1. Build the sourcing rules
In the Vinted or eBay workspace you create searches that describe what you are prepared to buy. Vinted has category-aware controls such as footwear/clothing/general product, title wording, alternative phrases, exclusions, sizes, colours, prices and conditions. eBay uses normal search terms plus required/excluded title phrases, price and postage limits, listing type, condition and UK-only filtering.
2. Start monitoring
When monitoring is started, each marketplace repeatedly processes its own enabled searches. Newly seen matches are recorded and the most relevant opportunities can be queued for deeper AI/market work. Vinted and eBay scanning remain independent.
3. Send the first signal
With normal Discord delivery, an explicit saved-search match can be sent quickly rather than waiting for a slower AI analysis. This means you can open the original listing while AutoBot continues the deeper work in the background.
Both Vinted and eBay also have an Only send AI analysis switch. When enabled, the match still appears in AutoBot immediately, but Discord waits for completed AI analysis. If AI fails or times out, that AI-only message is not sent.
4. Inspect the listing with local AI
The Vinted and eBay AI workers use your local Ollama installation. They analyse the actual listing evidence rather than assuming every product needs the same checks. Depending on the listing and settings, the AI can consider title, description, specifications and listing photos.
- Condition and visible wear.
- Damage risk and completeness.
- Identity/scope clues that affect what the item actually is.
- Evidence confidence and issues that should be checked manually.
Fast, Balanced and Deep modes change how much visual evidence is reviewed. Automatic hardware-adaptive mode is the recommended setting for most customers.
5. Compare current-market evidence
Market valuation runs alongside AI work where possible. Comparable listings are filtered for relevance so a wrong product, wrong scope or obvious parts-only listing cannot be treated like the full product. When there is not enough reliable market evidence, AutoBot can withhold a financial recommendation rather than fabricate a profit figure.
6. Review the deal
Deals pages combine the information you need to decide what deserves attention: buy cost, listing condition, AI notes, market estimate, potential profit/ROI, evidence confidence and the original listing link. This is the point where you make the decision.
7. For eBay auctions, keep refreshing the live situation
Tracked auctions keep live countdown/bid information and refresh more frequently as the auction approaches its end. Market evidence can be refreshed to support safe-bid guidance based on your target profit and ROI. AutoBot never places the bid.
8. Track the resale after purchase
Use Pipeline and Watchlist while deciding or preparing. If you buy an item, record the real purchase and costs in Resales. After sale, enter the actual selling price and date. True Profit then shows realised profit/ROI and AI Learning can compare the original forecast with the finished result.