TikTok Search Scraper

Search TikTok by keyword and get the results as data. One row per video, with the caption, who posted it and how many followers they have, the view, like, comment and share counts, the sound, the hashtags and the media URLs.

Give it as many search terms as you like, one per line. It searches each one the way the site does and walks the results until it has what you asked for or TikTok stops serving.

Results are regional, and that is not a detail

TikTok returns different videos to different countries for the same words. A search for a product term from the US and from the Netherlands returns two almost entirely different sets, in two different languages. The Country and Language inputs control this, and every row records which term produced it, so a run across several markets stays sortable rather than silently mixed.

If you are researching one market, set both. If you are comparing markets, run the same terms once per country and group by the country afterwards.

What you get per video

• description, created_at, url, video_id
• handle, nickname, author_id, author_verified
• follower_count, author_video_count
• play_count, like_count, comment_count, share_count, collect_count
• duration_seconds, width, height, cover_url
• play_url and download_url, the site's two media addresses
• music_title, music_author, music_id
• hashtags, is_ad
• search_term, search_page, position

follower_count next to play_count is the pairing most people are after. A video with 400,000 views from an account with 2,000 followers is a different signal from the same view count on an account with two million, and both come back on the same row so you never have to join them.

The two media addresses behave differently from ordinary links: they are authorised by the session that produced them, so pasting one into a browser returns an error. If you want files rather than statistics, use TikTok Video Downloader, which fetches and hosts them for you.

How many videos you can get

TikTok serves roughly 200 unique videos per keyword and then stops. Asking for more returns what exists rather than failing, and the run summary reports how many came back. If you need more coverage on a topic, use more terms: several related phrases return overlapping but distinct sets, and together they go well past what any single phrase will give you.

Videos come back around 30 to a page, so a 60-video term is two page loads and a 200-video term is about seven. A run of five terms at 60 each finishes in well under a minute.

Errors

A term that returns nothing becomes an entry in the errors record rather than a silently missing set of rows:

Code	Meaning
bad_input	No search terms supplied
no_results	The search ran and matched no videos
refused	TikTok declined to serve that page of results
signer_unavailable	The run could not prepare itself; rerun

no_results and refused are deliberately different. A term nobody has posted about and a request TikTok declined look identical if you only count rows, and only one of them is worth rerunning.

Run settings

The default memory of 2 GB is what this needs; it is not padding. Lower it and runs may fail to start. Timeout defaults to an hour, which is far more than a normal run takes.

Related actors

For a specific account's videos rather than a keyword search, use TikTok Scraper - Profiles, Videos & Hashtags. For the video files themselves, use TikTok Video Downloader. For shop listings and prices, use TikTok Shop Scraper, and for advertising, TikTok Ads Library Scraper.