Every AI decision in ListForge is backed by an evidence bundle — a traceable record of what the AI found, what sources it used, and why it reached its conclusions. This is how you verify the AI's work and make informed overrides.
What's in an Evidence Bundle?
An evidence bundle is attached to each researched item and contains:
- Source data — the specific tools the AI used (image recognition, barcode lookup, marketplace search, etc.)
- Search queries — what the AI searched for when finding comparables
- Comparable sales — the actual listings it found and used for pricing
- Reasoning — why the AI chose a particular identification, category, or price range
- Confidence scores — how certain the AI is about each piece of its analysis
Viewing Evidence
On any researched item's detail page, you can expand the evidence section to see the full research trace. This shows:
- What the AI identified and why — which visual features or text led to the identification
- What comps it found — links to actual marketplace listings with prices
- How it calculated pricing — which comps influenced each strategy and by how much
When to Check Evidence
You don't need to review evidence for every item. Focus on:
- Items with low conviction — the AI isn't confident, so verify its reasoning
- High-value items — the stakes are higher, worth a manual check
- Items where the price seems wrong — check if the comps are actually comparable
- Unfamiliar categories — if you don't know the market well, evidence helps you learn
Using Evidence to Override
If you spot an issue in the evidence — say the AI used comps for a different model, or included a damaged item's sale price for your mint-condition item — you can:
- Dismiss bad comps — remove comparables that aren't actually comparable
- Edit the price — override the AI suggestion with your own
- Reject and retry — if the identification is fundamentally wrong, reject the item and recapture with better photos or hints
Evidence bundles persist forever. Even after you override a price or edit a listing, the original AI reasoning is preserved. This gives you an audit trail of how every decision was made.
The Philosophy
ListForge's approach is "AI recommends, you decide." Evidence bundles are the mechanism that makes this possible. Instead of a black box that says "price it at $45," you get a transparent process that says "here are 8 comparable sales averaging $47, here's why I adjusted for condition, and here's my recommendation." You can agree or disagree with full information.