FAQ

Questions, answered plainly.

What counts as a "result"?+

A result is a scored agent run backed by its transcript: whether the agent selected you over a competitor, and if selected, whether it completed the task or failed. We report rates across many runs, and every rate links to the runs behind it.

Which agents do you test?+

By default Claude Code, Cursor, and Codex, across representative tasks. Enterprise engagements can add more agents and specific model versions.

How is this different from SEO or AEO?+

SEO optimizes for search rankings; AEO optimizes for answer-engine citations. Both target retrieval. We target the decision an agent makes while doing real work — which SDK it installs, which tool it calls, whether it succeeds — and the product surfaces that drive it.

Do you need access to our code?+

No. We work from your public product surfaces — docs, quickstarts, package, API, and MCP server. If you want us to fix private surfaces, we scope that separately.

How long does an audit take?+

Most first audits are delivered within one to two weeks, depending on how many workflows and competitors you want covered.

What do we get at the end?+

A report with your selection rate versus named competitors, a failure breakdown by cause, and a prioritized fix list — each finding linked to the transcript that produced it.