TikTok trend research

Produce a dated, sample-aware view of what TikTok is surfacing for a topic, hashtag, category, country, or period. Distinguish TikTok's different trend surfaces: a recommended feed, Creative Center rankings, hashtag activity, and Explore are not interchangeable measurements.

Setup and requests

Set CRAWLORA_API_KEY to a key from crawlora.net, then run the helper bundled with this skill. It sends x-api-key to https://api.crawlora.net/api/v1; keep the key in the environment. Read reference/endpoints.md for exact parameters.

Set `CRAWLORA_API_KEY` in the environment before running the helper.
scripts/crawlora.sh /tiktok/creative-center/hashtags country_code=US period=7 | jq '.'
scripts/crawlora.sh /tiktok/creative-center/videos \
  country_code=US period=7 sort_by=engagement organic_only=true | jq '.'

Inspect both HTTP success and the JSON envelope, including nested upstream code values. A 2xx response with no usable data is not automatically a successful trend observation.

Workflow

1. Clarify the topic or hashtag, country, lookback period, surface, requested sample size, and whether paid content should be excluded. Record retrieval time because trend results change.
2. Choose the matching surface:
   • /tiktok/trending for TikTok's current recommended feed; it has no geography or historical-period control.
   • /tiktok/creative-center/hashtags for ranked hashtags by country and period (7 or 30 days).
   • /tiktok/creative-center/videos for ranked videos by country and period, sorted by views, engagement, or six_second_views; use organic_only=true when the brief is organic-only.
   • /tiktok/search for keyword-matched videos when the brief starts from a topic rather than a known hashtag.
   • /tiktok/search/hashtag to discover candidate hashtags, then /tiktok/hashtag/{name} and /tiktok/hashtags to resolve a hashtag id and inspect its listed videos.
   • /tiktok/category followed by /tiktok/explore/{id} for a category feed. Do not describe an Explore feed as a global trend ranking.
3. For a topic brief, use search/hashtag discovery first, then validate a bounded set of returned video ids with /tiktok/post/{id}. Use identifiers returned by the API; do not invent ids from captions or display names.
4. Compare countries or topics only when the surface, period, sort, and filters match. Deduplicate by stable hashtag or video id and stop at the requested bound or the returned pagination terminal state.
5. Report the observed ranking/activity, source links or ids, filters, sample size, and retrieval time. Separate the measured signal from explanations or content recommendations.

Interpretation and limits

• Anonymous Creative Center access is partial: hashtag requests return at most three records, while video requests return only page one (four videos), even when the response reports a larger upstream total. Do not call that total accessible coverage.
• Creative Center video availability is uneven. US, JP, ID, VN, and TH have reliably returned populated samples; a valid empty result in another market can be genuine upstream no-data. Do not retry it indefinitely or claim an outage without corroboration.
• engagement_rate and six_seconds_vtr are ratios, not percentages. Keep period-scoped and lifetime fields separate. Counts and ranks are surface- specific and do not establish reach across all TikTok.
• Hashtag viewCount/videoCount and search results reflect TikTok's current index. They are not a complete historical archive. A search match is not proof that every returned video is about the intended brand or topic.
• Missing fields are unknown, not zero. A valid empty sample is an observation to report with its filters, not a reason to fabricate a conclusion.
• Do not claim to have watched a video's hook, heard its audio, or judged its visual format unless the environment actually exposed the media. Signed media URLs may expire.

Reliability and output

Use bounded pagination. On 429, back off; retry one transient 5xx when useful. Stop on 401/403 and report the access limitation. Preserve partial results when enrichment fails, labeling which observations are missing.

Return a concise table or brief with surface, topic/hashtag, country, period, rank or metric, source URL/id, retrieval time, and coverage caveats. Call patterns “observed in this sample”; label forecasts, content ideas, and causal claims as hypotheses.