The Steam Workshop's Copy-Paste Descriptions (2026): 189,543 Mods Share Their Blurb With a Stranger — and the Shared Text Looks Like It Helps Until You Compare Creators to Themselves

Every listing on the Steam Workshop has a description field. Most creators write something for it. What almost nobody measures is how often that text is not theirs — how often two mods in the catalogue carry, character for character, the same blurb.

We measured it. In the live subscribed catalogue, 51,180 descriptions are shared by two or more mods — and those descriptions carry 189,543 mods between them. That is 3.2% of the entire catalogue and, because shared-text mods skew large, 8.13% of every subscription on the platform: 1,012,918,512 installs sitting behind text that at least one other listing also printed.

This is the fourth field in a series that has now covered a mod's missing description (the blank-page problem), its missing image (the thumbnail divide), and its shared name (the name collision). The description had only ever been asked whether it exists. Nobody asked whether it was unique. It very often is not.

The gradient looks like an advantage

Split the 2,006,226 mods whose description runs to at least 100 characters by how many other mods carry the identical text, and a clean, monotone ladder appears:

Read at face value, a mod whose description is shared with fifty or more others is 1.29x more likely to reach a thousand subscribers than a mod with a blurb of its own, and its median install count is 1.69x higher. Every band is one way up. There is no sweet spot: the more your text is somebody else's, the better it looks.

That reading is wrong, and the reason it is wrong is the reason this series keeps repeating itself. Before believing a pooled gradient, you have to ask what kind of listing ends up in the top band — and the answer here is unusually blunt.

Who shares a description with 50 other mods

Of the 51,180 shared descriptions, 44,064 — 86.1% — belong to a single creator. They are not strangers colliding on the same words. They are one account that shipped a block of items behind one blurb. Those single-creator groups hold 158,280 of the 189,543 mods (83.5%). Only 7,116 groups (13.9%) are text that genuinely crosses between different creators, and only 3.9% cross between different games at all.

So the 50+ band is largely not "text that 50 rivals used." It is text that one prolific creator used 50 times. And a creator who publishes in blocks of fifty is, unsurprisingly, not an average creator: their catalogue is far larger than the platform norm, and large catalogues clear a thousand subscribers at a much higher rate than small ones. The shared text was never the cause. It was a marker of a prolific publisher, and prolific publishers win.

That is the whole mechanism, and you can watch it hold the gradient up. The ecosystems with the highest share of duplicated text are precisely the ecosystems where a single creator (or a handful) mass-produces listings: House Flipper at 63.0% of its long-description mods duplicated, Aimlabs 50.5%, Dungeon Alchemist 43.6%, Counter-Strike 2 32.0%, Wallpaper Engine 26.4%, Brick Rigs 23.9%, DayZ 22.4%. Four of those six are the documented in-game-storefront ecosystems where the Workshop page is not even the discovery channel — the same games that top the visitor-counter gap — and they are exactly where a template blurb is easiest to reuse.

The paired test reverses it

To separate the blurb from the publisher, compare the same creator to themselves. Take the 23,548 creators who published, in the same game, at least one mod with duplicated text and at least one mod with text of their own. Same account, same ecosystem, same audience:

The gradient inverts. Once you hold the creator fixed, a shared description is worth 2.55 points less, not 8 points more. The pooled 1.29x advantage was never an advantage; it was the difference between prolific creators and everyone else, wearing the shared text as a costume. This is the same shape the short-title problem, the tag cliff and the name collision each had to work through, and the answer is the same: publish the residue, not the pooled number. Here the residue is negative and small. The honest headline is that copy-pasted text is mildly worse, and mostly it is a symptom rather than a cause.

One caveat matters for anyone reading the ladder as advice: the paired cohort is only the creators who did both, and a creator who mass-produces behind one blurb often has few or no unique-text mods to compare against. The comparison is real but it does not cover the extreme publishers who define the top band — which is exactly why the top band is composition and not signal. Both halves of that sentence are the finding.

What the shared text actually is

The interesting part is not the gradient. It is the text. Because a description shared by hundreds of mods is, by definition, boilerplate — and the boilerplate reveals how a large slice of the Workshop is actually operated. Grouping the shared descriptions by what they say:

The single largest shared description on the platform is used by 856 mods, and it is not a description at all — it is a warning:

Do NOT reupload ANY of my builds without crediting me in the description AND modifying it first!

