Asking Analysis Questions
How to phrase questions to the Fake Products Library analyzer so you get a precise answer.
Analysis Questions (Analysis → New Question) let you ask the Fake Products Library — and, for some questions, every offer ever captured — a natural-language question. Each question re-runs on a schedule you set (default every 7 days), and you get a fresh markdown answer each time.
There's no fixed query syntax: type the question as you'd ask a colleague. But the engine only understands a specific set of things it can filter, count, and compare — knowing that set helps you phrase questions that get a precise answer instead of a vague one.
What you can ask about
| You can filter/search by | How to mention it |
|---|---|
| Product / description text | Just name the product, e.g. "Levi's jeans", "wireless earbuds" |
| Seller | Company name, nickname, or business name — exact or partial, e.g. "seller ACME Corp", "from user acme99" |
| Marketplace | Amazon, eBay, Fruugo, Allegro, BOL, ePrice, ManoMano, AliExpress, DIYTrade, etc. |
| Date range | "in the last 7 days", "since March 2026", "between January and June 2026" |
| Product category / tag | Whatever tags are used in your library, e.g. "counterfeit electronics", "apparel" |
| Takedown / legal status | "awaiting takedown", "resolved", "still active" |
| You can ask the engine to | Example phrasing |
|---|---|
| Count | "How many counterfeit products are in the library in total?" |
| Break down / rank | "What are the marketplaces with the most counterfeits?", "Who are the top 10 sellers by counterfeit listings?", "Break down counterfeits by product category." |
| Find sellers active on several marketplaces | "Which sellers operate across more than one marketplace?" |
| Search for a specific seller or offer, even before a verdict is made | "Has seller X shown up in the last 7 days?", "Find any offer from ACME Corp since March." |
| List matching entries | "List counterfeit products added in the last 30 days." |
- Most questions search the confirmed-fakes library — verified counterfeits only. This is the default, and it's what powers the counting/ranking breakdowns.
- Questions phrased as "has this seller/offer shown up," "did we capture anything from X," or similar pre-verdict language search every captured offer, confirmed fake or not — useful for spotting a seller before a human has reviewed their listings.
Good example questions
- "What are the marketplaces with the most counterfeits?"
- "How many counterfeit products are in the library in total?"
- "Which regions have the most fakes?"
- "Who are the top 10 sellers by counterfeit listings?"
- "Break down counterfeits by product category."
- "How many counterfeits are still awaiting takedown vs. resolved?"
- "Which sellers operate across more than one marketplace?"
- "List counterfeit products added in the last 30 days."
- "Has seller ACME Corp shown up in the last 7 days?"
- "Find any offer from user 'shoe_bargains99' since January 2026."
Current limitations
Worth knowing before you phrase a question, so you're not waiting on an answer the engine can't give yet:
- One date range at a time. You can filter to a period, but you can't yet ask the engine to compare "this period vs. a previous one" in a single question (e.g. "did this seller also appear before this current window, and where?"). Ask about the current window, then a separate question about the earlier window, and compare manually.
-
"Region" means marketplace domain, not country name. A region/marketplace breakdown
groups by domain (e.g.
amazon.de), so you'll see domain names in the answer rather than "Germany." - Blacklist matching is a check, not a search filter. If you name a seller, the answer tells you whether that seller is blacklisted — it can't yet search for "this seller OR anyone on the blacklist" as one combined query.
- Pre-verdict offer search returns at most the 200 most recent matches. For a very high-volume seller, older matching offers outside that window won't be reflected in the answer.
- Breakdowns are by tag/category, not literal product name. "Break down by product" returns categories/tags, not a count per exact product listing.
These are tracked as known gaps (PICD-61) and may be closed in a future update — this page will be kept current as that happens.