Skip to article
auraGEO
IN PRACTICE · Observations and results

How agencies should report clients' AI visibility

An agency report should connect work to evidence: what changed, what was published, which AI answers were obtained and which website events were observed. These are four separate layers. Do not replace them with a general ‘promotion score’ or promise ChatGPT query volume when you only have your own question panel.

Keep client projects separate

The website, domain, questions and connection must correspond to the specific client. Do not transfer another project's successful answer as the result for a new website. To compare several clients, agree on a common methodology beforehand and separately account for different offerings and audiences. A shared workspace helps organize work but does not make dissimilar industries into a fair ranking. Restrict a connected agent to selected projects and allowed actions so a report does not expose someone else's data.

Show work that can be verified

A useful row contains a specific action: instructions updated at a stable URL, material published or a detected block removed. Beside it, include the date, link, original rationale and live-check result. An accepted search submission is a separate stage. If a task is awaiting the owner, leave that status. A prepared patch cannot be presented as installed code, or a draft as a published article, merely because the agent finished its text response.

Then show observations

For the panel, state the questions, platforms, collection method and coverage. Distinguish the brand name, direct link and explicitly positive context. The educational label ‘a link was found in three of ten saved answers’ is better than an unsupported market-share claim. Actual figures come from the specific client; an example is not presented as a case study. If some checks are blocked, show them separately rather than treating service unavailability as a lack of recommendations.

Connect the findings to business without additions

The tracker's own events allow discussion of recognized referrals and, separately, goals configured for the client. They do not prove every visitor was a buyer or reconstruct all unknown sources. When examining changes, check that periods match and respect retention limits. If there are no inquiries yet or a goal is unconnected, the appropriate next step differs: improve the landing page, accumulate data or configure backend event confirmation rather than promise growth from an attractive graph.

Organizing this in AURA GEO

Through MCP, the agent can read the action log, summaries and client report. AURA GEO currently shows five monthly support levels: Free at €0 for one website; Starter at €19–29 for one website; Pro at €49–79 for up to three websites; Agency at €149–299 for up to five client projects; and Managed GEO at €500–2,000+ with scope agreed per project. These ranges describe the proposed package structure, not an active payment or subscription. Ongoing support still requires an executor that actually runs and the required permissions; MCP alone does not automatically perform recurring checks of every AI service.

Conditions and limitations
  • Connection, a technical check and a saved report do not guarantee an AI recommendation or a customer outcome.
PRIMARY SOURCES

Explore the details

  1. AURA GEO MCP documentation
  2. AURA GEO privacy
  3. Public AURA GEO early-access plans
NEXT STEP

A verifiable action plan and an explanation of the result with saved evidence and limitations.

Open the project summary
All articles