How does KOLens know a TikTok creator's category?

It infers it with fast, deterministic heuristics — no LLM, no guesswork you cannot audit. First it scans the bio for niche keywords and emoji (💄 makeup, 🏋️ gym, 🎮 gaming). If the bio is thin or emoji-only, it pools the captions and hashtags of the creator's recent videos and matches those. The result is one of 20 fixed category slugs, plus a category_source field telling you whether the bio or the hashtags produced it.

Why not just use the category TikTok provides?

TikTok, X and Douyin do not provide one. Only Instagram business accounts expose a self-declared category, so before this feature the category filter returned almost nothing for the other platforms — the column was simply null. KOLens' inferred category fills that gap so niche filtering works everywhere, not only on Instagram.

What categories can I filter by?

Twenty niches today: beauty, fashion, fitness, food, tech, gaming, finance, travel, parenting, pets, home, automotive, music, dance, comedy, art, education, sports, wellness and photography. The list is a controlled vocabulary so the filter is predictable; new niches are a one-line addition on our side as the dataset grows.

How accurate is the inferred category?

It is a high-precision heuristic, not a classifier you have to trust blindly. Each match is keyword-anchored and the category_source field shows the evidence (bio_heuristic or hashtag_heuristic), so you can challenge any row. When a creator's bio spans two niches, the higher-priority commercial niche wins deterministically — a 'skincare & wellness' bio resolves to beauty, not wellness — so results are stable across runs.

Can I combine category with other filters?

Yes — that is the point. Stack category with the country filter, a follower-count band, and a minimum engagement rate to go from a mixed 10,000-creator database to a single-niche, single-market, outreach-ready cohort in one request. Via MCP: list_kols(category='beauty', country='US', min_engagement_rate=0.04, require_email=True).

Does the category also improve similar-creator search?

Yes. find_similar_creators uses same-niche as one of its ranking signals. Previously that signal was dead on TikTok because the category was null; now it falls back to the inferred category, so 'find creators like this one' surfaces genuine same-niche peers on every platform.

All posts

May 29, 2026·KOLens teamTikTokInfluencer marketingCreator discoveryNiche targeting

Find TikTok creators by category (niche filtering)

Filter your whole creator database down to one niche in a single click — beauty, fitness, tech, finance — across TikTok, X and Douyin, where the platform itself never tells you the category.

Quick answer

KOLens now tags every scraped creator with a content category — one of 20 niche slugs like beauty, fitness, tech or gaming. The /kols category filter and the list_kols(category=…) MCP tool match it even on TikTok, X and Douyin, where the platform never declares a category. KOLens infers it from the bio first, then recent-video hashtags.

How do you find TikTok creators in one specific niche? It sounds trivial — surely you just filter by category — but until now you could not, and the reason is structural. TikTok, X and Douyin do not expose a self-declared category for a creator the way Instagram business accounts do. So a discovery database built from those platforms has a category column that is almost entirely null, and a category filter built on top of it silently returns nothing. You are left scrolling a mixed feed of beauty, gaming and finance creators looking for the ten that fit your brand.

This is the same shape of gap we closed for country targeting: the filter existed, but the data behind it was empty. The fix is the same too — stop waiting for the platform to hand us a field it never will, and infer it ourselves from signals the creator does give us. This article explains how KOLens now categorises every creator, what the 20 niches are, and the exact workflow to pull a single-niche shortlist.

How KOLens infers a creator's category

The categoriser runs automatically the moment a creator is scraped — no extra step, no backfill button to remember. It is a fast, deterministic heuristic (no LLM call, nothing you cannot audit) that reads two signals in priority order:

Bio keywords and emoji. The strongest signal. A bio reading "makeup artist 💄 honest skincare reviews" resolves to beauty; "powerlifting coach 🏋️" resolves to fitness; "iOS developer · SaaS founder" resolves to tech. Keywords are word-boundaried so "art" does not match inside "start" and "cat" does not match inside "category".
Recent-video captions and hashtags. The fallback for creators with a thin or emoji-only bio. KOLens pools the text of the creator's recent videos and matches the same keyword set — so a creator whose bio is just "✌️" but whose videos are all #skincare #beauty still lands in beauty.

Every categorised row also carries a category_source field — bio_heuristic or hashtag_heuristic — so you can see the evidence behind the label and challenge a false positive. When a bio spans two niches, the higher-priority commercial niche wins deterministically (a "skincare & wellness" bio is beauty, not wellness), which keeps results stable run to run.

Why deterministic, not an LLM classifier

A no-LLM heuristic is auditable, free, and fast enough to categorise the entire back-catalogue in seconds. You get a label you can trace to a specific keyword instead of a black-box guess — and the filter behaves identically every time you run it.

The 20 niches

The category filter matches a controlled vocabulary, so results are predictable. The current slugs:

Lifestyle & appearance: beauty, fashion, fitness, wellness, food
Interest & hobby: gaming, travel, pets, automotive, home, photography
Creative: music, dance, comedy, art
Knowledge & commerce: tech, finance, education, parenting, sports

That covers the overwhelming majority of brand-relevant creator marketing. The list is designed to grow — adding a niche is a one-line change on our side — so expect it to widen as the dataset does.

