One object per line.
No conversion step.

JSON and JSONL describe the same rows differently.

A JSON dataset is usually an array with square brackets and commas. JSONL (also called NDJSON) writes each object on its own physical line. A review containing a line break keeps it escaped inside its string.

{"type":"review","text":"First line\nSecond line", ...}
{"type":"review","text":"Another review", ...}

The shortened rows above explain the layout; the ellipses are not valid JSON. Download the complete runnable fixture with one summary and three review rows.

Blank lines, LF or CRLF endings, and a leading UTF-8 BOM are tolerated. The browser accepts .json, .jsonl and .ndjson; pasted input is detected automatically. Keep a UTF-8 copy of the original snapshot.

Try the same snapshot in two ways.

1. Open the browser-local checker and select Load passing JSONL demo, or choose your local file.
2. Keep Complete export checked when you expect both the summary and all review rows. Select Check export locally.
3. The synthetic fixture returns PASS (JSONL): 1 summary row, 3 review rows. Download the result JSON if you need an audit record.

For CLI or CI, extract the ZIP and run this from its folder:

python3 validate_output.py demo-export.jsonl --json --require-reviews
# Exit 0: ok:true, summaryRows:1, reviewRows:3

python3 validate_output.py your-export.ndjson --json --require-reviews > validation.json
# Keep the process exit code: 0=passed, 1=data failed, 2=input error.

The check does not fetch new reviews or start a paid Actor. The browser keeps file contents and results in the tab without upload or browser storage. Loading a demo fetches only that public fixture. Normal first-party page and CTA counts remain enabled.

Browser limits: 10 MiB, 20,000 rows, safe integer IDs/counts, conventional ISO dates and a current browser supporting JSON numeric source types. Use Python 3.10+ for larger exports, unusual dates or CI.

Export the whole snapshot before checking completeness.

Apify’s dataset documentation lists JSONL among its export formats. In the Console, open the existing dataset, choose Export, select JSONL and download the full dataset. Keep both TanLabs row types and all required fields; a Reviews-only view or a selected-field export may remove the summary or evidence fields needed by this checker.

For an existing dataset, the dataset items API supports format=jsonl. The request shape is GET /v2/datasets/{DATASET_ID}/items?format=jsonl. Account access and dataset visibility still apply; use Apify’s documented authentication and never put credentials in a public report. Downloading an existing dataset does not start a new Actor run; platform charges remain governed by Apify’s current plan.

A complete check needs the full dataset rather than just one API page. Preserve app IDs, country codes, titles, text and rating fields. If you deliberately keep only summaries, use partial mode and label source provenance as unchecked.

Fix the line, then check the report.

Whole-document JSON is tried first, so pretty-printed JSON arrays and {"items": [...]} wrappers continue to work. Escaped newlines inside review strings remain intact.