalogue's median age is 1,192 days. Removals cluster at the entry point of the Workshop like sediment at the mouth of a river.

Turned into a rate, the same data reads: a mod published in 2026 is removed at 0.737%, a mod published in 2018 at 0.047% — a 15.7x difference.

The tidy reading is that the Workshop is good at clearing out fresh, low-effort uploads, and that what survives to old age has earned it. The tidy reading is also, in part, an artifact, and we should say why.

Our observation window is fixed at six weeks. A mod published two weeks ago had its entire life inside that window. A mod published in 2018 had six weeks of an eight-year life inside it. If removals happen at a roughly constant rate per mod-year, the young mods will look disproportionately removed simply because we watched their whole youth and only a sliver of the veterans' old age. Survivorship runs the other way too: an eight-year-old mod still standing is, by definition, one that survived everything that might have removed it.

So "young mods disappear more" is partly true and partly a measurement of our own window. What survives the caveat is the shape: the removals are concentrated, not uniform, and the median removed mod is 274 days old — the Workshop's churn is a young phenomenon whatever the cause.

Where the floor drops out, game by game

Removals are rare in most ecosystems. On the median game, well under half a percent of the catalogue was removed in six weeks. Then there is the other end.

And the games that lost almost nothing, which is the more common story and worth printing for contrast: They Are Billions, Final Fantasy XV, Let Them Trade, PAYDAY 2, Megaquarium, Car Mechanic Simulator 2021 and Killing Floor 2 all removed zero mods. House Flipper removed one out of 33,224 — 0.003%.

Scale does not predict removal. Stick Fight: The Game and Project Zomboid sit at ~1.8% with 120,000 and 64,000 live mods respectively; Deadside tops the table with 1,669. The rate is a property of the community and the game's publishing culture, not of the catalogue's size.

The metric that matters more than count is the value the removal destroyed. Sort the same way by subscriptions stranded, and the ranking inverts:

Squad is the anomaly. It lost 338 mods — a rounding error against its catalogue — but those mods carried 582,587 subscriptions, or 7.86% of everything Squad has ever distributed. Sea Power, similarly, lost 5.15% of its subscriber base in 56 mods. Meanwhile Garry's Mod shed 247 mods and barely noticed (0.05%), and Counter-Strike 2 lost 1.1 million subscriptions and also barely noticed (0.26%).

A big game can absorb a giant loss. A small one cannot. If you want to know whether a Workshop ecosystem is fragile, do not ask how many mods it has or how many it removed. Ask what share of its installs live in mods that are no longer there. For Squad, that number is one in thirteen.

What this means

If you are a creator, your subscriber count is a historical fact, not a live one. Subscriptions do not decrement when a mod is removed — not in the public numbers, and not in your head. The 669,836 people who subscribed to a Counter-Strike 2 map are still counted among "Counter-Strike 2 Workshop subscribers" long after the map stopped existing. Anything built on those counters, including this dataset and every leaderboard on Steam, inherits the same ghost. A removal is not a subtraction; it is an orphan.

If you are a studio, watch the 76% figure. More than three quarters of everything that vanished in our window was published in the last two years. Whatever the mechanism — abandoned early projects, policy sweeps, creators testing and quitting — the Workshop's removal churn is a function of its intake, not its back catalogue. The mods that disappear are the ones you never had time to notice, plus, occasionally, the one your whole community was built on.

If you rely on mods to play, the loss is not evenly distributed and not predicted by size. Project Zomboid lost 1.5 million stranded installs in six weeks; Killing Floor 2 lost zero. The 460,467 subscribers of eris_food_expiry had no signal that a 2019 mod with 12,753 favourites was about to stop resolving. The only durable protection is the unglamorous one: keep local copies of anything your save depends on, because the Workshop's floor can drop out under a mod that is seven years old and by every public measure, thriving.

The ecosystems in this article

• Counter-Strike 2 — 146,570 live mods, lost a 669,836-subscriber map and 1.1M stranded subs; absorbed it in 0.26%.
• Project Zomboid — 64,382 live mods, 1,187 removed, and the largest stranded total on the Workshop: 1,517,447 subscriptions.
• Squad — the fragile case: 338 removals worth 7.86% of its entire subscriber base.
• Euro Truck Simulator 2 — 107 removals, 526,072 stranded subscriptions, including a 224,079-sub lorry pack.
• Garry's Mod — 247 removals and 783,124 stranded installs, but only 0.05% of its base.
• Wallpaper Engine — 284 removals, 1,486,594 stranded subscriptions, a 0.08% dent in the largest category on the Workshop.
• DayZ — the highest removal count by share among big games at 1.17% (1,328 mods).
• RimWorld, Left 4 Dead 2 and Don't Starve Together — steady, low-rate churn with high-value casualties.

Related reading from this series: the churn economy on the 134,000 mods losing subscribers right now, the one-subscriber floor on the 654,035 mods parked at a single install, the dormant giants on the mods that hold half the Workshop without being updated, and the thumbnail divide on the listings that were never really presented to anyone.

Method: 7,371,702 Steam Workshop items tracked by ModSignal as of 28 September 2026 — 7,353,985 live and 17,717 removed — across 396 games. A mod is counted as removed when a scheduled snapshot finds Steam's GetPublishedFileDetails no longer resolves it (fresh === null); ModSignal marks the row is_active = false and preserves every prior snapshot and counter rather than deleting it. Three consequences follow, and all three matter. First, this is a floor, not a total: ModSignal snapshots a tiered sample of the catalogue rather than every item continuously, so removals outside the sampled tier are not detected, and 17,717 is the number the pipeline actually observed. Second, the "detected" week of a removal reflects when our snapshot reached that item, not when Steam removed it, so the week-by-week distribution is lumpy and should not be read as a removal timeline. Third, the age-at-removal comparison is bounded by a fixed six-week observation window applied to mods of very different ages — a young mod's entire life falls inside that window while an old mod's does not — so the "young mods disappear more" pattern is partly survivorship and partly a property of the window, which is why we report it with the caveat in the body rather than as a headline. Removal reasons are not recoverable: Steam reports that an item no longer resolves, never why, so a takedown, a creator withdrawal and a mod absorbed into the base game (see Squad's SquidBots) are indistinguishable in the data. Subscriptions stranded are the last reported subscriptions_current for the removed row and are not decremented anywhere by Steam, so platform-wide subscription totals continue to include installs of mods that no longer exist. All figures are a single consistent read of the live database on 28 September 2026; the counters move continuously, so a re-run will differ by a handful of items. Per-game rates are reported only for games with at least 1,000 live items, to keep small denominators out of the table.