Data studies

By OnlyLookup Research

424 profiles published in the current catalog. Research figures below describe their dated snapshots.

How many OnlyFans categories are there — and how many are actually empty

Every OnlyFans directory sells the same promise: browse by category and find someone. The promise rests on a number that nobody audits — how many categories the directory has, and how many of them have anyone in them.

OnlyFans publishes no category taxonomy of its own. There is no canonical list to check against, which means each directory invents one — by reading public profiles, by harvesting the label sets other directories publish, or both. We run one of those pipelines and do both, so the figures below are our own full census of it: 723 category candidates, counted on 14 October 2026, with every derived number defined where it appears. The measurement queries sit in the pull request behind this article.

Where do 723 category candidates even come from?

Not from a plan, and mostly not from profiles. 78 of them were written by hand as the catalog's working taxonomy. The other 645 came out of harvesting: we read the labels that other public directories attach to their own listings, and any label that turns up on at least 30 profiles in our catalog becomes a candidate for a human to accept or drop. Reading a profile's own text is a separate job in our pipeline, and it cannot invent a category — it only files people under categories somebody has already accepted.

OnlyLookup catalog, all 723 rows of the category table, measured 14 October 2026, 06:51 UTC.

The split is the whole story. A proposal costs 30 profiles carrying the same harvested label, which sounds like work and is not: another directory did the labelling, and a labelled profile arrives wearing a lot of labels at once — a median of 22 of them, 54 on average, 631 on the heaviest. A published page costs a judgement about whether the phrase is a real niche, whether it duplicates a page we already have, and whether enough people fit it to be worth browsing. 635 of the 645 proposals arrived in a single crawl pass on 12 October, which tells you how fast the cheap side of that trade compounds.

How many of those categories have anyone in them?

Count profiles the way a visitor experiences them: distinct people with a live, publicly visible page, not rows in a join table. That distinction is not cosmetic — across live profiles our assignment table holds 1 703 rows for 1 326 distinct profile-to-category pairs, because the same assignment can arrive from more than one source. Counting rows would overstate the live catalog by 28.4% in total, and the error is not spread evenly: 14 of the 21 categories listed further down are not inflated at all, while the largest one would read 88.7% too high.

OnlyLookup catalog, count(distinct creator_id) per category among 321 profiles with status live, measured 14 October 2026, 06:51 UTC.

All 645 harvested proposals fall in the top row, and nothing automatic will move them: filing a profile under a category draws only on categories a human has accepted, so a proposal holds 0 profiles until it is accepted. Today all 645 hold 0, suppressed and non-live rows included. The 9 remaining categories with nothing in them are all hand-written and all published.

Which is the disclosure this article owes you, since it cuts against us. Of our 75 published category pages, 9 show no live profile at all today, and all 9 are US state pages. Our location data rests on what creators write about themselves rather than on anything verified, it is incomplete, and it is being rebuilt — which is why this article links to no state page and would not cite one as evidence. No interest or price page is empty.

What does the proposed side of a taxonomy actually look like?

This is the part no directory shows you, and it explains why category counts are not comparable across sites. Below is every one of the 645 proposals sorted into what it is.

OnlyLookup catalog, 645 category rows with status suggested, measured 14 October 2026, 09:29 UTC. Buckets are assigned in the order shown, so each proposal is counted once. Every test here compares whole words rather than substrings, because our slugs are word-sorted: across these rows the words of a 2-word state name never once land side by side in the order the name needs, so a substring test cannot see them at all. The place bucket is slugs carrying both states and united as tokens; 406 of the 645 carry the words of a US state name, and all 406 of those sit inside the place bucket. 22 of the 723 candidates are age-shaped on the same definition once overlapping buckets are ignored, and 0 of those 22 are published.

The largest bucket is not a niche at all. 411 of the 645 are geography, and 2 separate things make them look the way they do. The labels we harvested are location paths out of another directory's index, stored here exactly as we found them — united-states-oklahoma-tulsa. The scrambling is ours: our slug builder lowercases a label, splits it into words and sorts them alphabetically, which turns that path into oklahoma-states-tulsa-united; labels that reduce to the same sorted form collapse into a single candidate. So the garbled look is our canonical form, not a sign that somebody else reordered anything.

