Running an audit

On-demand: Run audit now

On the Monitoring page, 'Run audit now' triggers a full audit for the brand selected in the top bar, across every AI engine on your plan. It asks every question on every engine, so it usually takes a few minutes, and longer if you are tracking many questions. You can leave the page and the results save automatically.

You need at least one active prompt for the brand. If you don't, the button tells you to add one first.

What a run costs depends on your plan, and the unit is one question, not one run. A run over eight tracked questions spends eight of your monthly allowance, not one. On the free plan that allowance is 3 a month, so a full run of a brand with several questions will use all of it at once; the run is clamped to whatever you have left rather than refused. Paid plans have a much larger monthly allowance and the same accounting applies to them, so a run can be refused once the month's allowance is spent. Your remaining allowance is shown on the Billing screen.

Automatic weekly monitoring

Beyond on-demand runs, Pondral automatically re-audits your brands on a weekly schedule so you can see trends over time without doing anything. Recent runs appear on the Monitoring page with timestamps and status. History shows the full trend.

Because AI engines are non-deterministic, a single run is directional and a trend across several runs is the real signal. The weekly full audit is designed to smooth that out.

Which engines run

The number of AI engines an audit uses depends on your plan. Paid plans run all five (ChatGPT, Claude, Gemini, Perplexity, Grok). The free plan runs two per check. Each engine has a different knowledge base and retrieval method, so your score can differ a lot between them. That's expected, and the Engines page shows the per-engine detail.