How can my agency offer AI visibility (AEO) tracking to clients?

Agree a fixed list of 20 to 50 buyer questions with each client, run them on a schedule in ChatGPT, Perplexity, Google AI Overviews and AI Mode, Gemini, Claude and Copilot, and record whether the client is named, whether its pages are cited, and which competitors show up instead. Add GA4's AI Assistant channel for the visits that do arrive. Sell it as a monthly report section plus a short fix list, and tell clients up front that AI answers change from run to run.

What AI-visibility tracking actually measures

Rank tracking asks where a URL sits for a keyword. An AI answer has no position one, so you measure four other things.

• A prompt set per client. Questions written the way a buyer talks: "best bookkeeper for Shopify sellers in Austin". Mix category questions, comparisons and questions that name the client. Keep the list stable, or month-over-month numbers mean nothing.
• The engines. ChatGPT, Perplexity, Google AI Overviews and AI Mode, Gemini, Claude and Microsoft Copilot. They lean on different sources, so a client can do well in one and be missing from another.
• Mentions and citations. A mention is the brand named in the answer. A citation is a link to the client's own page as a source. They move separately: a brand can be recommended off a review site with no link to its own pages. Track both, plus who else is named.
• Traffic. In May 2026 GA4 added a default "AI Assistant" channel for recognized AI referrers, with Google naming ChatGPT, Gemini and Claude as examples (Search Engine Journal). It is the closest thing to a business number here, and it undercounts.

Doing it yourself, step by step

1. Write the prompts with the client. Pull them from sales calls, support emails and Search Console queries that read like questions. Get the list approved in writing so nobody argues later about what was measured.
2. Pick a tool, or script it. Otterly.AI starts at $29 a month for 15 prompts and $189 for 100, covering ChatGPT, Google AI Overviews, Perplexity and Copilot, with Claude, Gemini and AI Mode as paid add-ons. Ahrefs Brand Radar lists custom prompt tracking from $50 a month across AI Overviews, AI Mode, Gemini, Copilot, Perplexity and ChatGPT. Peec AI and Semrush's AI Visibility Toolkit are other common picks. Prices change often, so check before you quote a client. A script calling each model's API with web search on also works, but API answers can differ from what a buyer sees in the app.
3. Run each prompt more than once. Several runs per engine per month, then report the share of runs where the client appears. "Named in 7 of 10 runs" holds up in a client meeting. One screenshot does not.
4. Keep the raw answers. Full text and cited URLs for every run, dated. When a client asks why a number moved, you need the answers.
5. Benchmark two or three competitors the client names, on the same prompts and runs.
6. Pull GA4 monthly. AI Assistant sessions and conversions by landing page. Look at Referral and Direct too.

Packaging it as a client deliverable

Agencies already sell this. A vendor post on Product Hunt describes audits at EUR 1,200 to 2,500 and monitoring at EUR 800 to 1,500 a month. SE Ranking's 2026 reporting guide puts AI visibility in its own section after organic traffic, framed as an emerging channel.

The monthly section

• Share of prompts where the client was named, per engine, against last month.
• Citations to the client's pages, and which pages.
• The competitors named most often.
• AI Assistant sessions and conversions from GA4.
• Three to five answers quoted word for word.
• What you changed this month and what comes next.

What to fix

• Answer-first pages. One page per real buyer question, opening with a one or two sentence answer that is easy to quote.
• Structured data. JSON-LD for the organization and its services, and FAQPage where there is a real FAQ. Google says AI Overviews need no special schema, so sell this as hygiene, not a trick.
• llms.txt. A short plain-text map of the site at /llms.txt. Cheap to add, but Google has said it does not use it (Search Engine Land), so promise nothing from it.
• Third-party listings. Check which sites get cited for the client's prompts: review sites, directories, comparison articles, forums. Competitors' mentions often come from there. Get the client listed, with the same name, prices and service area everywhere.

What can go wrong

Answers change every run

The same prompt on the same day can name different brands, and location, login state and model updates all shift it. Report rates over many runs, never one answer as proof.

Small samples that look like trends

With fifteen prompts run once, one changed answer moves the score about 7 points. Say how many runs sit behind each number.

Attribution stops at the click

Most AI answers never produce a visit, and visits without a referrer fall into Direct. You cannot tie a mention to a sale. Say so in the proposal, so the retainer is not judged on GA4 alone.

Switching tools, promising rankings

Each tool queries engines differently, so pick one per client and keep it. And there is no position to guarantee: sell measurement and a fix list.

When it is worth paying someone

A tool plus one owner works for a handful of clients, if that person keeps the prompt lists current, reads the raw answers and writes the section each month. Pay someone when that upkeep across ten or more clients is the problem.

That upkeep is part of what we sell. We set up and run an AI operator for agencies that runs the checks per client, keeps every answer, and drafts the monthly section for your account manager to approve. We have not run it for an agency yet. We run AI-visibility checks on our own sites weekly, on the same setup that runs the rest of our studio:

Related: How to automate weekly client reports and Monitoring client sites with an AI agent.

Questions