Methodology & sources

AI Hardware Deals tracks trusted new and used AI PC parts by region (EU, DE, US) with price trends and anti-scam filters. A BranoLabs project.

Idealo.de (primary EU retail)

We treat Idealo as the primary EU price index. Our daily script politely requests public OffersOfProduct / Liste URLs (User-Agent identifies BranoLabs; multi-second delay; on-disk cache). Paths we use are not broadly Disallow'd for User-agent: * in Idealo robots.txt, but live fetches from our servers currently return HTTP 403. When blocked, we keep last-known prices, set a stale badge, and never invent numbers. Manual/seed updates fill gaps until Idealo is reachable or we add partner feeds.

Source list

Own daily source list for AI Hardware Deals. Idealo.de is the primary EU retail index. As of 2026-10-15 automated Idealo fetches return HTTP 403 from our egress (Geizhals homepage 200 but search pages 403); refresh script keeps last-known prices and marks them stale. Never invent prices.

Source	Region	Type	Trust	Status
Idealo.de	DE	new / price-index	85	blocked_bot (403)
Geizhals.de	EU	new / price-index	85	reachable (200)
Mindfactory	DE	new / retailer	80	reachable (200)
Alternate.de	DE	new / retailer	80	reachable (200)
Caseking	DE	new / retailer	80	manual (200)
eBay.de	DE	used / marketplace	60	blocked_bot (403)
Kleinanzeigen.de	DE	used / classified	45	manual_strict (200)
eBay.com	US	used / marketplace	60	blocked_bot (403)
Newegg	US	new / retailer	75	blocked_bot (403)
Memory.NET	US	new / retailer	70	blocked_bot (403)

Freshness badges

fresh — refreshed from the named source on lastChecked
stale — last-known kept after a failed/blocked fetch
seed — curated seed for trends/trust UX (labelled)

Catalog generatedAt: 2026-10-15T16:12:50.021Z. Mode: mixed.

Anti-scam rules

Reject below 55% of latest median for SKU + condition (+ region when available).
Suspicious below 70% of median.
Require HTTPS listing URLs.
Prefer allowlisted marketplaces/retailers/indexes; unknown domains lose trust.
Used/classified: seller age under 90 days or feedback under 95% lowers score; scam keywords in titles lower/reject.

Daily automation

npm run update-sources pings each source. npm run update-prices retries Idealo-linked offers, applies anti-scam, appends history when a parse succeeds. Both are polite and fail soft.