Steam Reviews Scraper

What does Steam Reviews Scraper do?

Collect public Steam reviews across multiple games for feedback analysis, reporting, or comparison. Get review text, recommendations, language, timestamps, helpful votes and playtime in a downloadable dataset.

Steam provides the underlying data for free. This Actor handles pagination, output limits and persistent duplicate suppression so you can continue collecting without receiving the same review again. No Steam API key or paid proxy is required.

How to use it

1. Enter one to twenty Steam app IDs, such as 570 or 730.
2. Choose a language and set the maximum number of reviews.
3. Set maxReviewsPerGame if you want to limit how much each game contributes.
4. Reuse the same stateStoreName to resume collection and suppress previously delivered reviews.
5. Download the reviews and check the OUTPUT coverage report for limits or errors.

{
  "appIds": ["570", "730"],
  "language": "all",
  "maxReviews": 100,
  "maxReviewsPerGame": 50,
  "maxPagesPerGame": 3
}

Input options

Option	Behavior
appIds	One to twenty Steam app IDs.
language	Steam language code, such as english; defaults to all.
maxReviews	Maximum new reviews across all games per run: 1–1000, default 100.
maxReviewsPerGame	Maximum new reviews per game: 1–1000. When omitted, equals maxReviews.
maxPagesPerGame	Maximum pages per game per run: 1–20, default 3.
pageSize	Reviews requested per page: 1–100, default 100.
stateStoreName	Persistent deduplication and resume store; defaults to steam-reviews-poc-ledger.

Games are processed in input order. A small global limit can leave later games untouched. Limits are ceilings, not promises of a particular number of results: empty games, upstream errors or previously collected reviews can reduce output.

Output

Each review includes appId, reviewId, text, language, recommended, creation and update timestamps, helpful votes, playtime, a source URL and a content hash. Optional data may be missing. Download the dataset as JSON, CSV, Excel or HTML.

Synthetic example — illustrative data, not a live Steam review:

{
  "appId": "570",
  "reviewId": "100",
  "text": "Strategic team play.",
  "language": "english",
  "recommended": true,
  "playtimeMinutes": 120,
  "votesHelpful": 3,
  "coverageComplete": false
}

The OUTPUT report includes per-game cursors, stopping reasons, duplicate and changed-review counts, and uncertain deliveries. Errors and empty results are not review rows. fetchedAt is the response fetch time, not a guarantee that Steam's data is fresh. coverageComplete remains false because collection is bounded and does not represent a complete snapshot taken at one instant.

Pricing and cost example

USD/result: 0.0002

The listed rate is $0.0002 per unique useful review delivered. Consult the Apify Store Pricing tab for the currently applied price and platform charges. There is no separate Actor start, per-game or empty-result fee.

On November 4, 2026, our own two tests collected 100 reviews each, with cloud infrastructure costs of approximately $0.012 per test run. These observations are from our own tests only; they are not evidence of customer revenue, sales, profit or customer costs. Net profit margin has not been measured. Actual costs vary with workload, retries and storage use.

Resume and duplicate suppression

Only one run may write to a given stateStoreName at a time. Concurrent writers are not supported. Different stores have separate review histories; changing or deleting a store loses duplicate suppression across runs.

Review IDs are remembered across runs and languages. An already reserved review is not automatically delivered again, even if its content changes. Content changes are counted rather than emitted as new reviews.

Resume continues from the saved cursor for each game and language. Partial pages retain their cursor so remaining reviews can be collected later. After an empty terminal page, the next run starts from the first page while retaining the review history. This is continuation of an existing traversal, not fresh daily tracking.

Data and reliability limits

Responses are cached for ten minutes, and Steam may also return cached data. Cache expiry does not guarantee freshness. Rate limits can pause collection until a later run; upstream failures and incomplete coverage appear in OUTPUT.

To avoid automatic duplicate delivery, each review is reserved before it is written. If delivery or charging cannot be confirmed, its reservation stays pending and is not replayed automatically. This can omit a review after a crash or an uncertain write. Delivery, charging and state updates are not one transaction, so exactly-once delivery is not guaranteed.

State and cache storage grow with use. Long-term retention needs and large-scale costs have not been measured. Reviews reflect publicly available user submissions; availability, completeness and accuracy depend on Steam and its reviewers.