GearScout is a comparison tool, not a marketplace — every listing links back to where it's actually sold. Here's exactly how the number on each card gets calculated.
For every category and condition (e.g. "power racks, used — good"), we take the median asking price across every other active listing in that group. That's the baseline a listing's price is compared against — not a guess, and not the manufacturer's RRP, but what comparable secondhand items are genuinely listed for right now.
Claude (Anthropic's AI) reads each listing's title and description and flags things a buyer should verify before committing — vague condition language, missing details a genuine seller would usually include, or inconsistencies between the stated condition and the description. It's given the fair-value number above as context, so its reasoning is grounded in a real figure rather than inventing one.
The 0–100 score is calculated deterministically from the two numbers above — how far below or above fair value a listing is priced, minus a penalty for each condition-risk flag. The AI informs the score by identifying risks; it doesn't invent the score itself. That keeps it consistent and explainable across every listing, not just plausible-sounding.
Currently eBay, via their official Browse API — never scraped. Outbound links go through the eBay Partner Network, so GearScout may earn a commission on qualifying purchases at no extra cost to you. That's the only way this stays free to use.