AI visibility tracking for agencies: pay per client check instead of per-seat plans

Agencies have two ways to track clients' brands in AI answers in 2026. Agency plans on dashboards such as OtterlyAI, Peec AI, Rankscale and the Semrush AI Visibility Toolkit give you client workspaces, white-label reports and daily tracking, from about $99 to $385 a month plus per-client add-ons. A pay-per-call API such as Cite42 charges per check instead: a client's 10 buyer questions, rerun weekly across ChatGPT, Perplexity and Gemini, cost $5.63 a month, with no plan, seats or workspace fees.

The first route suits an agency selling AI visibility as a packaged service with client logins. The second suits an agency that already builds its own reports and wants the data inside Claude, ChatGPT or a spreadsheet. This guide covers what each costs and how the pay-per-check setup works client by client.

Cite42 publishes this guide and is the pay-per-call option in it. The agency plan details come from each vendor's public pricing and agency pages, read on October 16, 2026 for OtterlyAI, Rankscale and Peec AI and on October 6, 2026 for Semrush and Scrunch. We did not buy the plans.

What agency plans cost

Two costs hide in the plan prices. Prompts are pooled, so a fifth client eats into the fourth client's allowance until you buy more. And engines are often add-ons: OtterlyAI charges $109 a month for Claude on Standard, and Peec AI plans track three models of your choice.

The pay-per-check setup, client by client

Cite42 has no client workspaces. What it has is a tracker: a named list of prompts, the brands to look for, the engines, and a weekly or monthly schedule. Create one tracker per client.

1. Write 10 to 20 buyer questions for the client, the questions a prospect asks an AI assistant before shortlisting: "Best payroll software for restaurants", "Gusto alternatives", "cheapest payroll with tip reporting".
2. Run them once as a baseline. From Claude Code, Codex, Cursor or ChatGPT with the Cite42 MCP server connected, ask: "Check these 10 questions for Acme against Gusto and Rippling across ChatGPT, Perplexity and Gemini." Ten questions cost $1.30, or call POST /v1/rankings from a script.
3. Save them as the client's tracker, named after the client, and pick weekly. Cite42 shows the maximum cost per run before the tracker starts, so you know the per-client cost on day one.
4. Pull the report when you write the client update. GET /v1/trackers/:id/report returns the mention rate and rank per prompt and per engine over time. The same report is one sentence away in Claude or ChatGPT: "Summarize the Acme tracker since last month." Add an AI Citations tracker for the same prompts if the client also wants to know which pages the answers cite.

What a client costs per month

Ten clients with 20 prompts each is 200 prompts a week, $112.67 a month in total across three AI models, billed from one prepaid balance that never expires.

There are no seats. Anyone on your team with the MCP server connected uses the same account, and you can create a separate API key per client or per person to keep scripts apart. The usage page breaks spend down by tool rather than by client, so take the per-client figure from each tracker's cost per run.

Pitching with data before the contract

The cheapest use of pay per check is the pitch. Before a first call, run the prospect's 10 questions for $1.30 and bring the table: how often they are named, who is named instead, and which pages those answers cite. OtterlyAI's pitch workspaces do the same inside its agency plan. With Cite42 the pitch costs the same $0.13 per question as the retainer work, and there is nothing to set up or tear down afterwards.

What you give up without a dashboard

• Clients do not log in to Cite42, and there is no white-label dashboard. You deliver the report in your own format: a doc written in Claude or ChatGPT, a Google Sheet filled from the API, or your existing reporting tool.
• There is no prompt discovery database. Ahrefs Brand Radar and Semrush suggest prompts from their indexes. With Cite42 you write the questions, which is also why they match the client's buyers.
• Scheduled trackers rerun weekly or monthly, not daily. On-demand calls run any time.
• You build the automation. The MCP server makes most of it a sentence in Claude or ChatGPT, but no dashboard does it for you.

Frequently asked questions

What is the best AI visibility tool for agencies?

For a packaged service with client logins and white-label reports, OtterlyAI's Agency Partners program, Peec AI's agency bundles and Rankscale's Growth plan are built for it, from $189 to $385 a month plus add-ons (as of October 2026). For an agency that builds its own reports, Cite42 charges per check with no plan: $0.13 per prompt across ChatGPT, Perplexity and Gemini.

How much does AI visibility tracking cost per client?

On Cite42, a client with 10 buyer questions checked weekly across three AI models costs $5.63 a month, and 20 questions cost $11.27. On a dashboard, a client is a share of the plan's prompt pool: Semrush charges $99 a month per extra domain, and OtterlyAI Standard spreads 150 prompts across all client workspaces.

Can I track several clients in one Cite42 account?

Yes. Create one tracker per client, each with its own prompts, brands, engines and schedule, and they all bill from the same prepaid balance. There are no seats or workspace fees. Clients do not get their own login, so you deliver results in your own reports.

Can I white-label Cite42 reports?

Not as a dashboard. Cite42 returns data over MCP and REST, so the report is whatever you build from it, under your own brand. Rankscale and Peec AI offer white-label dashboards, and OtterlyAI has a Data Studio connector for branded reporting.

Set up the first client

Create a Cite42 account, which includes $1.00 of credit, connect the MCP server to Claude, ChatGPT, Codex or Cursor, and run one client's 10 questions as a baseline. Save them as a weekly AI Visibility tracker and pull the report when the next client update is due. The pricing guide has the full plan comparison.