Take a query like "Find me eco-friendly skincare creators in Germany under 50K followers". A person reads that in a second and understands the intent behind it: the niche, the audience, the region, the size, and the implicit values that go with "eco-friendly". A keyword search reads none of it. AI Search closes that gap. It interprets the meaning of the query rather than its literal words, infers the synonyms and adjacent terms creators in this space actually use, and translates everything into a structured search the index can answer.
You Describe the Creator You Want
Your input can be between 3 and 150 characters and can describe a niche, an audience, a content style, or a specific username. The query goes through a quick safety check before it is passed to the AI model.
What the System Accepts and Rejects
AI Expands Your Brief Into a Structured Query
A handful of words carries meaning that a keyword search cannot interpret. AI reads the input and generates terms the index can match against: hashtags at three levels of specificity, plus a short list of phrases drawn from how creators in this space describe themselves. The result is cached so the same query is not re-processed.
What AI Returns for "Travel Creators in Lisbon"
Easy tags produce the broadest match. Hard tags narrow it. Medium tags cover the synonyms and alternate spellings creators commonly use. Without all three layers, the system would either return millions of loose matches or miss creators who used a different word for the same concept.
We Search the Index With a Hybrid Query
The expanded query runs against our creator index. Two query types run in parallel: a semantic similarity search across captions, biography, and full name; and a phrase match against bio and full name with a small word slop to catch near-matches. A creator needs to satisfy only one branch to be considered a candidate.
How the Two Branches Blend Into One Query
Expanded hashtag set used as a similarity vector across captions, biography, and full name. At least 8% of terms must match. Up to 70 query terms considered.
strength: meaning & contextEach of the 5 AI-generated phrases runs against bio and full name with a small word slop. Boost factor 1.75. At least one phrase must match.
strength: exact intent signalsA creator only needs to satisfy one branch to be considered. Matches that hit both branches surface higher because two independent signals agree.
We Rank by Content Match Plus Creator Quality
A creator can match the query and still be a low-quality account, so content match alone is not used for ranking. The final rank combines content match (70%) with the Creator Quality Index (30%), then adds bonuses for high-priority countries and verified creators in healthy follower bands.
How the Final Score Comes Together
This is the same blend used in Lookalikes. Content match receives the larger share because it is the strongest signal that a creator covers your topic, while CQI prevents low-quality accounts from rising in the rankings simply because their bio matches a few keywords.
Read how CQI works →We Apply the Same Quality Filters as the Rest of the Platform
Before results come back, we apply the same trust filters used everywhere else. English-speaking creators are prioritized, and your account-type preference is honored if you set one (creator, business, or both).
Filters Applied to Every Result Set
Bio, captions, and content are used to detect language. Creators producing in English rise; non-English creators are deprioritized.
If your search is set to creator, business, or both, the result set respects that preference.
Spam, inactive, and fake-follower accounts are excluded by the same checks Lookalikes and Matching use.
End-to-End Example
A short, word-only query enters at the top of the funnel. AI expands it into the terms creators actually use, the index narrows that into a candidate pool, CQI reorders what remains, and a ranked list of creators comes out the other end.
The first three creators all have travel as their primary niche, with matching profile data. BorisLifeLog is the contrast case: he occasionally posts travel vlogs, so his bio contains some travel terms, but his primary niche is lifestyle. The 25-point drop identifies the cutoff between creators who are genuinely about this niche and those who are only loosely related.
How We Validate AI Search 
AI Search outputs a ranked list rather than a binary answer, so the validation approach mirrors Lookalikes: two tracks run in parallel, with AI providing scale and human reviewers catching cases AI misses.
From Plain English to Ranked Creators
A brief is entered in plain English. AI translates it into a structured query the index can answer. Hybrid search runs. CQI re-ranks the results. The output is a list of creators that fit the brief, returned in seconds without requiring a single boolean filter.