Subscriber acquisition across the entire subscription economy has fallen by a third since 2021. What replaced it is not a bigger audience. It is a narrower one.
Two numbers from the same report, published four months apart, explain most of what has happened to paid content in the last two years.
The first: Instagram accounts with more than a million followers pull a median of 6,678 likes per post. The second: accounts between 10,000 and 50,000 followers pull 91. That gap is roughly seventy-three to one, and it is exactly what you would expect.
The engagement rates run the other way. Nano-influencers — 1,000 to 10,000 followers — post at a 1.78% engagement rate, according to HypeAuditor’s 2026 State of Influencer Marketing report as reported by eMarketer. Micro-influencers sit at 0.54%. The million-plus accounts sit at 0.33%, roughly a fifth of the nano rate.
One tier has the audience. The other has the attention.
The acquisition problem nobody solved
This would be an academic distinction if attracting new subscribers were still cheap. It is not.
Recurly’s ninth annual State of Subscriptions report, built on data from 67 million subscribers, found that subscriber acquisition rates fell to 2.8% — down from 4.1% in 2021. Free trial conversion fell harder, from 46% to 33%. The funnel that worked in 2021 leaks about a third more than it used to, at every stage.
“Even as customer acquisition declines, retention is increasing,” Recurly CEO Joe Rohrlich said in the release. “A clear sign that consumers want to find and stick with subscriptions that best fit in with their lives.”
That last clause is the whole thing. Best fit in with their lives is a polite way of saying: people are no longer collecting subscriptions. They are keeping the two or three that earn it and dropping the rest.
What creators did about it
The response was not to chase more followers. It was to chase fewer.
The 2026 Circle Community Trends Report found that 69% of creators now treat member outcomes — whether the person actually got what they came for — as their main retention driver, ahead of reach. More striking: 39% say they have deliberately de-prioritised growth in favour of higher-touch, higher-ticket offerings. Another 12% cap membership size outright, to protect the experience.
Those are creators voluntarily turning down subscribers. Two years ago that would have read as a rounding error or a rationalisation. Now it is a strategy with a name.
The logic holds up against the engagement data. If a nano-influencer converts at five times the rate of a celebrity account, the celebrity account is not five times more valuable — it is more expensive per unit of attention, and its audience is broader in exactly the way that makes conversion harder.
The category problem
Here is where paid content diverges from the rest of the subscription economy, and where the honest version gets uncomfortable.
Specialisation works. A creator who is unambiguously the thing a specific audience came for — a particular look, a particular register, a particular kind of interaction — converts better than a generalist, for the same reason the nano-influencer does. The audience is smaller and it is correct.
But specialisation only pays if the audience can find the specialist.
OnlyFans, the largest platform in the category, has no browse-by-category view. No filter, no directory, no recommendation surface. Its internal search works if you already know the username you are looking for, which means it works for confirming a name and nothing else. A platform architecture built around creator privacy has the side effect of making category discovery structurally impossible from the inside.
So a fan who wants a specific niche — say, busty creators on OnlyFans — has to leave the platform to find one. Outside indexes assemble that view from public profile data, sorted by category. The caveat matters: an index can only see creators who are already visible. The quiet accounts, the ones who never promoted anywhere, are under-represented everywhere including there. It is a workaround for a missing feature, not a census.
What the surgery data says about the same period
There is a parallel worth noting, because it cuts against the obvious assumption.
The American Society of Plastic Surgeons’ 2025 procedural statistics, released this September, show cosmetic procedures up 7% year over year. Breast lifts rose 8%. Breast augmentation stayed in the top three by demand — but the share of the US population getting implants has held at 0.09% since 1999, unchanged across twenty-six years.
“Today’s patients are increasingly seeking authentic restoration rather than transformation,” ASPS President C. Bob Basu said of the findings.
If the category has grown online while the underlying population has not moved, the growth is not in supply. It is in visibility and sorting — more of the people who were always there, findable by the people who were always looking.
What this does not settle
The engagement figures are Instagram and TikTok data, not subscription data. Nobody has published a clean study mapping engagement rate to paid conversion on adult platforms specifically, and it would be a stretch to assume the ratios carry across intact.
Recurly’s numbers cover the subscription economy broadly — streaming, retail, SaaS, media. Adult content sits inside that universe but has its own churn dynamics that the aggregate hides.
And the Circle findings describe creators who responded to a survey about community building. Survey populations skew toward the deliberate. The creator who is not thinking strategically about retention is also not filling out a report about it.
What holds across all three sources is narrower and duller than any single headline: acquisition got harder, attention concentrated in smaller audiences, and the operators who noticed early stopped optimising for size. The rest is still being counted.