The AI Assistant for load test reports is in Beta. This article describes how it handles your data.

Data Processing

When you start an analysis, the AI Assistant receives a summary of the selected load test and send to an AI provider. Review what data the AI Assistant sends, who processes it, and how long it is kept before you run an analysis.

The first analysis of a test can take longer while the AI Assistant processes the test data. The wait time depends on the test size.

Item Details
Test configuration sent for analysis Target URLs, EveryStep script or JMeter configuration, load curve, load injector settings, geographic distribution.
Test results sent for analysis Session counts, success and failure rates, response times per target, load injector CPU usage.
Additional data for follow-up questions Request-level results: URLs, timings, status codes, error messages.
Additional data for single-session analysis Request and response headers. Part of the response body.
AI provider OpenAI API.
Model training Data is not used to train AI models.
Encryption HTTPS (TLS) between your browser, LoadView, and the AI provider.
Provider retention OpenAI may keep API data for up to 30 days for abuse monitoring only.
Retention in LoadView Analysis conversations and related test data stay in your account.
Deletion No automatic deletion. To delete the AI analysis of a specific test, open a support ticket.

This removes the data from LoadView only. It does not affect data held by the AI provider.

What the AI Assistant Analyzes

Metrics

The AI Assistant analyzes the following load test metrics and results:

  • Session totals and failure rate
  • Average, minimum, maximum, and 90th-percentile response times
  • Step timings and variation
  • Bottleneck step and its share of total transaction time
  • Outliers
  • Response-time trends during the test
  • Failures by step, with error details and timing
  • Third-party errors versus application errors
  • Load injector CPU usage

The AI Assistant flags:

  • A high failure rate
  • Overloaded load injectors
  • Common test configuration issues

Each conclusion includes the supporting numbers, such as counts, timings, percentages, timestamps, and per-step tables. Use them to verify the finding against your report.

Sensitive Data

Load test reports can contain internal URLs, IP addresses, headers, cookies, session identifiers, authentication tokens, error details, and values entered in scripts. In the Beta, the AI Assistant receives the full test data. You cannot select or review fields before submission.

The AI Assistant saves each analysis in the chat log as a named conversation. Each conversation is linked to its load test and shows the date and time it was created.

To reduce risk:

  • Store credentials in Secure Vault.
  • Use dedicated test accounts and test tokens. Do not use production secrets.
  • Do not run AI analysis on tests whose data must not go to an external AI service.

Access Control and Audit

Account Admins can enable or disable the AI Assistant in Account Preferences. The setting is available at the parent account level and at the Department level. A setting on the parent account does not apply to its departments. Enable the AI Assistant separately in each department that needs it.

More granular, per-user permissions are under consideration for a future release.

Changes to the AI Assistant setting are recorded in the account Audit log.

Limitations

AI-generated analysis should be reviewed against the load test report before you act on it.

  • Baseline-versus-new comparison is not available. Each conversation with AI Assistant covers one load test.
  • AI analysis results are not available for export and appear in the AI Assistant panel only.
  • For very large tests, the AI Assistant analyzes aggregated data, not every request.
  • Explanations of root causes, such as server load or caching effects, are possible causes rather than confirmed diagnoses. Use these explanations as guidance for further investigation.

Usage Limits and Cost

Check your monthly token allowance, the actions that use it, and the current Beta pricing.

Item Details
Monthly allowance 15 million AI tokens per account (10 million input, 5 million output).
Reopening a saved analysis Uses no tokens.
New conversations and follow-up questions Use tokens from the allowance.
Pricing Included at no extra cost during the Beta. LoadView will notify customers in advance of any pricing change.

Support

If an analysis looks inaccurate or unclear, contact Support with the test ID. Support reviews the analysis against your test results.