Manual TikTok creator vetting takes roughly 45 minutes per creator: scroll their profile, eyeball engagement, check comment quality, look for brand safety red flags. API vetting takes under 2 minutes per creator with quantified metrics instead of gut feelings. For agencies vetting 50+ creators per campaign, the API approach turns a week of work into an afternoon.
Manual vetting: what it actually takes
The manual process for a single creator: open their TikTok profile (2 min), scroll through recent 20 posts noting views and likes (10 min), read comments on 5 posts checking for bots and spam (10 min), check follower count growth by comparing to cached data or third-party estimates (5 min), review bio and links for brand safety (3 min), cross-reference on other platforms (5 min), write up assessment notes (10 min). Total: about 45 minutes of focused work per creator, and the assessment is subjective.
API vetting: the pipeline
Python
import requests, os
from datetime import datetime, timedelta
API_KEY = os.environ["SCAVIO_API_KEY"]
BASE = "https://api.scavio.dev/api/v1/tiktok"
HEADERS = {"Authorization": f"Bearer {API_KEY}",
"Content-Type": "application/json"}
def tiktok_api(endpoint, payload):
resp = requests.post(f"{BASE}/{endpoint}",
headers=HEADERS, json=payload)
return resp.json()["data"]Step 1: Profile lookup
Python
def get_creator_profile(username):
"""Pull full profile data for a creator."""
data = tiktok_api("profile", {"username": username})
user = data["user"]
return {
"username": username,
"sec_uid": user["sec_uid"],
"followers": user["follower_count"],
"following": user["following_count"],
"videos": user["aweme_count"],
"total_likes": user["total_favorited"],
"verified": user.get("verification_type", 0) > 0,
"bio": user.get("signature", ""),
"bio_link": bool(user.get("bio_url", "")),
}
# 1 creditStep 2: Engagement calculation from recent posts
Python
def analyze_engagement(sec_uid, count=20):
"""Calculate real engagement metrics from recent posts."""
data = tiktok_api("user/posts",
{"sec_user_id": sec_uid, "count": count})
posts = data["aweme_list"]
engagements = []
post_dates = []
for p in posts:
stats = p["statistics"]
total_engagement = (stats["digg_count"] + stats["comment_count"]
+ stats["share_count"])
views = max(stats["play_count"], 1)
engagements.append({
"views": views,
"likes": stats["digg_count"],
"comments": stats["comment_count"],
"shares": stats["share_count"],
"engagement_rate": total_engagement / views,
})
post_dates.append(p["create_time"])
avg_views = sum(e["views"] for e in engagements) / len(engagements)
avg_engagement = (sum(e["engagement_rate"] for e in engagements)
/ len(engagements))
view_consistency = min(e["views"] for e in engagements) / max(
max(e["views"] for e in engagements), 1)
# Post frequency: days between oldest and newest post
if len(post_dates) > 1:
time_span = max(post_dates) - min(post_dates)
days = max(time_span / 86400, 1)
posts_per_week = (len(post_dates) / days) * 7
else:
posts_per_week = 0
return {
"avg_views": int(avg_views),
"avg_engagement_rate": round(avg_engagement * 100, 2),
"view_consistency": round(view_consistency, 2),
"posts_per_week": round(posts_per_week, 1),
"post_count_analyzed": len(posts),
}
# 1 creditStep 3: Follower quality check
Python
def check_follower_quality(sec_uid, sample_size=50):
"""Sample followers to detect bot networks."""
data = tiktok_api("user/followers",
{"sec_user_id": sec_uid, "count": sample_size})
followers = data.get("followers", [])
suspicious = 0
for f in followers:
red_flags = 0
# No profile picture
if not f.get("avatar_thumb", {}).get("url_list"):
red_flags += 1
# No videos posted
if f.get("aweme_count", 0) == 0:
red_flags += 1
# Following way more than followers (bot pattern)
following = f.get("following_count", 0)
follower_count = max(f.get("follower_count", 0), 1)
if following > follower_count * 10 and following > 500:
red_flags += 1
# Generic or empty bio
if not f.get("signature", "").strip():
red_flags += 1
if red_flags >= 3:
suspicious += 1
bot_rate = suspicious / max(len(followers), 1)
return {
"sample_size": len(followers),
"suspicious_count": suspicious,
"estimated_bot_rate": round(bot_rate * 100, 1),
"quality": "good" if bot_rate < 0.15 else
"warning" if bot_rate < 0.30 else "poor"
}
# 1 creditStep 4: Comment sentiment and quality
Python
def analyze_comments(sec_uid, video_count=3, comments_per=20):
"""Check comment quality on recent videos."""
