Steam Review & Comment Intelligence Scraper

Steam reviews are where players tell you exactly why they buy, refund, rage, recommend, complain, and keep playing.

This actor scrapes public Steam game reviews/comments and turns them into voice-of-customer intelligence: sentiment, complaints, praise, bugs, feature requests, pricing objections, opportunity scores, summaries, and recommended actions.

Use it with the Steam Market Intelligence actor: first find interesting games, then mine their reviews to understand the market’s actual language.

What this is really for

Use it to answer questions like:

• Why do players love this competitor?
• What are players complaining about repeatedly?
• Are negative reviews caused by bugs, pricing, missing content, balance, UX, or community issues?
• Which comments contain feature requests?
• Which reviews are useful enough for roadmap, ad copy, landing pages, or competitor teardown?
• What exact words do players use to describe value, frustration, fun, grind, bugs, and pricing?

Best use cases

• Competitor teardown — pull negative reviews from rival games and find recurring weaknesses.
• Game positioning — mine positive reviews for the exact language players use when they love a game.
• Feature research — detect “wish it had…” and “please add…” patterns.
• Bug/performance research — cluster crash, lag, stutter, and unplayable complaints.
• Pricing research — surface refund, overpriced, DLC, microtransaction, and paywall objections.
• AI/RAG datasets — every row can include clean markdown ready for LLM workflows.

What you get

Core Steam review fields

• App ID
• Game title
• Recommendation/review ID
• Review URL
• Review text
• Recommended / not recommended
• Language
• Created and updated timestamps
• Author Steam ID
• Playtime forever
• Playtime at review
• Playtime last two weeks
• Helpful votes
• Funny votes
• Weighted vote score
• Comment count
• Steam purchase flag
• Received-for-free flag
• Early-access review flag
• Steam Deck flag

Comment intelligence fields

• sentimentScore
• sentimentLabel
• painScore
• praiseScore
• usefulnessScore
• opportunityScore
• matchedTopics
• complaintSignals
• praiseSignals
• bugSignal
• featureRequestSignal
• pricingSignal
• toxicitySignal
• commentSummary
• recommendedAction
• ragMarkdown

The intelligence layer

sentimentScore and sentimentLabel

Combines the Steam recommendation flag with positive and negative language. Labels each review as positive, mixed, or negative.

painScore

Finds reviews that reveal useful pain: bugs, crashes, lag, pricing objections, lack of content, bad balance, bad UX, cheating, toxic community, refund language, and “not worth it” complaints.

High pain is useful for competitor displacement research.

praiseScore

Finds reviews that explain why players love a game: fun core loop, strong value, polish, multiplayer/friends, atmosphere, content depth, replayability, and smooth performance.

High praise is useful for positioning and copywriting.

usefulnessScore

Ranks comments by research value using length, topic coverage, helpful votes, and player playtime.

A short meme review is less useful than a detailed review from someone with 100+ hours.

opportunityScore

The headline score. Prioritizes reviews that contain useful pain, high detail, topic signals, helpful votes, and feature-request language.

Sort by this when you want the comments most worth reading first.

Topic and signal detection

The actor detects topics like:

• performance
• bugs
• multiplayer
• balance
• pricing
• content
• controls / UX
• graphics / audio
• modding

It also labels complaint and praise patterns so you can quickly cluster feedback.

Example input

{
  "appIds": ["730", "230410"],
  "maxCommentsPerGame": 100,
  "filter": "recent",
  "reviewType": "all",
  "purchaseType": "all",
  "language": "english",
  "includeIntelligence": true
}

Recommended presets

Mine competitor weaknesses

{
  "appIds": ["730"],
  "maxCommentsPerGame": 300,
  "filter": "recent",
  "reviewType": "negative",
  "language": "english",
  "includeIntelligence": true
}

Mine player praise for positioning

{
  "appIds": ["230410"],
  "maxCommentsPerGame": 300,
  "filter": "recent",
  "reviewType": "positive",
  "language": "english",
  "includeIntelligence": true
}

Build a feature-request dataset

{
  "steamUrls": ["https://store.steampowered.com/app/3124540/Far_Far_West/"],
  "maxCommentsPerGame": 500,
  "filter": "all",
  "reviewType": "all",
  "language": "english",
  "includeIntelligence": true
}

Example output

{
  "appId": 730,
  "gameTitle": "Counter-Strike 2",
  "recommendationId": "123456789",
  "isRecommended": false,
  "reviewText": "The game is fun but the cheating and stutters make it hard to recommend.",
  "playtimeForeverHours": 248.4,
  "votesUp": 12,
  "sentimentLabel": "negative",
  "painScore": 82,
  "praiseScore": 18,
  "usefulnessScore": 75,
  "opportunityScore": 79,
  "matchedTopics": ["performance", "bugs", "multiplayer"],
  "complaintSignals": ["buggy or unstable", "community or cheating problem"],
  "bugSignal": true,
  "featureRequestSignal": false,
  "pricingSignal": false,
  "commentSummary": "Player did not recommend the game; sentiment is negative; pain 82/100; praise 18/100. Topics: performance, bugs, multiplayer. Complaints: buggy or unstable, community or cheating problem.",
  "recommendedAction": "Prioritize as bug/performance evidence; cluster with similar reviews before roadmap decisions."
}

Input guide

Steam app IDs

Paste one or more app IDs. The app ID is the number in a Steam store URL.

Example:

https://store.steampowered.com/app/730/CounterStrike_2/

App ID: 730

Steam URLs

If you do not want to extract IDs manually, paste Steam app URLs. The actor extracts IDs automatically.

Comments per game

• 25 — fast test
• 100 — useful snapshot
• 300-500 — strong research set
• 1000+ — deeper VOC mining

Review feed

• Recent reviews — best for current sentiment.
• Updated reviews — useful for changed player opinions.
• All reviews — broader historical research.

Review sentiment

• All reviews — balanced research.
• Negative only — best for pain, bugs, pricing objections, and competitor weaknesses.
• Positive only — best for praise, positioning, and copywriting.

Language

English is recommended for best intelligence classification.

Use all if you want every available language, but keyword-based intelligence will be weaker outside English.

Notes and limitations

• Uses public Steam review endpoints.
• Does not log in.
• Does not scrape private user data.
• Does not bypass age gates, login walls, or account pages.
• Steam may not expose reviews for every app, unreleased title, demo, DLC-like item, or restricted page.
• Built-in intelligence is deterministic keyword scoring, not an LLM call, so it is fast and explainable.

Output destinations

Results are saved to the default Apify dataset.

A SUMMARY record is saved to the default key-value store with run settings, counts, and generated fields.