Review X-Ray

Review X-Ray turns the customer reviews already on the page you're looking at — Trustpilot, Amazon, the App Store, G2, Capterra, anywhere reviews live — into a real competitive-intelligence report. Not a summary. Not "AI found 17 themes." An honest read of what customers value, where a business is actually vulnerable, and what's worth building or testing next. FREE TO USE, NO SIGNUP Click the icon on any page with reviews and run it — no account, no API key, nothing to configure. Every install gets one free scan a day, up to 60 reviews. Want more than one scan a day, or to analyze larger review sets (up to 400)? Add your own free DeepSeek and OpenRouter API keys in Settings — entirely optional, never required to try the product. HOW IT WORKS Open any page listing customer reviews and click the Review X-Ray icon. It reads the reviews already rendered on that page — no copy-pasting, no spreadsheets — and runs them through a two-stage analysis: 1. Every individual review is classified — category, sentiment, explicit feature requests, implied unmet needs, and severity — by TypeSafe's Jev decision model. This is a structured, typed judgment per review, not a text summary, so every count in your report traces back to something Jev actually decided about an actual review. 2. Those pre-computed counts (never raw review text alone) are handed to an AI writer that turns them into a report. It never counts, estimates, or invents a number or a quote — every statistic you see was computed in code from Jev's real answers before the writer ever saw it. On the free daily scan, both steps run on our own server using our API keys. If you add your own keys instead, both steps run directly between your browser and DeepSeek/OpenRouter — see PRIVACY below for exactly what that changes. WHY THIS IS DIFFERENT FROM A GENERIC AI SUMMARIZER Most "AI review analysis" tools blur three very different things together: what a review literally says, what a pattern across several reviews might mean, and what you should actually do about it. Review X-Ray keeps those separate on purpose: - Evidence tiers, not vibes. Every finding is labeled Strong (5+ reviews), Emerging (2-4), or Signal (1) — so a single review's explicit request is still surfaced (it's real evidence), but it's never dressed up as a trend the way a 5-review pattern is. - A code-computed "Evidence Verdict." Separately from anything the AI writes, the report itself flags which findings are Strong Evidence and which ones you shouldn't overreact to — computed directly from the per-review counts, with zero AI judgment involved in that specific call. - Know / Suspect / Test reasoning. Reports separate what the data actually shows, from what that might mean, from what's worth researching next — instead of presenting a guess as a fact. - A single, evidence-gated hypothesis. Rather than forcing an "opportunity" out of every category, the report surfaces one commercial hypothesis only when the evidence actually clears a real bar — and explicitly lists what it does NOT prove. No strong hypothesis in the data? The report says so. That's a feature, not a gap. - A verdict built for one question: "If I were competing with this company, what would I do?" Findings are sorted into Match (capabilities you can't ignore), Exploit (real weaknesses), Investigate (patterns worth researching further), Build, Sell, and Position — and a bucket only appears when there's actual evidence behind it. WHAT YOU GET - Customer sentiment and a plain-language risk read (not just a percentage) - What customers value — and what a competitor would have to match to compete - Where the business struggles, honestly weighted by how many reviews back it up - Explicit feature requests vs. implied unmet needs, kept distinct - A pain map — frequency vs. severity, click any point to read the underlying reviews - The exact words and phrases reviewers use, tiered by how many people independently said them - Real quotes, picked for being specific and confidently classified, never invented - A direct export to a real, selectable, searchable PDF — same data as the on-screen report, so they never drift apart HOW IT SCANS Review X-Ray only reads what's already loaded on the page you're viewing when you click its icon — not other pages of the site, and not reviews hidden behind pagination or a "load more" button you haven't clicked yet. It automatically detects the block of repeated text on the page most likely to be real reviews and shows you an actual excerpt before running anything, so you can confirm it found the right thing. PRIVACY Free daily scan (default): review text is sent to our backend server to run the analysis and generate your report. We do not build a database of what you've scanned — review text is used only to produce your report and is not retained afterward. We keep a minimal usage record (a random device ID and a hashed IP, tied to a date, no review content) solely to enforce the one-scan-per-day limit. Bring your own keys (optional): add your own DeepSeek and OpenRouter API keys in Settings, and review text goes directly from your browser to those two providers using your own keys — your review text never passes through us. The first time you save a key, Chrome will ask you to approve access to that provider's domain specifically; this is requested only at that moment, never at install. Anonymous usage counts: to learn which features people use, the extension sends simple counts — for example "scan finished" or "PDF exported" — with the extension version and the kind of site (for example Amazon or Trustpilot). Each count gets a new random ID, so counts can't be linked to each other or to you. Page addresses, review text and reports are never included. The counts are processed by PostHog on EU servers, with IP addresses discarded. Either way, your API keys (if you add them) and your generated reports are stored only in your browser's local storage. Uninstalling the ex…