Product

Extract Product Review Insights with AI

Pros, cons, feature requests, rating — review text into feedback-board fields, with experienced-vs-wished kept strictly apart.

Overview

Review mining fails in one specific way: the model blends what the customer experienced with what they wish existed. This setup keeps them apart structurally — pros and cons hold experiences, feature_requests holds wishes — and the source guidance states the distinction outright. rating extracts the number on its stated scale, would_recommend answers true or false only from the text, and every list field deduplicates near-identical points. Conservative inference lets an obvious implied recommendation through while blocking sentiment invention.

Workflow

One review per extraction
Per-review records keep traceability — every con links back to the review that said it.
Check the experienced/wished split
"The strap feels flimsy" is a con; "a silent mode would be great" is a feature request. The guidance draws the line; the fields enforce it.
Average ratings only on stated scales
The rating rule keeps the number as given (4 of 5 stays 4) — normalize scales in your own pipeline, knowingly.

Why This Works

Structural separation beats asking the model to "distinguish" in prose
Deduplicated lists stop one repeated complaint from looking like five
Text-grounded true/false on would_recommend resists sentiment hallucination

Best for

Product teams mining reviews at volume
Feedback pipelines that route bugs and wishes differently
Reviews where the wish is phrased like a complaint

Not for

Scoring sentiment positive/negative/neutral — that's classification, not extraction
Aggregating across thousands of reviews — extract per review, aggregate in your tooling

Use cases

Feeding a feedback board from app-store or marketplace reviews
Separating real complaints from feature wishes automatically
Collecting ratings on their stated scale for honest averaging

Customize This Resource

Opens this setup in Extraction Prompt Generator. Generate to get the full extraction prompt — then adjust the fields, missing-data behavior, and ambiguity policy.

Open in Extraction Prompt Generator

Prompt Template

TASK
Extract structured insights from the customer review in the input for our product feedback board.

SOURCE
The input is a customer product review.
Reading guidance: Separate what the customer actually experienced from what they wish existed — those go to different fields.

FIELDS TO EXTRACT
- product_name (required): Product being reviewed
- rating (optional): Numeric rating if the reviewer gives one
- pros (required): What the reviewer liked
- cons (required): What the reviewer disliked
- feature_requests (optional): Things the reviewer wished the product did
- would_recommend (optional): Whether the reviewer would recommend it

EXTRACTION RULES
- product_name: The product's display name as written.
- rating: Extract the numeric value; if a scale is stated (e.g. 4 of 5), use the stated number.
- pros: Extract every distinct item as a separate list entry; deduplicate near-identical ones.
- cons: Extract every distinct item as a separate list entry; deduplicate near-identical ones.
- feature_requests: Extract every distinct item as a separate list entry; deduplicate near-identical ones.
- would_recommend: Answer true or false based only on what the text states.
- Extract values from the source text only — do not use outside knowledge to fill fields.
- Do not summarize or interpret the document — return only the fields listed above.

MISSING DATA
- If a field is not found in the source, include it with an empty string "".
- Never invent or guess a value for a missing field.

AMBIGUITY POLICY
- Prefer explicitly stated information; infer a value only when the text strongly implies it.
- Keep any inference minimal — one short step from the text, never a chain of assumptions.

OUTPUT FORMAT
Return the extracted data as a single valid JSON object using the field names below.

EXAMPLE OF A VALID EXTRACTION
{
  "product_name": "Aurora Desk Lamp",
  "rating": 4,
  "pros": [
    "Battery lasts two full days"
  ],
  "cons": [
    "The strap feels flimsy"
  ],
  "feature_requests": [
    "A silent mode for night use"
  ],
  "would_recommend": true
}

OUTPUT RULES
- Return only the JSON object — no text before or after it.
- Do not wrap the output in markdown code fences.
- Use the field names exactly as defined — no renaming, no extra fields.

Related Resources

Explore more resources that work well with this one.

Used in Playbooks

Complete workflows that include this resource as a step.

AI Customer Feedback Analysis Workflow Workflow · 3 steps

Tip: Save time by exploring related resources and tools that integrate with this workflow.

Explore all resources