No number has moved through business writing lately the way one MIT finding has: that 95% of enterprise AI pilots return nothing. You've met it, reworded, on a dozen blogs that share nothing but the stat. Most of them skip the part where the finding is contested and its methodology has been challenged in the open. Any one of those posts looks well sourced. Read them side by side and you're looking at a single number under a dozen different citations.
The pattern has a name: citation monoculture, a field whose numbers keep coming from the same few places. To find out how common it is, this study traced 1,006 blog citations across 46 SaaS domains to their final destinations. Common undersells it. The monoculture is the field's default state.
Citation Monoculture Defined and Measured
A citation monoculture is a body of writing whose numbers resolve to the same few destinations. Each post inside it reads as carefully sourced. Merge the reference lists across the field, though, and a short roster of names emerges, with hundreds of pages resting on a handful of shared origins.
The corpus is the one behind the study of citation chain depth: 961 posts on 46 SaaS blog domains, with the same 1,006 verified linked citations, each followed until its reference trail ran out.
That study ran vertical, following each citation down its chain to see how far it travels before the trail ends. This one runs horizontal across the endpoints. The claim extraction pipeline had already carried every citation to its terminal node, so the only new step is grouping: take where all 1,006 chains end, resolve each endpoint to its root domain, and count.
Forty-four of the 46 corpus domains carry at least one traced citation, so the breadth figures below divide by 44 and the share figures by the full 1,006. I re-ran each count against the frozen dataset with 95% Wilson score intervals before trusting a headline, and every number in this post comes back identical from a script anyone can run.
Set Your Source Diversity Bar
Before the data, write down your own rule. Most sourcing standards are written as a count: a stat becomes load-bearing once two links back it, or three. A count answers how many sources. It says nothing about the question underneath: how many independent sources. This study lives in the distance between those two answers.
Set your bar now, before the field's habits have a chance to argue with it.
Keep your answer close. Whether you can meet it depends on the pool the entire field draws from.
Whatever bar you hold, you wrote it in sources, and a link tells you nothing about the origin underneath it. A rule of "at least two" passes as soon as it sees two links pointing at two different pages, with no view of whether both pages rest on one dataset. The question that matters sits beyond your team's discipline, in the shared pool itself, and in how small that pool turns out to be.
Ten Domains Absorb Almost a Quarter of Citations
Group all 1,006 citations by the root domain they resolve to and you get 516 distinct destinations. The long tail is real; hundreds of domains collect one or two citations apiece. The finding lives in how the weight sits across them.
The three busiest destinations absorb 12.8% of every citation in the set. The 10 most-cited carry 23.1%. Ten domains out of 516 hold almost a quarter of an industry's sourcing, and the remaining 506 divide what's left.
Fold the whole distribution into a single figure and the skew sharpens. Run the standard concentration math across all 516 domains and the field's citations carry weight the way roughly 104 equally weighted sources would. Bibliometrics has documented this kind of skew in academic citation networks for years. Where a whole commercial field's numbers actually land, and how few destinations bear the load, stayed unmeasured until now.
The names at the top are the research houses a whole industry reaches for without thinking. Statista is where 6.4% of all traced citations end, and Gartner and McKinsey each sit near 3%. Nothing in that indicts the houses themselves. Each is a source many sites lean on at once, and that shared reliance is the field's own habit.
Breadth and Depth in a Citation Monoculture
Share of citations is half the picture, and the half most likely to mislead. One site can cite a favorite source 20 times and lift it up the rankings while every other publisher ignores it. The number that exposes a monoculture is reach: how many separate sites cite a source at all.
Gartner shows up in 15 of the 44 citing sites, McKinsey in 15, Statista in 14. Roughly a third of the independent SaaS blogs in the corpus lean on Gartner, the same share leans on McKinsey, and nobody coordinated any of it. These are the names writers pick when a stat needs to sound settled.
That reach is breadth, and breadth is a shared dependency: 15 sites leaning on one house share a single point of failure.
Statista shows a second shape. It reaches 14 sites but carries 64 citations, the most in the corpus, so the sites that use it return to it over and over. That is depth, one origin cited repeatedly inside a single blog, and it concentrates risk in its own way, because a post that cites the same source four times has built four claims on the same foundation.
Both shapes belong to the monoculture. A sourcing policy that watches only a most-cited list catches neither cleanly, since the wide-and-shallow pattern and the narrow-and-deep one attack a content library from opposite sides.
Shared Sources Fail Together
Concentration turns into damage through a dull mechanism. A house re-bases its figure. A newer edition ships. A stat gets retracted, or the report goes behind a login. Every post that leaned on the old number becomes wrong at the same moment, and no link checker notices, because every link still resolves.
The URL still answers 200 and the page renders. The number on it has moved away from the one the post claimed. The link stays blue, the citation looks healthy, and a correction that should reach a dozen posts reaches zero of them.
Posts written independently should fail independently. A monoculture removes the independence: one change upstream and the field's copy goes stale in unison.
This exposure is what LiquiChart's living content infrastructure watches. A Monitored Page tracks what a source page says, beyond whether its URL still loads. When the number on it moves, staleness flows to every claim of yours that cited it, and the Pulse timeline flags the source as a hub at risk, one alert for the dependency behind all of them.
Living Content drafts the correction into a review queue for your approval; nothing rewrites a published page on its own. On Visionary, a Canonical Source of Truth binds every place a number appears to one authoritative claim, so a single approval reaches all of them. Watching sources at scale and binding a number across posts are paid-tier capabilities.
I watched a number sit in a post for a year after the study behind it published a revision. Nobody involved was careless. The link never broke, and an unbroken link is where most sourcing checks end.
The Authority Objection
There's a fair objection here: the big houses earned this. Gartner earns its citations by being authoritative, and a whole field converging on the same sources might simply mean the field has settled on the best ones.
Two answers. First, a source everyone already cites adds no information gain. Cite the same house as the 10 posts ranking beside you and you've handed the reader a reference every competitor already carries. The same concentration explains why the same few houses win the citation leaderboard in AI answers: models favor the sources they see everywhere, and every new citation makes them more visible still.
Second, authority amplifies the exposure this study measures. The more a house gets treated as the safe citation, the more posts anchor to it alone, and the more posts break together when its number moves. Shared trust becomes shared exposure.
Own the Number
Seeing the concentration, your first instinct will be to diversify: more sources, wider spread.
The data explains why that fails. The next source you reach for is the one the field already trusts, so an added citation lands on a destination a competitor already used far more often than on genuinely new ground. A citation count rewards conformity: it rises each time you choose the source everyone chooses, and it has no term for independence. A post with 15 citations can share 14 of them with every post it competes against.
Leaving the monoculture takes a different move: own the number. Publish a measurement no competitor holds, a poll you ran with your own audience, a chart from data you gathered, a claim you can defend because you made the measurement. A number you measured starts from an origin nobody else in the field holds, so when an outside house re-bases, your claim stays where you put it. Starting requires no research budget, only a question worth asking the audience you already have and a vague industry stat worth replacing with your own data.
Before changing anything, you can measure your own exposure. Run the Content Health Scanner on your posts and it sorts every claim by origin, owned versus borrowed, and flags the numbers of yours that lead back to the same handful of outside sources; it's free, no account needed. It reads your content alone, so what it maps is your share of the field's habit: how much of your credibility rests on numbers someone else controls.
Your citation list reads as proof of diligence. Laid over the field's, it becomes a map of the dependencies you share with everyone ranking beside you, and every load-bearing figure on it that you borrowed can be pulled out from under you by a publisher who has never heard of you. A borrowed number moves at its owner's convenience. A number you own moves only when you move it.