Key takeaways
- The Promptwatch community connector pulls visibility, citations, sentiment, share of voice, prompts, and crawler data straight into Looker Studio. No spreadsheets, no manual exports.
- One Looker Studio data source maps to exactly one Promptwatch project. Agencies should create one read-only project API key and one data source per client, then blend inside the report.
- Use a read-only Project Key, never a Master Key. Master keys span all projects and need an
X-Project-Idheader on every call, which the connector doesn't handle for you. - Rate limits are shared across every API key in your organization. A dashboard with 20 charts refreshing at once can trip per-second and per-minute limits even when you have hourly quota left.
- The connector covers reporting data only. Shopping, ads, and content surfaces stay in the Promptwatch app, so plan your dashboard around what the connector actually exposes.
Why build a custom GEO dashboard at all
Most GEO tools give you a dashboard. The problem is it's their dashboard, with their charts, their branding, and their idea of what matters. That works fine until you need to put AI visibility next to Google Search Console data in one client-facing report, or until a client asks for their numbers in their own reporting format.
Looker Studio is free, everyone on a marketing team already has a Google account, and reports share like any Google Doc. That combination makes it hard to beat for recurring client reporting.
There's also a measurement argument. AI citation behavior moves fast, and month-over-month shifts can be dramatic. Reddit's share of ChatGPT Search citations fell from 6.11% in May 2026 to 3.71% in June 2026, and later dropped from roughly 4% to 0.5% in a single day in August, per Promptwatch's Reddit citation research. A one-off audit snapshot goes stale in weeks. A connected dashboard that refreshes itself is the only format that keeps up.
Promptwatch is the platform this guide builds on, because its connector and API expose the underlying data rather than just a rendered report.
What you can (and can't) pull through the connector
Before you design anything, know what's on the table. The Looker Studio connector reads through the Promptwatch API and exposes, per project:
- Visibility: time series by day, week, or month, filterable by model, monitor, and prompt
- Sentiment: scores plus distribution over time
- Share of voice: time series with period-over-period comparison
- Responses and mentions: totals, plus top competitors by mention count
- Citations: cited domains over time, self-citation frequency, citation rank analysis, top cited pages, and per-model source breakdowns
- Competitor heatmap: the model-by-brand matrix
- Prompts and monitors: usable as report dimensions
- Crawler and visitor analytics: AI crawler trends, top crawled pages, visit totals
What's deliberately left out: shopping, ads, and content surfaces stay app-only. If your client's KPI is ChatGPT shopping recommendations or sponsored placements in AI answers, that lives in the Promptwatch app, not in Looker Studio. Design around this rather than fighting it.
Also worth knowing: the data comes from a short server-side cache, and Looker Studio adds its own freshness interval on top. Promptwatch's own docs describe it as