TikTok AI Detector β AI-Generated Videos & Quality Scraper
Pricing
$3.00 / 1,000 videos
TikTok AI Detector β AI-Generated Videos & Quality Scraper
π€ TikTok AI detector β detect AI-generated TikToks, including UNDISCLOSED AI TikTok flagged but the creator never labeled. Reads TikTok's native AI-content labels + video quality score, filters & effects. Scan any keyword, niche or the trending feed. No login, no proxy.
Pricing
$3.00 / 1,000 videos
Rating
0.0
(0)
Developer
Yakugusa Yumitori
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
2 days ago
Last modified
Categories
Share
π€ See TikTok's own AI & quality signals on any video β is it AI-generated (and did the creator disclose it, or did TikTok catch it undisclosed?), TikTok's internal video-quality score, the filters/effects used, auto-captions, detected language, and algorithmic category. By keyword, niche, or the live trending feed. No login, no proxy.
These are TikTok's own native labels β not a third-party AI guess. Every video TikTok itself has tagged as AI-generated, plus the ones its moderation detected as AI even though the creator never labeled them.
Everyone sells "AI content detectors" that guess from pixels. This reads the label TikTok already assigned β including the AI it flagged that the creator tried to pass off as real.
π― Why this is different
Search Apify for "TikTok AI" and you get transcript extractors (AI transcribes the video) and LLM analysers (AI summarises it). None of them expose TikTok's own algorithmic signals. This one does β straight from TikTok's API:
- π΅οΈ Undisclosed-AI detection.
moderationAigcLabelTypeis TikTok's own system flagging a video as AI-generated. When that fires but the creator set no AI label, this actor marks itai_detected_undisclosedβ AI slipping through unlabeled. Nobody else surfaces this. - π Quality score. TikTok's internal video-quality score (0β100) β the number its ranking system sees. Plus quality tier, resolution, bitrate, codec, and HD flag.
- β¨ Filters & effects. Every effect/filter applied, with counts across a niche.
- π Auto-captions & language. Whether TikTok auto-captioned it, the detected spoken language, and caption language.
- π·οΈ Algorithmic category. TikTok's internal content category + diversification IDs.
π€ Output β per video
| Field | Description |
|---|---|
is_ai_generated | TikTok flags it as AI (disclosed or detected) |
ai_disclosed | the creator labeled it AI-generated |
ai_detected_undisclosed | TikTok detected AI the creator did NOT label π΅οΈ |
quality_score, quality_band | TikTok's video-quality score (0β100) + high/medium/low |
quality_tier, resolution, is_hd, bitrate_kbps, codec | full production-quality picture |
effects, effect_count | filters/effects applied |
auto_caption, spoken_language, caption_language | AI caption + language signals |
category_id, diversification_id | TikTok's algorithmic categorization |
views, save_rate, engagement_rate, posted, age_days, creator* | engagement, recency + creator context |
Per-query aggregates (with onePerVideo off): ai_content_pct, ai_disclosed_pct, ai_undisclosed_pct, avg_quality_score, quality_high/medium/low, hd_pct, auto_caption_pct, top_effects, top_languages, top_ai_creators.
βΆοΈ Example runs
Audit undisclosed AI in a niche:
{ "keywords": ["ai influencer", "faceless channel"], "maxVideos": 100, "onePerVideo": false }
Build a dataset of only AI-generated videos:
{ "niches": ["ai art", "ai animation"], "aiOnly": true, "maxVideos": 150 }
How much of what's going viral right now is AI?
{ "trendingVideos": true, "onePerVideo": false }
Is AI spiking in a niche this week? (date filter)
{ "keywords": ["ai influencer"], "postedWithin": "7", "onePerVideo": false }
π¦ Example output
Default β one row per video ({ "keywords": ["ai generated story"] }):
{"video_url": "https://www.tiktok.com/@animeseries156/video/7635050592662064397","creator": "animeseries156", "creator_name": "Anime series", "creator_followers": 203900,"is_ai_generated": true,"ai_disclosed": true,"ai_detected_undisclosed": false,"ai_label_type": 1,"quality_score": 70.5, "quality_band": "medium", "quality_tier": "normal","resolution": "720p", "is_hd": false, "bitrate_kbps": 1961, "codec": "h264", "duration": 70,"effects": [], "effect_count": 0,"auto_caption": true, "spoken_language": "pol-PL", "caption_language": null,"category_id": 100, "diversification_id": 10052,"views": 13600000, "likes": 200200, "comments": 1285, "shares": 3726, "saves": 16900,"save_rate": 0.12, "engagement_rate": 1.63, "posted": "2026-05-01","keyword": "ai generated story"}
An undisclosed-AI row looks identical but with
"ai_disclosed": false, "ai_detected_undisclosed": trueβ e.g.@sunnycartoon01, which TikTok flagged as AI-generated even though the creator never labeled it.
