Influencer Vetting API: How to Vet Creators With Public Data

Use an influencer vetting API to pull public profiles, posts and comments, score engagement and comment quality, then ask creators for first-party data.

By Scrape Everything team · · Updated · 12 min read

Last updated: November 6, 2026

TL;DR: An influencer vetting API workflow pulls a creator's public profile, recent posts and recent comments as JSON, then computes signals anyone can re-check: median engagement rate for the follower tier, views relative to followers, and how much of the comment section is emoji-only or duplicated. Public data flags creators for a closer look. It can't prove bought followers or show audience demographics, so ask for first-party analytics before you pay.

This is a screening method, not a vendor roundup: endpoints, math, a Python script for a shortlist, and what to ask creators who pass.

What does an influencer vetting API actually check?

Influencer vetting is the check a brand or agency runs before paying a creator: is the audience real, does it pay attention, and is the content safe for the brand? An API turns those questions into numbers you compute the same way for every creator on a shortlist, instead of scrolling a profile grid for ten minutes each.

Everything below uses public data, so nothing about who the followers are. We also don't sell a single "vetting score". When you can see the median engagement, reel views and comment sample behind a red flag, you can argue with it, and so can the creator.

Which public signals can you measure, and which not?

SignalMeasurable from public data?Endpoint and fields
Follower, following and post countsYesinstagram.photos (includes every profile field): follower_count, following_count, post_count; tiktok.profile: follower_count, like_count, video_count
Engagement per postYes, when counts are visibleinstagram.post (1 credit per post): like_count, comment_count, since the instagram.photos list leaves them null; instagram.reels: reels[].like_count, reels[].comment_count; tiktok.video: like_count, comment_count, share_count
Views vs followersYesinstagram.reels reels[].view_count, tiktok.posts posts[].play_count, youtube.videos videos[].view_count
SavesTikTok onlytiktok.video save_count; Instagram doesn't expose saves publicly
Posting cadence and gapsYestaken_at from instagram.post (Unix seconds) or the approximate reels[].taken_at; the posts list leaves it null
Comment qualityPartly: a sample onlyinstagram.comments (15 newest top-level), tiktok.comments
Sponsored vs organic performancePartly: from captionsposts[].caption (Instagram captions are cut at 300 characters)
Brand safetyPartly: from captions and bioposts[].caption, biography
Audience demographics, reach, impressionsNoAsk the creator for platform analytics
Bought or bot followers, directlyNoinstagram.followers lists only about the first 50 followers of large verified accounts; only indirect signals

One catalog detail changes the math: Instagram returns like_and_view_counts_disabled: true when the owner hid counts, and the instagram.photos posts list returned null likes, comments and taken_at for every post in our tests. Get per-post counts from instagram.post and drop posts that still come back null rather than counting them as zero.

How do you pull a creator's profile and recent posts?

Every endpoint is a POST to https://api.scrapeeverything.com/v1/<platform>/<action> with a JSON body and a bearer key. On Instagram, instagram.photos (2 credits) returns every field instagram.profile returns plus up to 60 recent posts, so calling both wastes a credit. Here's the call for a creator's last 30 posts:

bash
curl -X POST https://api.scrapeeverything.com/v1/instagram/photos \
  -H "Authorization: Bearer $SCRAPEEVERYTHING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"username": "example_creator", "limit": 30, "cache_days": 7}'

Trimmed response (values are placeholders):

json
{
  "success": true,
  "data": {
    "id": "1234567890",
    "username": "example_creator",
    "full_name": "Example Creator",
    "is_private": false,
    "is_verified": false,
    "is_business_account": true,
    "category_name": "Personal blog",
    "follower_count": 48200,
    "following_count": 610,
    "post_count": 734,
    "posts_fetched": 30,
    "has_more": true,
    "posts": [
      {
        "shortcode": "C0exAmPle12",
        "link": "https://www.instagram.com/p/C0exAmPle12/",
        "type": "album",
        "is_video": false,
        "like_count": null,
        "comment_count": null,
        "like_and_view_counts_disabled": null,
        "caption": "Three things I changed in my morning routine...",
        "taken_at": null,
        "taken_at_text": "October 28, 2026"
      }
    ]
  },
  "meta": {
    "request_id": "req_...",
    "cached": false,
    "credits_charged": 2,
    "credits_remaining": 98
  }
}

The full pull, per creator:

  1. instagram.photos, limit 12 (2 credits at any limit up to 60), then instagram.post on each of those posts for likes, comments and the publish time (1 credit per post; the list leaves them null).
  2. instagram.reels, limit 12, for view_count (2 credits). Skip collab reels (a different owner_id) and pinned ones (is_pinned: true, usually old hits).
  3. instagram.comments on 2 recent posts, passing the post url (1 credit each).
  4. TikTok: tiktok.profile and tiktok.posts (1 credit each). tiktok.posts returns plays only, so call tiktok.video on the 3 newest videos for likes, comments, shares and saves (1 credit each).
  5. YouTube: youtube.profile and youtube.videos (1 credit each). Counts there are rounded, so treat them as approximate.

