I run an agency and need to show clients how they rank in AI search results: how we report it
How we show clients where they stand in AI search results: fixed prompt panels, share of voice, citations, honest caveats and a reporting format that holds up.
We are an agency, and we have had this exact problem. A client asks, "where do we rank in ChatGPT?" and expects a number like the one in a rank tracker. There is no such number, and pretending otherwise would put us on the wrong side of the client's trust within a quarter.
What we do instead is build a report that answers the real question, which is: how visible are we, against whom, and what can be done about it?
Reframe "rank" in the first conversation
AI engines write answers. They do not list ten links. So we tell clients up front that we will report:
- How often the brand is mentioned across a fixed set of prompts.
- Share of voice against named competitors.
- Position within the answer.
- How the brand is described.
- Which sources the engines cited.
- AI-referred traffic and conversions, where they can be measured.
And we tell them that answers vary between runs, so we report trends, not single checks.
The panel is the foundation
For each client we write a prompt panel together: category questions, comparison questions and direct brand questions, grouped by topic and persona and, for local clients, by location. Clients are good at spotting prompts we missed, and it gets them invested in the report.
We run it daily. Promptwatch is the platform our client programs run on, and for agency work the agency plans matter: unlimited projects, unlimited prompts, unlimited pitch projects and 10 seats on every agency plan. Kick-off is $199/mo, Growth $399/mo and Scale $799/mo, and the tiers differ by response volume, crawler log volume and visitor events.
What goes in the monthly report
We keep it to three parts.
1. Trend. Visibility Score, Share of Voice and Position in Answer over time, per engine, with competitors. Promptwatch's custom dashboards have templates for competitor comparison, topic gaps, AI model breakdown and citations, and the heatmap compares any two dimensions.
2. What changed and why. Citation analytics show which pages and domains feed the answers. Agent Analytics shows IP-verified crawler logs, and visitor analytics shows AI-referred sessions and conversions. This part answers the client's next question, which is always "why."
3. What we do next. The Action Board gives a prioritized list, and we pick the work.
Reports go out as PDFs. The platform exports PDFs of any dashboard and can schedule them, with Excel and CSV exports for anyone who wants the raw data. We do not rebuild charts by hand. A Looker Studio connector is on every agency plan too.
Using the tool to win clients
Pitch projects are unlimited on agency plans. Before a prospect signs, we run their brand and two competitors through a panel, then show them where they stand. It is the best sales material we have, because it is about their business.
White-label
Promptwatch lists a white-label workspace and client portal for agencies, and white-label branding on PDF exports. We would not recommend buying on that basis without confirming exactly which plan includes what you need. Slack briefings are not available on white-label workspaces.
What agency plans do not include
They exclude AEO articles, the knowledge base, 360 Insights, Shopping Insights and the automated Content Agent. If a client needs the Content Agent, that sits on Business or Enterprise plans for a brand account. Plan your service scope around that.
How we price what we deliver
We do not price on rank. We price on the reporting, the analysis and the work we ship. Clients who ask for a guaranteed position get a polite explanation of why nobody can give one.
Where we have seen the pain
Other tools in this space can be fine for single brands, but agencies hit friction on reporting and on how credits or response limits scale across many clients. Ask any vendor how they count usage across projects before you commit, and ask to see a real client report.
Start small
Choose three clients, build a panel for each, run it for four weeks, and build one report format you can reuse. The free Explore plan lets you test the method on one brand, with 10 prompts on ChatGPT only, before you commit to an agency plan.