Clients keep asking about ChatGPT visibility and you have no answer: how agencies turn GEO into a service
How agencies turn GEO into a service when clients ask about ChatGPT visibility: what to sell, how to report, and what we run our own client programs on.
If clients keep asking your agency about ChatGPT visibility and you have no answer, you are not behind in knowledge as much as in structure. The agencies we see doing this well did not invent a new discipline. They took work they already did, pointed it at AI engines, and added a way to measure it. Here is how we have set it up for our own clients.
Say what you can and cannot promise
Do this first. You cannot guarantee that ChatGPT or any other engine will recommend a client. You can measure how often they appear, diagnose why they do not, and fix the causes that sit within your control. A service built on that promise is honest and sells fine.
The five parts of the service
A baseline. Before anything else, we show the client where they stand. A set of buyer prompts, run across the main engines, with who gets named and which sources are cited. This one deliverable is also the best sales tool we have. Clients who have never seen their own gap tend to want to talk about it.
Monitoring with a monthly read-out. Same prompts, tracked daily, summarized monthly. We report share of voice, position in the answer, sentiment and the trend. We spend most of the meeting on what changed and why, not on the charts.
Crawler access. A surprising number of sites block or break AI crawlers by accident. Fixing robots.txt rules, error pages and redirect problems is often the first visible win.
Content against gaps. We look at prompts where the client is absent and ask what the engines cite instead. Then we refresh existing pages or build new ones. Quality control matters here, and we review everything before it goes live.
Offsite presence. Engines lean on reviews, forums and third-party articles. We map where competitors appear and look for honest ways to get the client there.
What we run it on
Our client programs run on Promptwatch, and the choice was practical. The agency plans have unlimited projects, unlimited pitch projects and unlimited prompts, so adding a client does not change the platform bill in a way we have to explain. Ten seats are included. The Kick-off, Growth and Scale plans cost $199, $399 and $799 a month, and differ mainly by response volume and data allowances.
Here is how the features map to the five parts:
- Baseline: pitch projects let us run a prospect audit before a contract exists.
- Monitoring: daily prompt panels per model, country and language, with Visibility Score, Share of Voice, Sentiment and Position in Answer. State and city tracking is on every agency plan.
- Crawler access: Agent Analytics shows IP-verified crawler logs for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and others, with error tracking and a crawl-to-citation path per page.
- Content: content gap analysis shows where the client has no answer. Agency plans do not include AEO articles, the knowledge base or the automated Content Agent, so we plan content production around that.
- Offsite: citation analytics, offsite mentions, and Reddit, YouTube and LinkedIn reports.
For reporting, custom reports, the Looker Studio connector and PDF exports cover most of what clients ask for. Agent Chat helps us answer one-off questions from live data, and the MCP server and API let us pull numbers into our own tooling. Visitor analytics adds AI-referred traffic and conversions, which is how a retainer ties to business results.
Packaging
We avoid quoting a single price in public, because scope varies. We structure tiers around the variables that drive effort: how many prompts, how many markets or languages, how many engines, how much content per month, and how often we meet. A small pilot of three months, with a baseline, daily monitoring, one technical review and a short content backlog, is easier for a client to say yes to than an open-ended retainer.
Mistakes we see agencies make
Selling monitoring alone. A dashboard with no actions attached becomes a line item the client cancels. Always attach a diagnosis and a backlog.
Starting with a huge prompt set. Start with 50 prompts the client recognizes as their real buyers' questions.
Reporting a single answer. Because answers vary, report trends and rates.
Ignoring attribution. If you cannot show visits and conversions from AI engines, the service is hard to defend in a budget review.
Getting started this month
Pick two clients who have asked the question. Run a pitch project for each. Present the baseline in a 30 minute call. Offer the pilot. By the end of the month you will have a service, a template report and real experience to refine.
If you want to see how the agency setup works, start with the pitch projects on Promptwatch.