Understanding Confidence Scores

Updated February 14, 2026

What confidence percentages mean and when to trust them.

Understanding Confidence Scores

Each AI suggestion includes a confidence score (0–100%) that indicates how certain the model is about the recommendation.

What Confidence Means

Confidence is derived from:

Data volume – listings with more views and sales history get higher-confidence suggestions.

Market signal strength – categories with many similar listings provide stronger benchmarks.

Model certainty – the underlying language model's internal token probability for the generated suggestion.

Confidence Tiers

Confidence	Meaning
80–100%	High – the suggestion is strongly supported by data. Apply it.
60–79%	Medium – the suggestion is likely beneficial. Review before applying.
Below 60%	Low – limited data. Use your own judgement.

When Confidence is Low

Low confidence is not a signal that the suggestion is wrong — just that we have less evidence. New listings (fewer than 50 views) will often have low-confidence scores. As your listing accumulates performance data, re-running the AI analysis will produce higher-confidence results.

Feedback Loop

Use the thumbs up / thumbs down on each suggestion to help train the model. Your feedback is stored locally and used to bias future suggestions for your shop.

Was this article helpful?