posts = tiktok_api("user/posts",
{"sec_user_id": sec_uid, "count": video_count})
videos = posts["aweme_list"][:video_count]
all_comments = []
for video in videos:
video_id = video["aweme_id"]
comments = tiktok_api("video/comments",
{"video_id": video_id, "count": comments_per})
for c in comments.get("comments", []):
text = c.get("text", "")
all_comments.append({
"text": text,
"likes": c.get("digg_count", 0),
"length": len(text),
})
if not all_comments:
return {"quality": "no data", "avg_length": 0}
avg_length = sum(c["length"] for c in all_comments) / len(all_comments)
# Short generic comments ("nice", "wow", single emoji) = bot signals
generic = sum(1 for c in all_comments if c["length"] < 10)
generic_rate = generic / len(all_comments)
return {
"total_comments_analyzed": len(all_comments),
"avg_comment_length": round(avg_length, 1),
"generic_comment_rate": round(generic_rate * 100, 1),
"quality": "good" if generic_rate < 0.40 else
"warning" if generic_rate < 0.60 else "poor"
}
# 1 credit for posts + 3 credits for comments = 4 creditsStep 5: Full vetting report
Python
def vet_creator(username):
"""Complete creator vetting pipeline."""
print(f"Vetting @{username}...")
profile = get_creator_profile(username)
engagement = analyze_engagement(profile["sec_uid"])
followers = check_follower_quality(profile["sec_uid"])
comments = analyze_comments(profile["sec_uid"])
# Scoring
score = 0
flags = []
if engagement["avg_engagement_rate"] > 3.0:
score += 25
elif engagement["avg_engagement_rate"] > 1.5:
score += 15
else:
flags.append("low engagement rate")
if engagement["view_consistency"] > 0.3:
score += 20
else:
flags.append("inconsistent views (possible viral-only creator)")
if engagement["posts_per_week"] >= 2:
score += 15
else:
flags.append("infrequent posting")
if followers["quality"] == "good":
score += 20
elif followers["quality"] == "warning":
score += 10
flags.append(f"bot rate: {followers['estimated_bot_rate']}%")
else:
flags.append(f"high bot rate: {followers['estimated_bot_rate']}%")
if comments["quality"] == "good":
score += 20
elif comments["quality"] == "warning":
score += 10
flags.append("generic comments suggest low-quality engagement")
verdict = ("RECOMMENDED" if score >= 70 else
"PROCEED WITH CAUTION" if score >= 45 else
"NOT RECOMMENDED")
report = {
"username": username,
"followers": profile["followers"],
"engagement_rate": engagement["avg_engagement_rate"],
"view_consistency": engagement["view_consistency"],
"posts_per_week": engagement["posts_per_week"],
"bot_rate": followers["estimated_bot_rate"],
"comment_quality": comments["quality"],
"score": score,
"flags": flags,
"verdict": verdict,
}
print(f" Score: {score}/100 [{verdict}]")
print(f" Engagement: {engagement['avg_engagement_rate']}%")
print(f" Bot rate: {followers['estimated_bot_rate']}%")
if flags:
print(f" Flags: {', '.join(flags)}")
return report
# Full vet: 1 + 1 + 1 + 4 = 7 credits = $0.035 per creator
# Manual: 45 minutes of staff time
report = vet_creator("example_creator")Cost comparison at campaign scale
- Manual vetting (50 creators): 50 x 45 min = 37.5 hours of staff time. At $30/hour = $1,125.
- API vetting (50 creators): 50 x 7 credits = 350 credits = $1.75 on Scavio. Plus 30 min to review reports.
- Time saved: 36+ hours per campaign cycle.
- Quality difference: API gives quantified metrics (3.2% engagement, 12% bot rate) vs gut feeling ("seems good").
The API does not replace human judgment -- you still decide which creators fit your brand. It replaces the hours of manual data gathering that precede that judgment. Seven API credits per creator gives you a quantified vetting report that would take 45 minutes to assemble by hand, and the numbers do not lie the way eyeballed estimates do.