> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.beatapi.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.beatapi.io/_mcp/server.

# How do Zhihu comments help verify an answer's conditions, corrections and counterexamples?

> Read selected-answer comments and relevant replies, preserving original claims and context before refining the research conclusion.

**Write down the answer's specific claim first, then read the answer's comments with `zhihu.web.fetch_comment_v5` for added conditions, counterexamples or corrections of the object discussed.** Useful comments show whom an experience applies to and which material is still missing.

For example, under an answer claiming that home projectors work during the day, comments may reveal that the room was shaded, that the answer discussed a different product generation, or how the author actually set up the projector. Keep the original answer and the discussion relationships before updating your research notes.

## Select an answer that needs clarification

Reuse a verified answer ID. Read comments first for answers whose claim lacks environment details, whose evidence is unclear, whose comments mention a model difference or whose author explains further in the comments. Simple agreement and unrelated arguments can stay as background without reply requests.

Pass the `answer_id`, not the question ID, to `zhihu.web.fetch_comment_v5`. `limit` and `offset` are strings:

```json
{
  "action": "zhihu.web.fetch_comment_v5",
  "params": {"answer_id": "<confirmed_answer_id>", "order_by": "score", "limit": "20", "offset": ""}
}
```

Send the JSON to `POST https://api.beatapi.io/v1/social-data/call` with normal BeatAPI Bearer authentication. `order_by: score` returns popular comments first; `order_by: ts` returns the newest first. To review newer additions, save a separate observation with the applied ordering and collection time, and match repeated comments by actual comment identifier.

## Which replies deserve expansion?

Expand a comment's replies when the replies could change how the answer reads, for example “That is a different model” or the author explaining that the curtains were closed. Confirm the selected comment's identifier, then call `zhihu.web.fetch_sub_comment_v5`:

```json
{
  "action": "zhihu.web.fetch_sub_comment_v5",
  "params": {"comment_id": "<confirmed_comment_id>", "order_by": "score", "limit": "20", "offset": ""}
}
```

Keep the parent comment, the answer ID and the question. Because `zhihu.web.fetch_comment_v5` and `zhihu.web.fetch_sub_comment_v5` return raw platform payloads without fixed top-level paths for the comment array, text, author, or continuation token, inspect the saved result to locate the actual comments. For the next batch, use the continuation values the response actually returned; do not copy the answer list's integer `offset` or `cursor`.

## How do you update the research table?

The table below illustrates how comment findings update a research conclusion:

| Discussion                                     | What to preserve                                        | Effect on the conclusion                                |
| ---------------------------------------------- | ------------------------------------------------------- | ------------------------------------------------------- |
| Author adds that blackout curtains were closed | Original claim and sourced clarification                | Record the shading conditions explicitly                |
| A reader reports different results             | Equipment and environment, with missing conditions open | Keep a separate experience pending comparability checks |
| A product generation is corrected              | Original statement, correction and follow-up            | Verify objects before grouping                          |
| Participants repeat their positions            | Representative excerpts and discussion relationships    | Reply counts alone do not establish evidence strength   |

In the worksheet, record the original claim, the exact comment and reply wording, the object, the conditions, the author identity when confirmed, the source, your interpretation and the open items. Judge each comment's claim by verifiable evidence; if a comment cites external material, read the cited source before writing a confirmed conclusion.

## What should the team receive?

Sort the findings into three groups: clarified conditions, counterexamples that still need verification and corrections of the object discussed. Attach the original answer and comment context to each finding, and state which part of your summary the finding changed.

For example, your notes say “The projector works during the day.” If the author replies that the blackout curtains were closed, change the note to “The author tried daytime viewing with the blackout curtains closed.” Put the answer link and the clarification on the same row, and move “performance with the curtains open” to the follow-up column. Record which comment pages and replies you read, and keep any failed pages on the follow-up list.

Prompt for the assistant:

> Read this projector answer's comments for conditions, model corrections and concrete counterexamples. Expand relevant replies where needed, preserving the claim, parent comment and author additions. Identify conclusions needing conditions or further verification, with original excerpts and sources.

Budget the selected answer-comment pages and reply pages separately with current [pricing](https://beatapi.io/pricing). Reuse saved material, and expand only for a concrete question.

[Question-answer research](/guides/zhihu-question-archive) · [Objects and identifiers](/faq/zhihu-object-identifiers)