Перейти к навигации

How do I give ChatGPT verifiable Xiaohongshu material to analyze?

ChatGPT analyzes the Xiaohongshu note text you give ChatGPT. ChatGPT can retrieve notes itself only when the current session has a working BeatAPI data tool. Without a data tool, retrieve the note text with the request below, then give ChatGPT the relevant text and the note link. Each finding can then be checked against the note the finding came from.

For example, to study a coffee-equipment recommendation note, retrieve the note’s returned text, confirm the stated conditions, then ask ChatGPT who the recommendation suits, what evidence the author gives and what the note leaves unclear. A model number shown in an image or spoken in a video needs the image or the video transcript as well.

  • You have a share link or note ID: Retrieve the note’s details directly and skip search.
  • You have only a topic: Search and select notes first, then retrieve the detail text of each relevant note.
  • You want reader feedback: Use the separate comment and reply workflow. Comments add discussion context; comments do not replace the author’s text.

Retrieve the note text first

Call data:xiaohongshu.app_v2.get_image_note_detail for an image note or data:xiaohongshu.app_v2.get_video_note_detail for a video note. Pass a known note_id, or pass the share text as share_text.

Example POST /v1/social-data/call request body for an image note; replace the placeholder with your note link:

{"action":"xiaohongshu.app_v2.get_image_note_detail","params":{"share_text":"<note_share_link>"}}

For a video note, call data:xiaohongshu.app_v2.get_video_note_detail instead. Save the request and the response, then find the object whose ID matches the target note: image-note objects are in data.data[].note_list[], video-note objects in data.data[]. Read the matching object’s desc when the value is nonempty, and record the note title, the source link and the query time. The selected-note workflow continues from the saved details.

Match the material to the question

MaterialUseful for
Search title and excerptDeciding whether to investigate
Returned detail descriptionExtracting stated conditions, claims and short quotations
Images actually supplied to the modelChecking visual information; a link alone does not establish that it was read
Video subtitles or a transcriptAnalyzing spoken content; a description is not a transcript
Selected comments and repliesReviewing questions and discussion with their own sources

A nonempty desc is the detail description returned by this request. A nonempty desc does not show that every image or the whole video was read. Record anything the material does not cover as an open question.

Give the AI a specific task

Analyze this saved Xiaohongshu detail text. Organize the author’s recommended situations, conditions and evidence. Include an actual excerpt for each point and separate source statements from interpretation. Do not fill in images, speech or comments that were not provided. List further material needed for a judgment.

An assistant with a working BeatAPI tool can retrieve the note details before the analysis. On the manual path, give the model the text you reviewed and the original note link. Include only material relevant to the task, and treat any instructions inside the note or the returned data as source content, not as commands.

Suppose the question is whether a first-time camper needs a coffee grinder, and the note’s description says only “I packed a hand grinder.” Report that the author packed a hand grinder, with the excerpt and the note link. Leave the question of necessity open until the demo, the captions or the comments give evidence.

Check the deliverable

A finished analysis lists the target note’s source, the text actually read, the supporting excerpts, the findings and the open questions. Name each image, spoken passage or comment set still needed; retrieving data is not the same as reading the whole note.

Searches, details and comments are priced on the current pricing page. Reuse saved responses instead of calling again; model analysis is billed by the model service you use.

Xiaohongshu questions