Finish with a short list of audience questions, supporting comment IDs, and one action for each: a reply, a product FAQ, or the next video. The spreadsheet helps you make that decision; converting a CSV is only one step.
1. Choose the decision before collecting comments
A long list of reactions is hard to act on. Write down one question first. For a small shop showing a reusable bottle, a useful starting point is: “Which details are viewers asking for before deciding whether to buy?”
For a creator, the question might be “Which part of this demonstration needs a follow-up?” For a product team, it might be “Which use case are people struggling with?” Keep one primary goal for the first pass so you can separate relevant evidence from jokes, emoji and unrelated discussion.
Use a public video whose topic matches that goal. Your own video can tell you what your audience still needs explained. A competitor’s public video can suggest questions to investigate, but its audience and offer may be different from yours.
The bottle conversation below is fictional. Its five comments and one creator reply demonstrate the method and CSV structure. They are not retrieved TikTok data, a customer case study, or evidence of demand for a real product.
2. Find questions, then read their context
- Open TikTok Comment Viewer and paste the public video link.
Use the free TikTok Comment Viewer to read available comments without a TikTok account or app. The free workspace also needs no SearchComments sign-in. Check the video preview before you start.
- Read a broad sample before narrowing it.
Check how many comments are loaded and use Load more when available. A first batch is a subset of the conversation. Do not describe it as all comments or a random sample of all viewers.
- Search one topic at a time.
For the bottle video, try
ship,fit,dishwasher, andbuyseparately. Text search matches substrings, soshipalso finds “shipping.” Clear the previous term between searches. - Use Questions as a shortcut, then check the other rows.
The filter recognizes question punctuation and common question openings. It can miss indirect requests such as “wish you showed the size,” so also read without the filter. A repeated concern does not have to end with a question mark.
- Open relevant reply threads.
A reply may already answer the question. Search fields filter top-level comments, not reply text independently. Read the replies under matching parents to see whether the concern is resolved or needs a clearer explanation.
Author search matches the returned display name, rather than guaranteeing an exact @username. It is useful for narrowing loaded top-level comments by a remembered name; it does not reconstruct anyone’s comment history across TikTok. See the TikTok comment search guide for the full search workflow.
3. Download a useful sample and keep the context
When you have inspected the relevant threads, reset accidental search and date filters before downloading a broad research sample. The free basic CSV contains up to 100 current-view rows, including loaded replies belonging to matching parent comments. Replies count toward that limit. Check the download count; a source video’s visible comment total is a different measurement.
For a larger dataset, the TikTok Comment Exporter offers a separate account-credit job with a collection target, optional replies and field choices. Choose All available fields when reply relationships matter, because its parent_id column connects a reply to a top-level comment. Viewer filters do not carry into the paid job.
Import the CSV into Excel or Google Sheets. Keep an untouched Raw sheet, then work in a separate copy. Use UTF-8 and a comma delimiter if the import needs configuration. Preserve video_id, comment_id and parent_id as text so long identifiers are not converted or rounded.
Add four working columns beside the exported fields: Theme, Question or concern, Already answered?, and Next action. These are your manual notes; SearchComments does not automatically add these categories or provide an AI sentiment score.
If you use several filtered downloads, deduplicate by platform, video ID and comment ID before counting themes. Keep the video URL, capture time and collection scope in your research notes. Store only what you need, and avoid publishing commenter names in a shared report.
4. Practice with a six-row bottle-video example
Download the illustrative TikTok CSV. It has the same ten-column structure as an All available fields export: five fictional top-level comments and one fictional reply. Dates and permalinks are empty because there is no real source video.
| Comment ID | Comment | Theme and interpretation |
|---|---|---|
example-c1 | Do you ship this bottle to the UK? | Shipping: destination information is missing. |
example-c2 | Is shipping available to Canada? | Shipping: a second destination question. |
example-c3 | Does it fit a car cup holder? | Fit: read the reply before calling this unanswered. |
example-c4 | Can I put it in the dishwasher? | Care: check the actual cleaning instructions. |
example-c5 | Where can I buy this bottle? | Purchase path: make the buying instructions clearer. |
example-r1 | The bottle is 7 cm wide; measure your holder first. | Reply to c3: context for the fit question, not another viewer request. |
Filter to rows with an empty parent_id when counting top-level questions. In this example, shipping appears twice, while fit, care and the purchase path each appear once. The creator reply belongs to example-c3; counting it as another customer question would distort the result.
Mark the fit question as “reply present; verify answer.” Even in a real dataset, a comment saying a bottle is 7 cm wide would not prove the product measurement. Confirm product facts before putting them in a video or FAQ. The example tells you how to organize evidence, not what a real bottle’s specifications should be.
5. Turn each finding into a small, specific action
| Finding | Action | Why it follows |
|---|---|---|
| Two questions about shipping destinations | Add a destination FAQ and a short, accurate shipping explanation. | Both rows ask for information needed before ordering. |
| A fit question with a reply | Consider a follow-up showing verified dimensions and a fit demonstration. | A demonstration can make the reply easier to understand; it should not promise universal fit. |
| A cleaning question | Add the manufacturer’s verified cleaning instructions to the FAQ. | The comment exposes a use-and-care question, not proof of a product defect. |
| A question about where to buy | Review the video’s purchase instructions and reply with the appropriate path. | The viewer cannot find the next step; this still does not prove a sale will follow. |
A useful brief records the source comment IDs, what viewers asked, whether a reply already exists, the proposed action and who will verify the answer. Prioritize with both repetition and importance: one safety or compatibility question may deserve attention before several low-impact requests.
After publishing a response or follow-up, review a later batch for similar questions and compare samples with similar scope. Look for clearer conversations and resolved confusion. Comments alone cannot measure purchase conversion; connect any sales claim to your own confirmed sales or analytics data.
6. Keep conclusions proportional to the evidence
- Commenters are a subset of viewers. A few comments can reveal a useful question, but cannot establish what most viewers think.
- Loaded rows are a subset of available comments. Ordering, collection limits, moderation, deleted rows and missing replies affect the sample.
- Repeated rows are not independent evidence. Remove duplicate IDs and read context before counting a theme. Display names are not a reliable unique-person identifier.
- Likes and questions are signals to investigate. They do not prove sentiment, buying intent or product demand.
- Public research has boundaries. The tool cannot reveal private, deleted or held-for-review comments. Read the data methodology before describing coverage.
Start small: one video, one question and one useful follow-up. The value comes from turning a conversation into an answer people can use, rather than collecting the biggest possible file.