Aggregate β one row per query ({ "keywords": ["ai animation"], "onePerVideo": false }):
{"keyword": "ai animation","video_count": 20,"ai_content_pct": 90,"ai_disclosed_pct": 55,"ai_undisclosed_pct": 35,"avg_quality_score": 69.8,"quality_high": 0, "quality_medium": 18, "quality_low": 1,"hd_pct": 0,"auto_caption_pct": 95,"top_languages": [{ "language": "eng-US", "count": 6 }, { "language": "kor-KR", "count": 2 }],"top_ai_creators": ["levkudrin", "craziest_a4", "chopstory_", "sunnycartoon01"],"top_effects": []}
90% of "ai animation" videos are AI-generated β and 35% of them are undisclosed (TikTok detected the AI; the creator never labeled it). Turn on
aiOnlyto keep just the flagged videos, ortrendingVideosto run the same breakdown on the live For-You feed.
Filters & effects (TikTok Effect House) β effect-heavy niches name the exact filters used, and each trending filter links to its effect page ({ "keywords": ["beauty filter"] }):
// per video β every filter/effect applied, by name"effects": ["Bold Glamour", "Beautify", "Pink Skin bloom", "Idol Retro Blue"],"effect_count": 4// aggregate (onePerVideo: false) β trending filters, each linking to every video that uses it"top_effects": [{ "effect": "Bold Glamour", "count": 10, "effect_url": "https://www.tiktok.com/effect/bold-glamour-14519845" },{ "effect": "Idol Retro Blue", "count": 10, "effect_url": "https://www.tiktok.com/effect/idol-retro-blue-2125959484" },{ "effect": "Pink Skin bloom", "count": 10, "effect_url": "https://www.tiktok.com/effect/pink-skin-bloom-282517011" }]
These filters are TikTok effects β many built by creators in TikTok Effect House.
effectsnames every filter on a video;top_effectsranks the filters a niche leans on and links each to its effect page (which lists every video using it). So you can spot the Effect House filter behind a look that's blowing up β or check whether a creator's "natural" glow is really Bold Glamour.
π‘ What people use it for
- Brand safety & ad verification β filter out AI-slop before placing spend or partnering; flag undisclosed AI in a category.
- AI-content researchers & journalists β measure how fast AI-generated content is spreading by niche, and how much is going unlabeled.
- Platform & policy teams β benchmark undisclosed-AI rates and AI-label compliance across topics and markets.
- Agencies & studios β benchmark production quality (quality score, HD%, bitrate) against competitors and spot the filters trending in a niche.
- Trust & safety / academic β reproducible datasets of TikTok's own AI labels without manual review.
- Filter / "catfish" check β see whether a creator leans on beauty filters (Bold Glamour & co.) or AI enhancement rather than posting unfiltered β flags heavily-filtered or AI-generated profiles at a glance (handy before paying for or partnering with a creator).
π€ Agentic AI use-cases
- "Is this TikTok AI?" checker β an agent takes a keyword/creator, runs this actor, and answers whether the content is AI-generated (disclosed vs. TikTok-detected-undisclosed) with the quality score attached.
- AI-slop gatekeeper β before an agent surfaces or reposts TikToks, drop anything
is_ai_generated(or onlyai_detected_undisclosed). - Undisclosed-AI monitor β schedule on a niche; alert when
ai_undisclosed_pctclimbs (unlabeled AI rising) β an early-warning feed for trust & safety. - AI-proliferation tracker β log
ai_content_pctper niche daily to chart how fast AI-generated content is taking over a category. - Quality-bar filter β keep only
quality_band == "high"+is_hdclips for a highlight reel or competitor benchmark. - Filter-trend agent β track
top_effectsweek over week to catch the effect/filter a niche is adopting. - Brand-safety pre-check β an agent vets a creator's recent videos for AI-generated content and quality before greenlighting a partnership.
π§© Wire it into an AI agent (Claude, ChatGPT, n8nβ¦)
This actor is a clean tool for agents β deterministic JSON in, structured signals out. Concrete scenarios:
- Claude (via Apify MCP / tool use): expose this actor as a tool and ask "Vet @creator's niche for AI slop before we sponsor them." Claude runs it, drops anything
ai_detected_undisclosedor lowquality_band, and returns a go/no-go with the exact flagged clips and reasons. - ChatGPT (Custom GPT Action / Assistants API): build an "Is this TikTok AI?" GPT with the actor as an Action. A user types a niche or handle; the GPT calls it and replies with the AI %, the undisclosed-AI %, and which creators are posting unlabeled AI.
- Autonomous / scheduled agents (n8n, LangChain, CrewAI): schedule a trust-&-safety agent to run the actor nightly on a watchlist of niches; when
ai_undisclosed_pctcrosses your threshold it fires a Slack/email alert with the offending videos. - Analyst agent (any LLM): pipe the
onePerVideo: falseaggregate rows into your model to auto-write a weekly "State of AI content" brief per niche β AI %, quality trend, top filters, top AI creators. - Content-pipeline guardrail (Gemini/Llama/any): before an agent reposts or recommends TikToks, gate on
is_ai_generated+quality_band == "high"so only human, high-quality clips pass.
