Reviews Scraper & AI Insights — Trustpilot, G2, Capterra, App Store & Google Play

All public reviews for a company or app across five sources in one schema, with ratings, dates and company replies. Optional AI insights: themes, pain points, praise, feature requests and competitor mentions, in English or Arabic. Alerts for new and negative reviews.

جمع تقييمات ومراجعات أي شركة أو تطبيق من آب ستور وجوجل بلاي وترست بايلوت في مخطط واحد موحّد، مع تحليل ذكي للمواضيع ونقاط الألم وطلبات الميزات — بالعربية الفصحى أو الإنجليزية، وتنبيهات فورية للمراجعات السلبية الجديدة.

Who it's for

• Product teams turning hundreds of reviews into a themed backlog instead of reading them all.
• CX and support leads who need to know within a day when a bad review lands.
• Agencies and competitive-intel analysts tracking several companies at once.
• Investors checking whether a portfolio company's ratings are trending up or down.
• App publishers watching Gulf storefronts (sa, ae, eg) alongside us and gb.

What makes it different

• Five sources, one schema. A Trustpilot review and an App Store review come back with the same fields and the same 1–5 rating scale.
• Company replies included, so you can see which complaints were answered.
• Arabic and English insights. Analysis is written in Modern Standard Arabic when you ask for it, while quotes stay in the language they were written in.
• Alerts, not dumps. Change mode returns only new reviews, negative reviews, rating drops and newly added replies.
• Anonymized at the source. No reviewer names, handles, avatars or profile links ever enter the dataset — they are stripped in the source adapters, before a record is built.

Output

Four item types, told apart by itemType, all documented in .actor/dataset_schema.json: review, summary (per target × source), insights (per target) and alert (change mode). The Console shows Reviews, Insights and Alerts views.

A review, from a real run:

{
    "itemType": "review",
    "source": "googleplay",
    "target": "Notion",
    "productName": "Notion: Notes, Tasks, AI",
    "reviewId": "5f56eb4a-baf7-4ff8-b2ae-e33c15873454",
    "rating": 1,
    "title": null,
    "text": "when I try to open the settings page it will stuck",
    "language": "en",
    "publishedAt": "2026-09-06T16:07:11.184Z",
    "country": "us",
    "appVersion": "0.6.4114",
    "helpfulCount": 2,
    "companyReply": null,
    "scrapedAt": "2026-09-07T06:00:00.000Z"
}

An insights item (shape; requires an LLM key — see Limitations):

{
    "itemType": "insights",
    "target": "Notion",
    "overallSentiment": -0.15,
    "themes": [
        {
            "theme": "Sync reliability",
            "sentiment": -0.6,
            "count": 14,
            "shortQuotes": ["sync stops working after a few hours"]
        }
    ],
    "painPoints": ["Offline mode loses edits"],
    "praise": ["Flexible database views"],
    "featureRequests": ["Native Arabic RTL support"],
    "competitorMentions": [{ "competitor": "Obsidian", "count": 4, "context": "cited as faster for local notes" }],
    "suggestedActions": ["Investigate sync failures on Android 14"],
    "sourceBreakdown": { "appstore": 40, "googleplay": 40 },
    "reviewsAnalyzed": 80,
    "language": "en"
}

Alerts and scheduling

1. Set emitChangesOnly and a trackingId, then Save as task.
2. Add a Schedule — daily is typical, hourly if you are in a launch week.
3. Under Integrations, send results to Slack, Email, a Webhook, Make or Zapier on Run succeeded. In change mode the dataset is empty when nothing happened, so you only hear from it when something did.
4. Route by type: send negative_review and rating_drop to a human, log new_review.

The first run for a tracking ID records a baseline silently, so your first alert is a real one.

n8n and Make

Use the Apify node with Run Actor and get dataset, then filter on type == "negative_review" to open a ticket, or on rating_drop to notify a lead.

Use as an API / MCP tool

With Standby enabled:

GET https://<your-actor>.apify.actor/?target=Notion&sources=appstore,googleplay&aiInsights=1

Parameters mirror the input (target, website, sources, countries, lastNDays, maxReviewsPerSource, aiInsights, insightsLanguage, trackingId). GET /health returns {"ok":true}.

Pricing

Pay-per-event. You pay for sources read, reviews returned and analysis produced.

Event	Price	When
Source checked	$0.005	Per target × source fetched successfully. A blocked source is free.
Review	$0.002	Per normalized review stored.
AI insight batch	$0.03	Per batch of 10 reviews analyzed; the rollup is included.
Alert	$0.01	Change mode, per alert. Source errors are free.

• 500 reviews across 3 sources for one company: 3 × $0.005 + 500 × $0.002 = $1.015.
• AI insights over those 500 reviews (50 batches): $1.50.
• Daily alerts for 10 competitors: about $0.15 a day, plus $0.01 per new review.

Default memory 512 MB is enough; the Actor makes no browser calls.

Per-source notes and limitations

• Apple App Store — Apple's public customer-reviews feed, per storefront. Reliable and fast. The feed carries no developer replies, so companyReply is null for App Store reviews.
• Google Play — through the maintained google-play-scraper package, including replies. A Play breakage is usually fixed by bumping that dependency.
• Trustpilot — read from the page's embedded JSON rather than CSS, so it survives redesigns. It is fronted by bot protection that refuses most datacenter traffic; if you see a source error, switch proxyConfiguration to residential proxies.
• G2 and Capterra — off by default. Both block automated traffic aggressively and publish no feed. Selecting them will normally return a source error rather than reviews. They are listed so the input is honest about what exists, not because they work reliably today.
• AI insights need a key. Without OPENROUTER_API_KEY or OPENAI_API_KEY set on the Actor, aiInsights is skipped with a warning and nothing is charged for it.
• Reviews with no date are kept rather than dropped, so a source that omits dates is not silently filtered away by lastNDays.

Related actors

• ../hiring-signals-monitor — track companies' career pages.
• ../ad-library-intelligence — track competitors' live ads.
• ../page-change-monitor — track any page for changes.

Legal note

The Actor reads publicly published reviews — the same content any visitor sees — and never logs in or sends cookies. It collects no reviewer identity: names, handles, avatars and profile links are stripped in the source adapters before a record is built, and contact details pasted into review text are redacted. Quotes inside AI insights are short excerpts of at most 15 words, never whole reviews. Review text remains the work of its authors and the property of the platform that hosts it; use the output for analysis in line with each source's terms and the laws that apply to you. Trustpilot, G2, Capterra, Apple and Google are trademarks of their owners; this Actor is not affiliated with them.