How a Nestor property report is produced

1. Match the exact property

Nestor starts with the listing URL, address, or screenshot supplied by the user. It corroborates identifying details across retrieved pages. If more than one property is plausible or confidence is below the product threshold, the analysis returns a not-found result instead of silently choosing a nearby home.

2. Retrieve decision-useful evidence

The report prefers the original listing and official primary sources. It records links actually retrieved during the analysis. Missing evidence stays missing: an estimate is not presented as an official fact and a neutral score is not inserted merely to fill a gap.

See the source hierarchy and limitations.

3. Score six disclosed dimensions

The current report evaluates value (25%), schools (15%), commute (15%), safety (15%), lifestyle (10%), and condition (20%). The overall score is calculated in application code as the rounded weighted mean; it is not independently chosen by the AI.

Each dimension includes its score rationale, evidence confidence, limitations, and links. Scores are comparative judgements, not facts, valuations, guarantees, or predictions.

4. Validate the response

The application checks the report schema, metric order and weights, overall-score arithmetic, source URLs, exact-listing provenance, and required sections before showing the result. Invalid outputs fail rather than being displayed as complete reports.

5. Explain AI use and human responsibility

AI is used to search, synthesise, compare, and write the report. Automated checks reduce common failure modes but cannot prove that every publisher record is complete or current. Users must verify important claims and seek qualified advice before making a purchase decision.