AI visibility tracking for agencies

Recomma (recomma.ai) measures how ChatGPT, Gemini, Google AI Overviews and AI Mode describe each of your clients, from one prompt budget you spend where it is needed — and turns every gap into work you can show was worth doing.

One prompt budget, spent where you choose

A prompt is a question you want a client named in the answer to, asked on a schedule on every engine the project tracks. Recomma counts prompts across the whole workspace rather than per brand, so the client in a crowded category can have forty questions and the quiet one can have five. Per-brand allowances push an agency into tracking its biggest accounts properly and guessing at the rest; a pooled budget lets the coverage follow the work.

How many clients each plan holds

Each client is a project. The plan sets how many projects the workspace can hold and how many prompts they share:

Competitors and team members are unlimited on every plan, so adding a strategist or a client's rival costs nothing. Every plan samples daily, and a client's long tail can still go on weekly sampling — cadence is set per project.

  • Starter — $30/mo · 1 brand · 50 prompts pooled · sampled daily
  • Growth — $299/mo · 3 brands · 300 prompts pooled · sampled daily
  • Scale — $999/mo · unlimited brands · 500 prompts pooled · sampled daily

Numbers a client cannot argue with

An AI answer is sampled, not looked up: ask twice and the brands in it can change. Every figure Recomma reports — visibility, share of voice, position, sentiment — carries the number of answers behind it and a 95% confidence interval, and a figure whose range is wider than the reading is shown as the guess it is. When a client asks whether last month's movement was real, the answer is already on the page. The reasoning is in why every number comes with its error range.

Work you can show was worth doing

Each gap becomes a ticket: a source that keeps feeding the answer and skips your client, or a markup fault on their site from the 15-rule audit. When a ticket closes, the reading on its prompts is frozen, and fourteen days later the same questions are asked again on fresh answers. The verdict is kept whichever way it goes — so a quarterly review is a list of what moved the number, not a chart and a hope.

From the client's repository

Everything in the workspace is available over MCP, so Claude, ChatGPT, Claude Code, Codex and other clients can read a client's answers, gaps and site findings where the fix happens. The tools that change anything preview first and need a separate write scope, so an agent can propose prompts or close a ticket but cannot change a workspace by accident. The walkthrough is in fixing what AI search gets wrong from Claude Code.

Or hand the work to us

For clients you would rather not staff, Recomma's team works the queue: managed services start at $1500 a month for the pages a client's own site is missing, each one re-measured against its baseline. The tiers are on the pricing page.

Comparing tools for your agency?

We have written up how Recomma sits beside the tools agencies usually shortlist, with every fact about them linked and dated: Otterly.AI alternatives and Recomma vs Peec AI. Recomma does not offer white-label reports today; if your clients need your brand on every chart, check that first.

Start with one client

Add a client's domain. We read the site and propose the topics, questions and competitors; you keep what you want, and the first answers are captured while you set up the next one.