TikTok Comment Sentiment and Follower Growth Tracking

A creator's follower count only moves in one direction on their profile page — up, with no context for why. A brand vetting them for a partnership, or a creator checking their own account's health, needs two signals read together: is the audience actually growing, and does the sentiment in their comment section support that growth or contradict it. A creator can gain followers off one viral moment while their regular audience is souring in the comments underneath it — a follower count alone won't show you that split.

If you're an agent (or building one) reading this rather than a human, the full machine-readable schema for every endpoint below lives at mindcase.co/skills.md.

What does a follower count alone not tell you?

Say you're vetting a creator for a paid partnership, or checking your own brand account after a format change. Three questions a raw follower number can't answer on its own:

• Is growth recent or stalled? A profile snapshot is one point in time. Whether that number is climbing, flat, or falling only shows up by comparing snapshots taken days or weeks apart.
• Is the audience engaged or just large? A big follower count with a quiet, low-effort comment section reads differently than a smaller count with an active one.
• What's the tone underneath the numbers? A spike in negative comments on recent videos is a signal a follower count won't surface at all — it takes reading the comments themselves.

How do you pull a creator's current follower count?

TikTok Profiles API takes a search query — a handle works as well as a keyword — and returns 19 fields per matching account, including followers, following, totalLikes, videos, and accountCreated.

curl -X POST "https://api.mindcase.co/v1/data/tiktok/profiles/run?wait=true" \
-H "Authorization: Bearer $MINDCASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "params": {
    "searchQueries": "skincare.with.mia",
    "maxProfilesPerQuery": 5
  }
}'

# $0.0025 per profile returned

There's no separate "look up one exact account" input — searchQueries runs the same search TikTok's own search box would, so passing a handle usually returns that account as the top result, but the response can carry more than one match for a common handle fragment. Matching on the returned handle field, as above, is what actually pins down the right row rather than trusting result order.

How do you pull comments from their recent videos?

TikTok Comments API has a "by profile" mode: pass the handle through profiles instead of a video URL, and it pulls comments from that account's own recent videos, controlled by resultsPerPage (videos per profile) and commentsPerPost (comments per video).

curl -X POST "https://api.mindcase.co/v1/data/tiktok/comments/run?wait=true" \
-H "Authorization: Bearer $MINDCASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "params": {
    "profiles": "skincare.with.mia",
    "resultsPerPage": 5,
    "commentsPerPost": 40,
    "profileSorting": "latest"
  }
}'

# $0.0005 per comment returned

profileSorting: "latest" pulls comments from the five most recent videos rather than the five most popular — recent comments are what reflect the account's current direction, not whichever old video happens to have the most engagement. Each returned row carries comment, likes, replies, and postedDate — the endpoint returns the comment text itself, not a sentiment label, so scoring tone is a step you run on top of the pull, the same way a raw star rating isn't a sentiment score on its own.

How do you turn repeated pulls into a growth and sentiment trend?

Neither endpoint is schedule-aware or returns history — each call answers "what does this look like right now." A trend comes from storing what each pull returns and comparing it to the last one:

Every 7 days, for each creator on the watchlist:
1. Pull their profile (as above) and store {handle, followers,
   totalLikes, videos, date}.
2. Pull comments from their 5 most recent videos.
3. Score each comment's text for sentiment (a small LLM call per
   batch works well here — the endpoint gives you the text, not a score).
4. Store {date, followerDelta since last pull, negative comment share}.
5. Flag any creator where followers are still rising but negative
   comment share has climbed two weeks running — a real signal, since it
   means growth is coming from a source the existing audience is
   souring on.

That last check is the actual point of pulling both signals together: a rising follower count with softening sentiment underneath it is a pattern a follower count by itself can't show, and it's the kind of thing worth catching before a partnership deal goes out, not after.

What can't you get from Mindcase?

Two cases.

A sentiment score. Comments come back as text, likes, and reply counts — real signal, no invented positive/negative label attached. Scoring tone from that text is a step you run yourself, same as with any other review or comment dataset.

Follower history. Every pull is a snapshot of the account right now. There's no endpoint that returns "followers on this date in the past" — a growth trend only exists once you've stored more than one of your own snapshots and compared them.

Which endpoint should you use for which job?

FAQ