Influencer Finder: Score Your Creator List Against a Brief

Bring your creator list and score it against your campaign brief. This Actor does not search for creators — you bring the list: a dataset or CSV from a YouTube, Instagram or TikTok scraper, or your own spreadsheet. Instagram and TikTok creators need their bio and posts in that data (from a profile or post scraper); a bare Instagram or TikTok handle comes back unclear. YouTube channels typed in are read from their public About page. Then, per creator, one row with a verdict against your brief and the exact sentence it rests on, an AI fit review with the audience and risk flags (paid plans), and one ranked shortlist of the best fits with the reasons the rest were excluded.

Free plan vs paid plans

	Free plan	Paid plans
Creators scored per run	the first 1 in full; the rest as counts only (how many strong or possible, weak, excluded, unclear — not who)	up to 5,000 in full (maxCreators, default 500)
Rule verdict: mustHave keywords, exclude keywords, follower range — each match with the quoted sentence and where it is (bio, post n)	✅	✅
AI fit verdict vs your brief: strong / possible / weak / unclear, quotes from the creator's own text, stated audience, topics, risk flags	🔒 locked — the preview row counts how many creators a paid run could send	✅ up to aiMaxItems (default 200)
Ranked shortlist row: up to N strong/possible fits + why the others were excluded	🔒 locked — the preview row counts the candidates by the rules	✅ one per run
Contact fields you brought in (e-mail, phone, website)	carried over as given	carried over as given

The free run ends with a free-plan-preview row with numbers counted in that run, each saying what it counts: the creators read, the ones a paid run takes (maxCreators), the rule counts over them, how many could go to the AI review (requests — a review that fails our checks is not a verdict and is not charged), the shortlist candidates and the most it lists, and an estimate of this Actor's events for a paid run (your Apify plan and the scrapers that made your data are not in it). The free plan uses your own Apify credits for its run-start fee and the one creator it judges.

What the paid AI layer adds — a fixed example

Illustrative only: made-up sample data, not one of your creators and not a measured model answer. Every free run's preview row carries the same example (example).

{
  "handle": "trailcook.example",
  "ruleFit": { "verdict": "possible", "reasons": [{ "term": "camping", "quote": "Camping recipes for two-day hikes.", "source": "bio" }], "mustHaveMissing": ["tent"] },
  "fit": { "status": "reviewed", "fit": "strong",
    "reasons": [{ "quote": "Testing a new backpacking stove on the ridge", "why": "Reviews the stove category the brief names." }],
    "audience": "beginner backpackers", "topics": ["camp cooking", "backpacking"], "flags": ["sponsorship-mentioned"] }
}

The rules found one keyword (possible); the AI review read a post the keywords missed and rated it strong, with the quote, the stated audience and a sponsorship flag.

Paid-plan discounts

Event	FREE	BRONZE (−20%)	SILVER (−35%)	GOLD and above (−48%)
Run started (once per run)	$0.005	$0.004	$0.00325	$0.0026
Creator checked (rule verdict, one per creator)	$0.005	$0.004	$0.00325	$0.0026
AI fit reviewed (one per verdict that passed our checks)	$0.03	$0.024	$0.0195	$0.0156
Shortlist ranked (one per run)	$0.10	$0.08	$0.065	$0.052
Source access checked (a typed YouTube channel YouTube refused, 403/429 — instead of a creator check)	$0.005	$0.004	$0.00325	$0.0026

The store card shows AI fit reviewed (the primary event): it is charged per AI verdict on top of the creator check, the run start and the shortlist. Example: 100 creators on BRONZE with an AI verdict for each = $0.004 + 100 × $0.004 + 100 × $0.024 + $0.08 = $2.884 (this Actor's events; the scraper runs that made your list are separate). Your run's maximum charge is always respected: when it is reached the run stops cleanly (SUCCEEDED) and says what was left out.

What you give it

• datasetIds — datasets of other runs on your account (any scraper). Field names of common scrapers are read as they come:
  • YouTube: channelUrl, channelId, handle, title / channelName, description, subscriberCount / subscribers, email / emails (our YouTube Channel Email Finder and YouTube Channel Monitor, and the common YouTube e-mail scrapers);
  • Instagram: username, fullName, biography, followersCount, latestPosts[].caption, businessEmail, externalUrl (Apify's Instagram Profile Scraper), and post rows with ownerUsername + caption;
  • TikTok: authorMeta.name, authorMeta.signature, authorMeta.fans, text, hashtags (one row per video — the rows of one creator are merged).
• csvUrls — public CSV or JSON files with the same columns (Apify's flattened export headers like authorMeta/name are read).
• creators — profiles typed in: https://www.youtube.com/@name, instagram.com/name, tiktok.com/@name, @handle, instagram:name, tiktok:name. For YouTube channels typed here the public About page is read once (no login). For Instagram and TikTok entries typed here only the handle is known, so their verdict is usually unclear — bring a dataset with their bio and posts.
• brief — one or two sentences about the campaign. mustHave and exclude keyword lists, minFollowers / maxFollowers.

