Provenance|AI Memory Audit Infrastructure
v1 · ChatGPT Memory
Memory Integrity Scanner

Audit your ChatGPT memory.
Detect what's contaminated before it contaminates your outputs.

Import your ChatGPT memory export. Provenance runs multi-layer detection across temporal staleness, overconfidence, unknown origin, decision risk, and downstream scaffolding effects — then lets you quarantine bad memories and export a clean seed file.

How to export your ChatGPT memories
  1. 01Open ChatGPT → Settings → Data Controls
  2. 02Click “Export data” and confirm
  3. 03Download the ZIP file from your email
  4. 04Locate memory.json inside the archive and import it here
Detection Layers
Temporal Staleness

Detects year references, version numbers, prices, and relative time markers that suggest a claim may have drifted.

Overconfidence

Flags absolute language — always, never, definitely, 100% — that records opinions as facts.

Source Gap

Identifies short declarative claims with no identifiable source, origin, or qualifying context.

Decision Risk

Surfaces claims that may directly inform tool choices, hiring, strategy, or financial decisions.

Classification

Checks preference, aversion, and role-attribution claims that commonly reverse over time.

Scaffolding Effect

Detects first-person identity, ownership, and intent claims that orient all downstream AI outputs.