A private account returns is_private: true with an empty posts array, and the reels and comments calls return 404 with PRIVATE_ACCOUNT. For a paid partnership, that's a finding in itself. Field details are on the Instagram profile scraper page, the Instagram Reels scraper page and the TikTok profile scraper page.

One pull is a snapshot. A jump of thousands of followers in a week with no viral post behind it is a classic influencer fraud signal, and spotting it needs history: call the profile endpoint on a schedule without cache_days and store follower_count with a date. See tracking follower counts over time for a snapshot script.

How do you check influencer engagement per follower tier?

Engagement rate per post is (likes + comments) / followers x 100. Take the median, not the mean: one viral post drags the mean up and says nothing about the next sponsored post.

Rival IQ's 2025 Social Media Industry Benchmark Report uses the same denominator (interactions divided by followers) and reports rates fell year over year, by 16% on Instagram and 34% on TikTok. Its sample is brand accounts, not creators. Other vendors divide by reach or impressions, which is why published "good engagement rate" numbers disagree.

For a starting point, our influencer-vetting Agent Skill uses these rough norms. They're heuristics, not standards:

PlatformRough median engagement norm (our heuristic)Flag when
Instagramlikes + comments / followers, roughly 1 to 3% (lower for 1M+ accounts)far below the band for the size tier
TikToklikes / plays, roughly 3 to 10%far below the band
YouTubelikes / views, roughly 2 to 5%far below the band
Reels and TikTokviews / followersunder 3% on a large account

Replace them with your own soon. Bucket every creator you screen by follower band (under 10k, 10k to 100k, 100k to 1M, over 1M), since engagement tends to fall as accounts grow, and keep the median per band per platform. A few dozen creators in a band gives you numbers that fit your niche and formula. Flag creators far below their band, and also those far above it with flat views: likes without views is what you'd expect from engagement pods or paid likes. That's a question to ask, not a verdict.

Three more numbers from the same pull:

  • Views-to-followers: median reel or TikTok plays divided by followers. Big follower counts with tiny views suggest an audience that doesn't see the content.
  • Post-to-post variance: the coefficient of variation (standard deviation / mean) of engagement. The skill treats anything above 1.5 as hit-driven: a couple of breakout posts and a lot of quiet ones.
  • Sponsored vs organic: split median engagement by captions containing #ad, #sponsored, "paid partnership" or a discount code. A creator can look great overall and go quiet every time a brand pays. Captions are cut at 300 characters, so a late disclosure can be missed.

On TikTok, add saves: save_count from tiktok.video shows people keeping a video, a different kind of interest than a quick like.

How do you spot suspicious engagement in comments?

Comments are the cheapest fake-follower check you can run, because bot and pod activity leaves a pattern in text. From a sample of recent comments, compute:

  • Emoji-only ratio: comments with no letters or digits, such as a row of fire emojis.
  • Duplicate ratio: comments whose normalized text (lowercased, whitespace collapsed) appears more than once across the sampled posts. Repeats across different posts are the stronger pod signal.
  • Short generic ratio: two words or fewer, like "nice pic" or "love this".

None of these is proof. A fitness creator with a loyal audience honestly gets plenty of emoji replies. Worry when most comments carry no content and the same phrases show up on every post. Then read the comments yourself; whether they respond to the caption is still a human call. If the brief targets the UK and most comments are in another language, note it. Weak signal, free to check.

Mind the sample size: instagram.comments returns at most the 15 newest top-level comments per post, so two posts give you 30 at most. And keep the report about the account. The Instagram comments scraper returns author usernames; a vetting report doesn't need them.

How do you check brand safety from captions?

You already have the captions and bio from instagram.photos. A keyword scan against the brand's blocklist (gambling, competitor names, topics legal has ruled out) catches the obvious problems before a human reviews the rest. It misses anything said only in a video's audio; scan a transcript for that.

What does a full screening script look like?

