White-label GEO reports for agencies

Clients have started asking their agencies a new question: "what does ChatGPT say about us?" A GEO report — who the AI engines name, who they cite, how that moved this month — is the deliverable that answers it. The agency problem is producing that report under your own brand, for twenty clients, without twenty dashboard subscriptions. The data-layer answer: fetch each client's raw records through one API, run one report template over them, and put your logo on the output. AgentGEO has no end-client surface at all — there is nothing to white-label because there was never a vendor UI in the deliverable to begin with.

The unit of work is one client's month: a fixed query set, fetched on schedule, reduced to the four numbers clients actually read — mention rate, share of voice, citation share, and what changed — with the answer excerpts as receipts.

How the agency workflow runs

1. One workspace, per-client query sets. Keep each client's queries in a file next to their brand and competitor list. Twenty clients is twenty files and one loop — not twenty logins.
2. Fetch on the reporting cadence. A monthly run of 30 queries × 3 engines is 90 records per client — 90 credits, a knowable line on your COGS that prices the retainer. Failed fetches cost zero.
3. Render your template. Compute the numbers from the stored records and lay them out in your house style — a PDF, a Notion page, a slide. The geo-report skill produces the full write-up if an agent is doing the drafting.

while read -r query; do
  curl -s https://api.agentgeo.org/v1/fetches \
    -H "Authorization: Bearer $AGENTGEO_KEY" \
    -H "Content-Type: application/json" \
    -d "{\"query\": \"$query\", \"surfaces\": [\"chatgpt\", \"perplexity\", \"gemini\"]}" \
    --max-time 200 \
    | jq -c --arg q "$query" '.answers[] | {q: $q, engine: .surfaceKey, answer: .answerText, sources}'
done < clients/acme/queries.txt >> clients/acme/2026-07.jsonl

Everything downstream reads that JSONL: this month's numbers, the diff against last month's file, and the quoted answer lines that make the report credible. When a client asks where a number came from, the answer is a line in a file you keep — not "the tool said so".

Let your agent draft it

Over MCP, the geo-report skill — one of eight shipped in agentgeo-mcp — takes a brand, fetches the set, computes the metrics and drafts the narrative in one run. An account manager's edit pass turns it into the deliverable; the agent did the fetching and the arithmetic.

claude mcp add agentgeo -- npx -y agentgeo-mcp --key ag_live_...

Alternatives

Some GEO dashboards sell agency tiers with white-label exports — workable if their template is your template and per-seat-per-client math clears your margin. The structural difference: their export is the ceiling of what you can deliver, while raw records make the report format yours to evolve. Client wants a Looker page instead of a PDF? Same records, different renderer. The comparison pages lay out where dashboards genuinely win.

More use cases

Agencies typically pair the report with ongoing share-of-voice monitoring between cycles, or go further and embed AI visibility into their own client portal — see all use cases.