The workflow: niche + market + quality in one query

Category on its own is useful, but the real leverage is stacking it with the filters you already use. The goal is to go from a mixed, many-thousand-creator database to a single-niche, single-market, outreach-ready cohort in one pass.

1
Pick the niche.
On /kols set the category filter to the niche you sell into — say beauty. Every creator KOLens has ever scraped in that niche, across platforms, collapses into view.
2
Add the market.
Stack the country filter (e.g. US). KOLens matches the creator's self-declared country or its inferred country, so US-targeting brands drop everyone else even when the scraper never captured a location.
3
Set a quality floor.
Add a minimum engagement rate of 4% (KOLens' own quality floor) and a follower band that fits your budget — e.g. 10k–100k for micro creators. Toggle require email so every survivor is contactable.
4
Save the cohort to a KOL list.
Bulk-select the survivors and save them to a KOL list to run them through the outreach pipeline. You now have a single-niche, single-market shortlist instead of a mixed feed.

Doing it from Claude (MCP)

If you drive KOLens from an AI agent over MCP, the same filter is one argument on list_kols. The whole recipe above is a single call:

list_kols(category="beauty", country="US", min_engagement_rate=0.04, require_email=True, limit=50)

Each returned row carries category_inferred and category_source alongside the usual stats and the pre-computed llm digest, so the agent can see why each creator matched the niche and reason over the cohort without a follow-up call.

A bonus: better similar-creator search

The inferred category does double duty. KOLens' find_similar_creators ranks candidates partly on whether they share the seed creator's niche. On TikTok that signal used to be dead — the category was null, so same-niche matching never fired. With an inferred category on every creator, "find more creators like this one" now surfaces genuine same-niche peers on every platform, not just Instagram.

Why this beats the manual alternative

Tool	Gap	KOLens
Scrolling a mixed keyword feed by eye	Beauty, gaming and finance creators interleaved; minutes per creator	One category filter collapses the database to a single niche instantly
Relying on the platform's category field	Null on TikTok / X / Douyin — filter returns nothing	Inferred from bio + hashtags, so it works on every platform
Tagging creators by hand in a spreadsheet	Stale the moment you scrape more; no audit trail	Auto-tagged on every scrape with a traceable category_source
An LLM 'guess the niche' prompt per creator	Slow, costs tokens, non-deterministic across runs	Deterministic keyword heuristic — free, instant, repeatable

Next step

Niche filtering is live now — no setup, no backfill. Open /kols, set the category filter, and stack it with country and an engagement floor to pull your first single-niche shortlist. New accounts come with free credits — enough to run a niche-targeted search end to end.

Find TikTok creators by country — the market-targeting filter that pairs with category.
How to find micro influencers on TikTok — the follower-band + engagement recipe to stack on top.
How to find TikTok KOLs by keyword — the discovery workflow that fills the database in the first place.

READY?

Try it now — 50 free credits on signup.

Filter creators by niche free

Frequently asked

How does KOLens know a TikTok creator's category?: It infers it with fast, deterministic heuristics — no LLM, no guesswork you cannot audit. First it scans the bio for niche keywords and emoji (💄 makeup, 🏋️ gym, 🎮 gaming). If the bio is thin or emoji-only, it pools the captions and hashtags of the creator's recent videos and matches those. The result is one of 20 fixed category slugs, plus a category_source field telling you whether the bio or the hashtags produced it.
Why not just use the category TikTok provides?: TikTok, X and Douyin do not provide one. Only Instagram business accounts expose a self-declared category, so before this feature the category filter returned almost nothing for the other platforms — the column was simply null. KOLens' inferred category fills that gap so niche filtering works everywhere, not only on Instagram.
What categories can I filter by?: Twenty niches today: beauty, fashion, fitness, food, tech, gaming, finance, travel, parenting, pets, home, automotive, music, dance, comedy, art, education, sports, wellness and photography. The list is a controlled vocabulary so the filter is predictable; new niches are a one-line addition on our side as the dataset grows.
How accurate is the inferred category?: It is a high-precision heuristic, not a classifier you have to trust blindly. Each match is keyword-anchored and the category_source field shows the evidence (bio_heuristic or hashtag_heuristic), so you can challenge any row. When a creator's bio spans two niches, the higher-priority commercial niche wins deterministically — a 'skincare & wellness' bio resolves to beauty, not wellness — so results are stable across runs.
Can I combine category with other filters?: Yes — that is the point. Stack category with the country filter, a follower-count band, and a minimum engagement rate to go from a mixed 10,000-creator database to a single-niche, single-market, outreach-ready cohort in one request. Via MCP: list_kols(category='beauty', country='US', min_engagement_rate=0.04, require_email=True).
Does the category also improve similar-creator search?: Yes. find_similar_creators uses same-niche as one of its ranking signals. Previously that signal was dead on TikTok because the category was null; now it falls back to the inferred category, so 'find creators like this one' surfaces genuine same-niche peers on every platform.