Understand the data behind the report.

Start with a client project

A project holds one brand, its website, aliases, competitors and tracked questions. Country and prompt language are recorded with each collection. Keep questions specific enough to reflect a real buying decision. Prompt language is your label; it does not control the answer language.

Collect and inspect answers

Monitoring covers ChatGPT and Google AI Mode. Each scheduled prompt on one platform in one location consumes one answer from the shared allowance, including failed attempts. Confirm the count before starting. Daily monitoring requires explicit opt-in; collections run at 03:15 UTC, with at most one collection per project per UTC day.

Each result preserves the question, answer and source evidence. Explicit citations, search results and attached links are separate. Incomplete extraction remains visible.

Read metrics with their coverage

• Collection coverage: successful answers divided by scheduled answers.
• Mention rate: analyzed answers that mention a brand, divided by analyzed answers. Repeated mentions in one answer count once.
• Recommendation rate: analyzed answers that recommend a brand, divided by analyzed answers.
• Share of voice: a brand’s answer-level mentions divided by all tracked-brand mentions in analyzed answers.
• Citation rate: explicit citations to a brand’s website across successful answers. Incomplete citation data can leave the rate unavailable.

Failed collection and missing analysis are not zero visibility. Compare only matching questions, platforms, locations and competitor sets. AI answers vary; these measurements describe collected answers, not every answer a user might see.

Share a stable report

Publish a dated snapshot with your summary, logo and accent color. Readers need access to that client project. PDF, CSV and Markdown exports preserve the report; sending a report link does not grant access. Evidence and reports are retained for 12 months.

Choose weekly or monthly email delivery in UTC. Only the latest published report is sent, once per schedule. Drafts stay private. Turn delivery off whenever you need.

Use an assistant with permission

Agent access is read-only and limited to explicitly approved projects and published reports. It cannot edit prompts, start paid collections or change billing. See the ChatGPT integration plan and tool reference.