How to Export Amazon Reviews to JSON

When JSON is the better format

Choose JSON when a script, database, notebook, data pipeline, or analysis tool benefits from explicit types and nested structures. CSV is easier for most spreadsheets; JSON is clearer when you need arrays such as image links, distinguish missing values from empty text, or inspect job-level metadata separately from review records.

Use case	Why JSON helps
Python or JavaScript analysis	Objects and arrays can be parsed without flattening.
Database ingestion	Job metadata can be stored separately from review records.
AI file analysis	Field names and source context remain explicit.
Quality monitoring	Warnings and errors retain code, message, and recoverability structure.

Export steps

Open an Amazon product page, reviews page, or enter an ASIN manually.
Select JSON in the INCIO Reviews popup.
Choose the requested count, rating filter, and sort option.
Start the user-initiated export and wait for collection to finish or stop safely.
Open the downloaded file in a code editor or import it into your analysis environment.

How the JSON export is organized

Schema version 1.2 places the schema version at the top level, keeps job metadata in a meta object, preserves warnings and errors, and stores review records in a reviews array. Review records can include text, rating, author display name, date, verified status, helpful count, variation, identifiers, URLs, image-link arrays, marketplace context, and source page number.

Metadata records the request mode, source URL when available, requested, available, target, and collected counts, the 100-review extension limit, processed pages, selected filters, completion reason, dataset completeness, and timestamps. A downstream workflow should inspect those fields before assuming the dataset is complete.

Inspect a schema 1.2 example

Synthetic JSON exportTwo invented review objects with complete job metadata.Download sample JSON

Field-by-field referenceReview fields, completion values, warnings, and compatibility.Open schema reference

Read schema_version first, then validate meta,warnings, and errors before iterating overreviews.

Minimal JavaScript import

Parse the file, reject an unsupported schema version, check the final export status, and only then pass the reviews array to a downstream transform. Keep the original file with the analysis output so warning codes and collection scope remain available for audit.

Minimal Python import

Load the object with a JSON parser, inspect data["meta"], and build a table from data["reviews"]. An empty array and a null field are different states: the first is a known empty collection; the second means a scalar value was unavailable.

Quality checks before analysis

Confirm export_status and completion_reason.
Check source_dataset_completeness before generalizing.
Review warning and error codes instead of silently ignoring them.
Deduplicate by review ID only when the ID is present.
Keep nulls as unknown values rather than converting them to false facts.
Use source page number and review URL when auditing a finding.

Metadata check	Decision
COMPLETED + LIMITED_BY_REQUEST	The requested scope was reached; do not imply all source reviews were collected.
COMPLETED + LIMITED_BY_EXTENSION	The 100-review product limit defined the result.
PARTIALLY_COMPLETED	Use collected rows only after reading the final warning and disclosing the partial scope.
LOGIN_REQUIRED or BLOCKED_BY_CAPTCHA	Do not retry automatically; resolve the source-page state manually.
Selector or missing-field warning	Inspect affected records before trusting absent values.

Rules for a reliable downstream pipeline

Store the raw JSON unchanged and derive transformed data separately.
Key each import by job_id and retain the schema version.
Use review ID for deduplication only when it is non-null.
Partition by marketplace, ASIN, and variation before category-level comparison.
Propagate warning and error objects into logs or analysis notes.
Do not make a missing optional field fail the entire import.
Validate future schema versions before assuming the same field contract.

Local export and responsible handling

The JSON file is created locally and is not sent to INCIO. It may contain public reviewer display names, review text, identifiers, and URLs. Limit unnecessary sharing, remove fields you do not need, follow applicable law and platform terms, and review the privacy terms of any service to which you upload the file.

Does INCIO provide a JSON API?

No. Version 1.0 creates a local downloadable file after a user-initiated export. It does not send the dataset to an INCIO endpoint and does not provide unattended cloud collection.

Can JSON be uploaded directly to an AI tool?

Often, but first remove fields you do not need and review the provider's data terms. Require the model to cite review IDs and verify important output against the original file.