They are also far thinner than 411 suggests, and this is the number to keep. Count the carriers the way the candidate itself was built — every profile holding that label in any spelling, collapsed into the one canonical form above: 249 of the candidates rest on the same 69 profiles, another 125 on the same 52, and all 411 together rest on 210 profiles, of which 85 are live, measured 14 October 2026, 07:31 UTC. That is the size of the thing — 411 candidate pages over 85 live people.

One more detail from the same rows, because it shows what a category count is worth. Every proposal carries a size and 5 example handles in its metadata, and both are ours, not the source's: our script counts the profiles in our own catalog that carry the label. That makes it a measurement of a batch rather than of a category. Across the 411 place candidates there are only 14 distinct sizes and 8 distinct example lists — 274 of them repeat one identical list of handles and 131 repeat another — because labelled profiles arrive wearing the whole label set at once. We store the number and we publish nothing on the strength of it. If a directory's category page tells you it holds 52 profiles, the question worth asking is what the 52 counts: people you can browse there today, or profiles that once carried a label.

How does a category get filled in the first place?

Almost entirely from what creators write about themselves. Automatic filing has 2 routes: read the profile's own public text, or inherit the assignment from a cross-reference against another public catalog. An editor can also assign a category by hand, and no live pair in the table below came that way today. Among live profiles the 2 automatic routes produce very different amounts of work.

OnlyLookup catalog, 1 326 distinct pairs across 321 live profiles, measured 14 October 2026, 06:51 UTC.

Read the bottom row. Across the entire live catalog, exactly 2 category placements exist that the profile's own words would not have produced on their own. Put that next to the section above and the shape of the study is plain: harvesting other directories is where our category names come from in bulk, and it is almost nowhere in the filing. That is also why totals do not compare across sites — a total counts candidate names, which travel between directories as easily as ours travelled to us, while the populated part of a catalog is built out of what profiles say about themselves. 317 of the 321 live profiles carry at least 1 category, with a median of 4 each and a maximum of 20.

There is also a third state between proposed and published that directories rarely expose. 3 interest categories in our taxonomy are accepted as real and still have no page, because they hold 1, 4 and 5 live profiles — real niches, too thin to browse. They sit in the taxonomy waiting for population, not for a decision.

So what is the bar for publishing a category, and who clears it?

10 live profiles. Below that a category page stops being a browsing surface and becomes a dead end: it answers a query with a list too short to choose from, and both the visitor and the search engine are right to leave. 4 of our published interest pages sit under the mark today and carry a no-index instruction rather than a padded list.

26 of the 723 candidates clear it. 21 are interest or price categories, and 5 are state pages we would not put weight on for the reason given above.

OnlyLookup catalog, count(distinct creator_id) among live profiles, measured 14 October 2026, 06:51 UTC. All 21 are published and browsable from the category hub.

Notice the shape of that list: the top category holds more live profiles than the bottom 10 of the 21 combined — 168 against 156. Category spaces here are not evenly populated and never will be: they are built from how people describe themselves, and that happens in a small number of common ways. A directory with 700 categories and a directory with 25 can be showing you the same 300 people.

How do you read a category count on any OnlyFans directory?

4 checks, none of which need access to anything private.

1. Open the category list and click the least popular entry you can find. If it returns an empty or near-empty list, the headline count is a count of pages, not of populated pages.
2. Look for a stated bar. A directory that will tell you its minimum profile count per page has thought about the problem; one that quotes only a total has not.
3. Check whether the entries read like phrases a person would write about themselves, or like an index somebody copied: location paths, word-sorted variants of one phrase, and the same phrase in 3 spellings are marks of a harvested label set rather than of profiles that were read. Our own candidate list holds 411 location paths, none of them published.
4. Compare 2 directories on the same category, not on their totals. The overlap is usually far higher than the difference in category counts suggests.

And the honest limit on everything above: these are our numbers about our pipeline, measured on one day. They describe what reading public OnlyFans profiles and other people's lists at scale produces, and they do not describe OnlyFans. Our methodology page sets out which fields we read, how often we re-read them, the 10-profile indexing bar, and how to be removed. Any figure here older than 90 days gets re-measured or deleted.

Want to see the populated end of this rather than the count? 21 interest and price categories clear the bar today. Browse the catalog.

Published October 14, 2026 · Updated October 14, 2026