π» Run it from your code
Python
from apify_client import ApifyClientclient = ApifyClient("YOUR_APIFY_TOKEN")run = client.actor("yumitori/tiktok-ai-signals-scraper").call(run_input={"keywords": ["ai animation"], "onePerVideo": False,})for row in client.dataset(run["defaultDatasetId"]).iterate_items():print(row["keyword"], row["ai_content_pct"], row["ai_undisclosed_pct"], row["avg_quality_score"])
CLI
$apify call yumitori/tiktok-ai-signals-scraper --input '{ "niches": ["ai art"], "aiOnly": true }'
β‘ Getting the most results
- Up to 200 videos per query β set
maxVideosto 200. - Add many niches/keywords β each is its own query and they run in parallel, so 20 keywords Γ 200 = up to 4,000 videos in one run. This is how you scale.
- Turn on
trendingVideosfor a big, no-keyword pull β it rotates TikTok's For-You feed across regions to aggregate a large, fresh set of viral videos. - Use
aiOnlyto spend your budget only on AI-flagged videos (a denser AI dataset per credit). - Raise
concurrency(up to 10) to process more queries at once on bigger runs.
π Proxies, cookies & safety
Do I need a proxy? No. It signs TikTok's own public API directly, so it runs proxy-free on Apify's servers. No residential proxy, no proxy config, nothing extra to buy or set up.
Does it use my TikTok login or cookies?
No login, no account, no cookies from you. Each run mints its own fresh anonymous session token (the same ttwid any logged-out visitor gets). You never paste credentials, and none are stored.
Is my TikTok account at risk? Could I get banned? There's no account involved to put at risk β it never logs in, follows, likes, or touches any account. Nothing is tied to your identity, so there's nothing to flag or ban.
What data does it actually access? Only publicly visible fields TikTok returns for public videos β the same information anyone browsing TikTok logged-out can see. No private accounts, no DMs, no personal data beyond public creator handles.
Will it get blocked / rate-limited? It calls TikTok's own signed endpoints the way a normal web client does, with pacing built in, so it behaves like ordinary traffic rather than aggressive scraping. Runs are resilient (automatic retries on transient hiccups).
Where does my scraped data go? Into your own private Apify dataset. It isn't shared with anyone; you export or delete it whenever you want.
β FAQ
Is this TikTok AI? How do I tell if a TikTok video is AI-generated?
Run the video's keyword or creator through this actor β it returns TikTok's own AI label for each video. is_ai_generated is true when TikTok considers it AI-generated; ai_disclosed means the creator labeled it; ai_detected_undisclosed means TikTok's systems flagged it as AI even though the creator never labeled it. It's the fastest way to check if a TikTok is AI without eyeballing pixels.
What's the best TikTok AI detector?
Most "AI detectors" analyse pixels or text and guess. This TikTok AI detector reads the label TikTok itself already assigned to the video (aigcLabelType / moderationAigcLabelType), so it's accurate for TikTok content and β uniquely β flags undisclosed AI other detectors miss.
How do I detect undisclosed AI content on TikTok?
Turn on the analysis and filter for ai_detected_undisclosed = true, or read the per-query ai_undisclosed_pct. Those are the videos TikTok's moderation flagged as AI-generated where the creator set no AI label β unlabeled AI slipping through.
How do I find AI-generated TikToks in a niche or hashtag?
Enter the niche/keyword and turn on AI-generated videos only (aiOnly). You get a clean dataset of every AI-generated TikTok for that topic, with quality scores and the creators posting them.
How do you detect AI-generated TikToks?
We don't guess β we read TikTok's own labels. aigcLabelType is what the creator disclosed; moderationAigcLabelType is what TikTok's own systems detected. When the second fires without the first, the video is AI that wasn't disclosed.
What is the quality score? TikTok's internal video-quality score (0β100) attached to the video object β the quality signal its ranking system uses. We expose it raw plus a high/medium/low band, alongside resolution, bitrate, codec, and HD.
Can I get only AI videos?
Yes β turn on AI-generated videos only (aiOnly) to return just the flagged ones.
Does it work on the trending feed? Yes β enable Trending videos to analyse the live For-You feed with no keyword: a real-time read on how AI-heavy and how high-quality what's going viral is.
Is this a third-party AI detector? No. Generic AI detectors analyse pixels/text and guess. This reports the label TikTok itself assigned, which is why it can flag undisclosed AI that a pixel-detector would miss or mislabel.
Do I need a login, proxy, or seed URL? None. It signs TikTok's own public endpoints directly.
Is it legal? It reads only publicly visible fields TikTok returns for public videos. You're responsible for complying with TikTok's Terms and applicable law.
β οΈ Unofficial. Not affiliated with, endorsed by, or sponsored by TikTok or ByteDance. Reads only publicly visible data. Respect TikTok's Terms of Service and applicable law.