This script screens one Instagram creator and returns every signal above. It needs requests and the standard library, and it costs 18 credits per creator (2 + 12 + 2 + 1 + 1).

python
import os, re, statistics, requests

API = "https://api.scrapeeverything.com/v1"
HEADERS = {"Authorization": f"Bearer {os.environ['SCRAPEEVERYTHING_API_KEY']}"}
SPONSORED = re.compile(r"#ad\b|#sponsored|paid partnership|use my code", re.I)
BLOCKLIST = ["casino", "betting", "competitorbrand"]  # replace with the brand's list

def call(endpoint: str, body: dict) -> dict:
    res = requests.post(f"{API}/{endpoint}", headers=HEADERS,
                        json={**body, "cache_days": 7}, timeout=90)
    if "json" not in res.headers.get("content-type", ""):
        raise RuntimeError(f"{endpoint}: HTTP {res.status_code}")  # 502/504 arrive as HTML
    data = res.json()
    if not data.get("success"):
        raise RuntimeError(f"{endpoint}: {data.get('code')} {data.get('message')}")
    return data["data"]

def med(values: list, digits: int = 2):
    return round(statistics.median(values), digits) if values else None

def comment_quality(comments: list[dict]) -> dict:
    texts = [re.sub(r"\s+", " ", c["text"].strip().lower())
             for c in comments if c["text"].strip()]
    n = len(texts)
    if not n:
        return {"sample": 0}
    emoji_only = sum(1 for t in texts if not re.search(r"\w", t))
    short = sum(1 for t in texts if len(t.split()) <= 2)
    dupes = n - len(set(texts))  # pooled, so cross-post repeats count
    return {"sample": n, "emoji_only": round(emoji_only / n, 2),
            "short": round(short / n, 2), "duplicate": round(dupes / n, 2)}

def vet_instagram(username: str) -> dict:
    prof = call("instagram/photos", {"username": username, "limit": 12})
    if prof["is_private"]:
        return {"username": prof["username"], "private": True}
    followers = prof["follower_count"]
    if not followers:
        return {"username": prof["username"], "error": "follower count missing"}

    # the posts list leaves likes, comments and taken_at null, so read each post (1 credit)
    details = [call("instagram/post", {"url": p["link"]}) for p in prof["posts"]]
    posts = [p for p in details if p["like_count"] is not None
             and not p.get("like_and_view_counts_disabled")]
    er = {p["link"]: (p["like_count"] + (p["comment_count"] or 0)) / followers * 100
          for p in posts}
    sponsored = [er[p["link"]] for p in posts if SPONSORED.search(p["caption"])]
    organic = [er[p["link"]] for p in posts if not SPONSORED.search(p["caption"])]
    rates = list(er.values())

    reels = call("instagram/reels", {"username": username, "limit": 12})["reels"]
    own = [r for r in reels if r.get("owner_id") in (None, prof["id"])
           and not r["is_pinned"] and not r.get("like_and_view_counts_disabled")]
    views = [r["view_count"] for r in own if r["view_count"] is not None]

    times = sorted(p["taken_at"] for p in details if p["taken_at"])
    gaps = [(b - a) / 86400 for a, b in zip(times, times[1:])]

    text = " ".join([prof["biography"]] + [p["caption"] for p in prof["posts"]]).lower()
    blocklist_hits = [w for w in BLOCKLIST if w in text]

    comments = []
    for p in prof["posts"][:2]:
        comments += call("instagram/comments", {"url": p["link"]})["comments"]

    mean = statistics.mean(rates) if rates else 0
    return {
        "username": prof["username"],
        "followers": followers,
        "following_to_followers": round((prof["following_count"] or 0) / followers, 3),
        "median_er_pct": med(rates),
        "median_er_sponsored_pct": med(sponsored),
        "median_er_organic_pct": med(organic),
        "er_cv": round(statistics.pstdev(rates) / mean, 2) if mean else None,
        "median_views_to_followers_pct": round(statistics.median(views) / followers * 100, 1) if views else None,
        "median_days_between_posts": med(gaps, 1),
        "longest_gap_days": round(max(gaps), 1) if gaps else None,
        "blocklist_hits": blocklist_hits,
        "comments": comment_quality(comments),
    }

print(vet_instagram("example_creator"))

A missing follower count returns an error rather than a silently inflated rate. The photos response has no pinned flag (reels do), so if longest_gap_days looks absurd, check for an old pinned post.

With cache_days set, re-running the script inside the window returns stored results for 0 credits, so fixing your scoring doesn't cost a second pull. Blocks, timeouts and errors on our side are refunded automatically; definitive 404s such as PRIVATE_ACCOUNT are charged. For which errors to retry, see the errors reference.

How do you run the same checks with one prompt?

If you'd rather not maintain a script, the influencer-vetting Agent Skill does this inside Claude Code, Cursor, Codex or Gemini CLI. An Agent Skill is a folder with a SKILL.md file that an AI agent loads when a request matches it. This one tells the agent which endpoints to call, how to compute median engagement, views-to-followers, cadence and consistency, and how to write a go / no-go report.

bash
npx skills add scrapeeverything/skills --skill influencer-vetting
export SCRAPEEVERYTHING_API_KEY=ssk_live_...

Then ask: "Vet @example_creator for a sponsorship. We sell running shoes to women 25-40 in the UK." The skill checks Instagram, TikTok, YouTube and X, marks each red flag with its evidence, scans captions for brand-safety issues and competitor sponsorships, and scores audience fit against your brief. Fake-follower signals are presented as indicators, not accusations.