The same creator from several sources is scored and charged once.

What you get

One creator row per creator (an example on made-up data — the fields are those of a real row):

{
  "recordType": "creator",
  "platform": "instagram",
  "handle": "trailcook.example",
  "name": "Trail Cook",
  "followers": 48200,
  "verdict": "strong",
  "verdictSource": "ai",
  "ruleFit": {
    "verdict": "possible",
    "score": 3,
    "reasons": [{ "term": "camping", "quote": "Camping recipes for two-day hikes.", "source": "bio" }],
    "mustHaveFound": ["camping"],
    "mustHaveMissing": ["tent"],
    "excludeHits": [],
    "followerRange": "in",
    "excluded": false
  },
  "fit": {
    "status": "reviewed",
    "fit": "strong",
    "reasons": [{ "quote": "Testing a new backpacking stove on the ridge", "why": "Reviews the stove category the brief names." }],
    "audience": "beginner backpackers",
    "audienceEvidence": "for beginner backpackers",
    "topics": ["camp cooking", "backpacking"],
    "flags": ["sponsorship-mentioned"],
    "model": "claude-haiku-5-5"
  },
  "topics": ["camp cooking", "backpacking", "camping"],
  "riskFlags": ["ai:sponsorship-mentioned"],
  "contacts": { "emails": ["hello@trailcook.example"], "phone": null, "website": null },
  "plan": "paid",
  "billedEvents": ["creator-checked", "fit-reviewed"]
}

One shortlist row per run (paid plans): verdictCounts, top (rank, handle, verdict, top reason, risk flags), excludedCount, exclusionReasons (for example exclude-term:casino, followers-below-min, ai:exclusion-match) and notFitReasons.

Plus notice rows (never charged) and one limits row with the run's own report. A run that uses the AI review writes one ai-life notice before its first model request (one more after each restart of the run): it is a record for the run's own restart checks, not a result.

How the verdicts are made

• Rules (every plan, no AI): your keywords and the topic words of your brief are matched against the creator's own public text. strong = every mustHave found (or four or more brief topics without mustHave), possible = some, weak = none or excluded, unclear = no public text in your data. Every match comes back with the sentence it was found in, as written.
• AI (paid plans, beta): the creator's public text and your brief go to Anthropic's API (claude-haiku-5-5). The answer is checked before it is used: every quote must be in the creator's text letter for letter, the audience must be quoted, an exclusion flag must be quoted, and no number may appear that is in neither the text nor your brief. An answer that fails is thrown away and not charged; the row keeps the rule verdict. Creators without public text and creators your rules already exclude are not sent.
• Privacy: with the AI review on, the creators' public text and your brief are sent to Anthropic's API (United States). E-mail and web addresses are removed from the text first, and no contact field is ever sent. Turn aiFit off to keep everything on Apify.

What is charged, and what is not

• Charged: the run-start fee once per run, when the input checks passed and before the sources are read — it pays for reading and merging what you gave, also when it holds no creator or a source cannot be read. Every creator row the rules scored — also weak and unclear ones (the verdict is the product), and a YouTube channel YouTube answers does not exist. A typed YouTube channel whose lookup YouTube refused (HTTP 403/429): a row with accessRefused: true and no verdict, charged once as a source access check instead. AI verdicts that passed the checks. The shortlist row.
• Not charged: an input with nothing to read (no dataset ID, file address or valid creator) or nothing to judge by (no brief or mustHave), a run whose maximum charge is below the start fee (it stops before reading), profiles that failed on our side (also a refusal right after one of our failed attempts), invalid entries, the same creator twice, AI answers we threw away, notices and the report row.
• Two runs of the same input at the same time each score and charge on their own.

Good to know

• How much a run reads: at most 3 rows per creator of maxCreators (default 500 → 1,500 rows; at most 20,000 rows of datasets and files in all), 5 MB of files and 120 seconds of reading. Past one of these, reading stops with a source-budget notice (not charged) and the creators read so far are judged. A video-per-row dataset (TikTok, Reels) has many rows per creator: raise maxCreators to read more of it.

• No contact details are collected. Contact fields in your data are carried over as they were. For YouTube business e-mails, run our YouTube Channel Email Finder on the same list first, then feed its dataset here.

• Keyword matching does not understand meaning (a word can appear in a negative sentence) — the quoted sentence is there so you can see why. The AI verdict is the layer that reads meaning.

• Verdicts are an aid to your own review, not a statement of fact about a person.

Responsibility

Unofficial, independent tool, not affiliated with, endorsed by or sponsored by YouTube, Google, Instagram, Meta or TikTok. It judges the creator data you bring in (and, for YouTube channels you type in, the public About page read without logging in); it collects no contact details and sends no messages. You are responsible for having the right to use the data you bring in and a lawful basis to process it (for example under GDPR, CCPA or PIPA), for complying with anti-spam law (for example CAN-SPAM) when you contact anyone, and with the terms of the platforms the data came from. You are the data controller for the output.