Skip to main content
Neural Defend

Reading NeuroVerify API Results

Read the response status before any score or risk band. Success means the media was scored. Rejected means the file did not pass a validation or face-policy condition and no authenticity score is available. Error means analysis did not complete. Image results use one risk score and band; video results keep video and audio fields separate. Low risk is not proof of authenticity, medium is inconclusive, and high risk should trigger review rather than an automatic rejection by itself.

Maintained byNeural Defend
Updated August 7, 2026
Siva Shankar, Co-founder and CTO of Neural Defend.
Siva ShankarCo-founder & CTO
An article-specific diagram for Reading NeuroVerify API Results organizes key sections including branch on status first, use risk as evidence, record what the decision needs.

Branch on status first

The wire contract defines success, rejected, and error. Some official SDKs also preserve an unknown forward-compatible outcome when a newer server response cannot be safely mapped. Code an explicit branch for every state and deny silent fallthrough to approval.

The billable field is independent of whether a score exists. In documented face-policy cases, the service can return a billable rejection on HTTP 200 with null risk fields.

Use risk as evidence

For images, risk_score is an opaque 0.1 through 10.0 value and risk_level is low, medium, or high. The contract advises using the risk band for business rules because thresholds may be tuned.

For videos, apply the same status-first rule, then inspect video and audio fields independently. A silent video can have a valid video result and no audio score. Preserve that distinction in storage, review screens, and downstream events.

An article-specific topic map for Reading NeuroVerify API Results organizes key sections including use risk as evidence, record what the decision needs, branch on status first.

Record what the decision needs

Retain the transaction identifier, outcome status, billable indicator, relevant risk bands, and the policy decision made by your system. Keep the original modality values when a client-side policy derives an overall video decision.

Human-readable messages can support user recovery and review, but raw diagnostic envelopes, filenames, local paths, and headers may contain sensitive operational data. Follow the selected SDK's documented serialization surface and sanitize support artifacts.

Evidence boundary

Limitations

  • No public result field proves authenticity or replaces a broader decision process, human review, or other evidence required by the use case.

Source record

Sources and references

  1. Neural DefendNeuroVerify response schemasVerified August 7, 2026.
  2. Neural DefendTypeScript SDK working with resultsVerified August 7, 2026.

Plan recovery

Separate errors from business rejections

Apply the documented retry policy without duplicating transactions or retrying unchanged invalid media.

Review errors and retries