A quick one-platform vet costs about 6 credits. A full four-platform vet costs about 17: 13 credits for the Instagram, TikTok and YouTube calls, plus 2 youtube.video calls for exact views and likes on the latest uploads and 2 credits for x.posts. Setup is on the Agent Skills integration page, and how Agent Skills work with our API covers each agent.

My take: use the script to screen 200 creators on a schedule, and the skill for the five finalists where you want a written report and a judgment call.

What does vetting 50 creators cost in credits?

Using the per-creator calls from the pull list:

Scope per creatorCallsCredits per creator50 creators
Instagram quick screeninstagram.photos + 12x instagram.post + instagram.reels + 2x instagram.comments18900
Add TikToktiktok.profile + tiktok.posts + 3x tiktok.video5250
Add YouTubeyoutube.profile + youtube.videos2100
All three platforms251,250

And in dollars, using our one-time credit packs (credits never expire):

PackPriceCreditsInstagram screens (18 credits)Three-platform screens (25 credits)Cost per three-platform screen
Free signup$010054$0
Builder$19.9925,0001,3881,000about $0.020
Growth$299500,00027,77720,000about $0.015
Scale$5991,200,00066,66648,000about $0.012

Cached re-reads inside cache_days cost nothing, so repeat screens come in under these numbers. Plug your own volume into the credit calculator.

Audience-analytics tools such as HypeAuditor or Modash report demographics we can't (the "No" rows above). Use the public screen to cut a long list, then spend paid audience reports on finalists. Two independent reads beat one.

Rate limits: defaults of 60 requests per minute and 5 concurrent, rising to 300 per minute and 20 concurrent once purchases total $19.99 or more (any pack). Agencies running this weekly will find more on the agencies use case page.

Get 100 free credits and screen your first 5 creators on Instagram before you buy anything.

What should you ask the creator that no scraper can give you?

Once a creator passes the screen, the rest of your influencer audit checklist comes from them. Ask for:

  1. Screenshots or an export from Instagram Insights or TikTok analytics for the last 30 to 90 days: audience age, gender and top countries.
  2. Reach and impressions on their last 5 to 10 posts, and story views if stories are in the deal.
  3. Results from a past sponsored post: link clicks, code redemptions, or whatever the previous brand measured.
  4. Follower growth over the last 6 to 12 months, so you can see whether a jump lines up with a viral post or with nothing.
  5. How they label paid posts. The FTC's Disclosures 101 for Social Media Influencers is the US reference, and a creator who can't describe how they disclose ads is a contract risk.

If their first-party numbers contradict the public ones (high claimed reach, flat public views), raise it before you sign.

Keep reports to the creator's professional presence, without personal details or individual commenters. Our post on whether scraping public social media data is legal lists questions for your own counsel.

Next step: run the script on three creators you know, one you trust and two you doubt. If the signals don't separate them, adjust the thresholds before screening anyone new.

FAQ

Can an API detect fake followers?

Not directly. Follower lists are only partly available (about the first 50 on large verified accounts), so no scraper-based check can label an audience's followers as bots. You can measure influencer fraud signals: low views for the follower count, low engagement for the tier, following close to followers, sudden growth with no viral post, and mostly emoji or duplicate comments. Treat them as reasons to ask questions.

What is a good engagement rate for an influencer?

There's no single number; it depends on platform, niche, tier and formula. As a rough heuristic, our vetting skill uses about 1 to 3% for Instagram (likes plus comments over followers). Your own median per platform and follower band is the better yardstick.

How many posts should I sample?

Twelve recent posts is a practical minimum for a median that one outlier can't dominate. instagram.photos accepts a limit up to 60 for a longer window at the same 2 credits, plus 1 credit for each post you read with instagram.post for its counts.

Does this work for private accounts?

No. Private accounts return profile fields only: instagram.photos gives an empty posts array, and instagram.reels and instagram.comments return 404 with PRIVATE_ACCOUNT. For a paid partnership, ask the creator to make the account public or share analytics directly.

Which platforms can I vet?

The workflow here covers Instagram, TikTok and YouTube, and the Agent Skill adds X. Scrape Everything covers 9 platforms in total: Instagram, TikTok, X, YouTube, Facebook, Telegram, Spotify, Pinterest and Twitch. LinkedIn, Reddit, Threads and Snapchat aren't supported.

Can I run this in n8n or through MCP instead of Python?

Yes. The same endpoints are available through the n8n node (n8n-nodes-scrapeeverything), the MCP server (@scrapeeverything/mcp) and the OpenAPI spec at https://scrapeeverything.com/openapi.json, so you can generate a client in any language.

More from the blog

Start with 100 free credits.

Continue with Google, create an API key and send your first request in under a minute.

Get 100 free credits