CSV · XLSX · JSON for analysis
Comments for research and sentiment analysis
Paste the post link, export, and load one row per comment into R, Python, NVivo, Excel or a sentiment model. Every file states how many comments were read next to the count the network reports. Free per post: Instagram and YouTube up to 1,000; TikTok and Facebook up to 300 comments.
The short answer
For a public post you do not run, the practical route is an export from the post link: no developer app, no API review, no login. CommentsToSheets reads the public comments and replies and gives you a file with one row per comment, ready for a sentiment model or manual coding. For accounts you own, the networks' official APIs are the other route, with their own setup and limits.
What the columns give an analyst
- comment: the text as written, emoji and line breaks kept. This is the column your sentiment model or codebook reads.
- posted_at_utc: a UTC timestamp for time series, event windows and before-and-after comparisons.
- likes: likes on each comment when it was read, a rough weight for agreement or visibility.
- is_reply, reply_to_comment_id, reply_to_username: the thread structure, so you can analyse top-level reactions apart from conversations, or rebuild reply trees.
- username, user_id: who wrote it, to count comments per person or remove one heavy commenter from a sample. Pseudonymise these before sharing a dataset.
- mentions, hashtags: already extracted, so tagging behaviour and topics need no regular expressions.
- platform: which network each row came from, so posts from several networks can sit in one dataset.
Full column list: #, comment_id, username, user_id, comment, posted_at_utc, likes, is_reply, reply_to_comment_id, reply_to_username, mentions, hashtags, platform.
Reporting completeness in your method
Every network's own comment counter includes comments it hides from everyone: spam-filtered, hidden words, restricted accounts, deleted comments. No tool can read those. The export page shows comments read next to the count the network reports, and the Read me sheet of the workbook records both numbers with the read time. Quote them in your method section ("2,184 of 2,240 reported comments were publicly readable on 3 November") rather than presenting the file as every comment ever written.
Above the free limit you get a labelled preview with both numbers in the file name; one payment for that post reads the rest: $4.99 up to 15,000, $9.99 up to 50,000, $14.99 up to 200,000 comments.
Formats for each tool
- CSV (UTF-8 with a byte-order mark):
pandas.read_csv, Rreadr::read_csv, SPSS, NVivo, Atlas.ti. - JSON: the post details and the comments as an array, for scripts and pipelines.
- Excel workbook: Comments, Commenters, Mentions, Per day, Hashtags and Read me sheets, for hand coding with filters.
- Google Sheets: Copy for Sheets, or a live
=IMPORTDATAlink a whole team can read.
What it cannot do
- Score sentiment for you. The file is the input; your model or codebook does the scoring.
- Read private accounts, personal Facebook profiles, groups, or comments a network hides from everyone.
- Collect by keyword or hashtag across a whole network. It works one post at a time from its link.
- Give demographics. Networks do not publish age, gender or location for commenters.
Ethics boards often ask how public data is stored: YouTube exports are deleted after 30 days under YouTube's terms, other exports after 30 days. Keep your own copy under your data plan.
Questions
Do I need an API key or developer account?
No. Paste the link of a public post. There is no API key, no login to the network and no account here.
How many comments can I download for free?
Per post: Instagram and YouTube up to 1,000; TikTok and Facebook up to 300 comments, every one of them. Bigger posts take one payment for that post: $4.99 up to 15,000, $9.99 up to 50,000, $14.99 up to 200,000 comments.
Can I compare several posts or networks?
Yes. Export each post; every file has the same columns and a platform column, so you can stack them into one dataset.
Is the timestamp in local time?
No, posted_at_utc is UTC, which keeps posts from different time zones comparable. Convert in your analysis if you need local time.
Does the file include replies?
Yes, each reply is its own row with the id and author of the comment it answers. On a free TikTok or Facebook export a long set of reply threads can be cut short, and the page says how many replies were not read.