The top member of that 856-strong group is a Brick Rigs build called The Grumobile (Remastered 2025) with 5,341 subscribers — we checked it against Steam's own API and it reports 5,346 current installs, and its description opens with that exact notice and then tells you the build "was made by Zombiehunter2.0. If the author is listed as something else, then this is a reupload." A mod of a Despicable Me car, carrying a copyright notice, listed under a warning that the listing itself might be stolen. That is the Workshop in one artefact.

The other classes are just as legible. A block of 845 mods in Age of Empires II share the disclaimer that they "may alter how Age of Empires II: HD Edition functions, and are not created by or endorsed by" the developers — a modder's own EULA, pasted 845 times. 523 mods carry a Patreon line. 488 mods in Counter-Strike 2 share a bilingual Russian-and-English server-redirect notice, and its near-twin (388 mods, one character different) proves how brittle an "identical text" test is: two creators, two notices, 876 mods, and a single typo in the Cyrillic keeps them in separate groups. 345 mods carry a French translation notice — "Pensez à vous abonner au mod original" — because translating one popular mod into a language and re-uploading it is its own cottage industry.

And then there is the class that carries the money: 1,495 Wallpaper Engine packs holding 43,441,969 subscriptions, averaging nearly 29,000 installs each, all sharing a description that is a raw steamusercontent image URL. That is a creator storing their preview images in the only field the Workshop gives them, pasted across an entire catalogue. It is the largest subscription total of any named class and its "description" is a line of machine text.

What it means

If you are a creator, the shared blurb is not a lever — it is a flag. The paired test says a mod with borrowed text does slightly worse than your own mods with original text, by about two and a half points of clearance. That is not a large penalty, and the honest reading is that the text is mostly neutral: what actually separates the bands is who is publishing. Reusing a house blurb across a catalogue is fine. Expecting it to lift a listing is not supported by the data.

If you are reading a Workshop statistic, this is the fourth field where the pooled average lies. Name, title length, description presence, and now description uniqueness have all produced a convincing pooled gap that a paired control dismantled or reversed. The pattern is now strong enough to treat as a rule: any per-listing attribute that a prolific creator can mass-assign will show a pooled advantage that is really a productivity effect. Check the paired number before you quote the pooled one — including, on this page, the first table.

If you are looking for the operationally interesting thing, it is the boilerplate itself. 51,180 descriptions are shared, and the ones big enough to matter are notices, licences, support links, translation disclaimers and template text. The Workshop's description field, on a meaningful fraction of the catalogue, is being used as a place to put a warning, a link, or a filename — not a description. The blank-page problem counted the mods that wrote nothing. This is the other half: the mods that wrote something, just not about themselves.

Related reading: the one-subscriber floor on the 654,035 mods parked at a single install, the discovery lottery on the 94% of active mods that get zero views in a week, and the tag cliff on the 1.8 million mods carrying no labels at all.

Method: 5,990,283 Steam Workshop items tracked by ModSignal as of 30 September 2026 across 392 games — every item that is is_active and carries at least one subscription, holding 12,453,944,440 subscriptions in total. Descriptions are compared as md5(lower(trim(description_short))) across the whole platform, not within a game, so identical text in two different ecosystems counts as a shared description; the mods compared are the 2,006,226 whose description is at least 100 characters long, which excludes short blurbs where a collision is coincidental rather than structural — including the sub-100-character population would raise every duplicated count and is deliberately excluded. "Reached 1,000 subscribers" means subscriptions_current >= 1000 at a single consistent read; it is a point-in-time threshold, not a tracked trajectory. The paired test compares creators who published both a shared-text and a unique-text mod in the same game and reports the aggregate clearance rate on each side; it is an unweighted comparison of totals, not a per-creator median, and it under-covers the extreme mass-publishers whose catalogues contain few unique-text mods — which is precisely why the pooled top band is composition. Class buckets are keyword regexes over the shared text (Latin, Cyrillic and CJK terms) and are a lower bound: the 90.0% "uncategorised" bucket contains many more notices, guides and templates that the patterns did not catch, so read the named classes as examples, not as a partition. The class table's subscription totals are the sums over the mods in each class, not per-creator. The cited Brick Rigs mod was live-verified against Steam's ISteamRemoteStorage/GetPublishedFileDetails (pfid 3441042568, 5,346 subscriptions, 2,029 unique visitors) and its description text matched. Subscriptions are the last reported values and are never decremented by Steam, so a delisted mod, if any, would still be counted. Counters move continuously; a re-run will differ by a handful of items.