Export TikTok comments: three ways that work

There is no export button on TikTok. If you want to export TikTok comments to a spreadsheet you have three options: copy them by hand, run a third-party extension or web tool over the video's URL, or ask an AI agent connected to TikTok's data to pull them and write the CSV for you.

The first is free and slow. The second is cheap and sits in a grey area of TikTok's terms. The third is the one we use, and it is the only one that also does the part that matters: turning 761 comments into six things your audience actually wants.

Below is each route with its real limits, then a worked example on a real video with 761 comments, coded into themes you can act on.

What TikTok gives you, and what it doesn't

TikTok has no comment export. Not in the app, not in TikTok Studio, not in the web dashboard. The only manual moves are long-pressing a comment to copy its text, and screenshotting.

The obvious workaround is Settings and privacy → Account → Download your data. It does not solve this. That export is your own activity: the comments you posted, your likes, your searches, your login history. It does not include the comments other people left on your videos, because those are other people's data. Request the file if you want your own comment history; do not request it hoping for your video's comment section.

TikTok comment scraper tools and extensions

These take a video URL and hand you a file. They are the fastest route if you need one video's comments right now and you are willing to paste a link into somebody else's website.

Prices and limits as published on each vendor's page, checked 19 September 2026. They change often.

Three things to know before you paste a URL. Browser extensions read the page you are logged into, so they run as you and can trip TikTok's rate limits on your own account. Web tools send the video to their server, so whatever they collect, they keep. And every one of these caps free use at a number chosen to be just below what a mid-size comment section holds, which is why 761 comments is an awkward figure.

Export TikTok comments with an AI agent

The route we use skips the browser. TokConnect is an MCP server, so Claude, ChatGPT, Cursor or Gemini can call TikTok's search and comment endpoints directly. The video_comments tool takes a video id and a count, returns up to 50 comments a page with the text, timestamp, reply count and the commenter's handle, and pages with a cursor until the section runs out.

That matters for one reason: the agent that fetches the comments can also read them. You get the CSV and the summary in the same reply, instead of a spreadsheet you will never open.

Two practical notes. The id in a TikTok URL after /video/ is the id the tool wants. And the first page you get back is TikTok's own ranking, so it is the top comments, not the oldest, which is usually what you want for theme work.

If you would rather write the loop yourself, the comments API guide has the request and pagination in Python.

The part nobody covers: what to do with them

Exporting is easy. Reading 761 comments is not, and a spreadsheet of raw text is not an insight. Here is the full method on one video.

We searched TikTok for car detailing sorted by most liked, and took a video from Active Auto Detailing (@activedetails) posted in June 2025: a mobile detailing van tour, 759,000 views, 45,300 likes, 761 comments. We pulled the first 50 comments and coded every one of them by hand.

Data: TikTok comments via TokConnect, pulled 19 September 2026. All 50 comments coded by hand. Open the snapshot, including the coding rules.

Three of the comments, shortened, with handles left out:

"How much for the whole setup + the van?"

"What do u do about hard water"

"how many vehicles can u do on a full water tank?"

Now read the table as a content plan. Half the section is noise, so ignore it. The other half is 25 comments that together say: tell us what it cost, tell us what you bought, and tell us how the water and power actually work. That is three videos, and the creator already knows people will watch them, because they asked.

The cost question is the sharpest signal in the set. Three people asked it directly and seven more circled it with "I wish I could afford a setup like this" and "that's the dream". Ten of fifty comments are about money on a video that never mentions a price.

The method generalises. Code every comment into a theme, count, drop the praise, then rank what is left by count. For a brand, run the same pass over a competitor's video instead of your own and you get their unanswered objections for free. The competitor analysis guide uses this as one of its steps.

Handle the data properly

Comments are public, but they are still written by people with names attached.

• Strip handles before you share. Keep them in your working file if you need to reply. Take them out of anything you send a client or publish, as we did above.
• Quote short and quote accurately. A screenshot of somebody's comment on a slide is a person's words in a room they didn't enter.
• Don't build a profile. Counting themes is research. Following one commenter across videos to work out who they are is not.
• Re-pull rather than archive. Comments get deleted and edited. A file from March is a claim about March.

If you want the demand side of the same picture, comments tell you what viewers of one video asked, and search data tells you what everyone else is typing into the search bar. The audience insights guide covers who those searchers are. Used together they are the cheapest content research there is.