When a Client Sends an AI Screenshot: A Calm Workflow for Account Managers

Author : wipelaw twugg | Published On : 18 Nov 2026

An illustrative workflow for account managers who receive a client screenshot showing an AI assistant naming a competitor. It covers logging the missing context, checking the question against an agreed panel, rerunning it, reviewing description accuracy and replying without turning one answer into a score.  Get more information about AI visibility measurement

 

The message usually arrives with little context: a screenshot of an AI assistant recommending a competitor, and a short note asking why the client is missing. The instinct is to answer with a number. A better response treats the image as a lead worth logging, not as a measurement.

 

Step one is to capture what the screenshot cannot show. Ask the client, politely, which platform they used, the date, the language and the exact wording of the question. If they can paste the full prompt and answer as text, even better. Store everything with the original image in a shared folder so it does not live only in one inbox.

 

Step two is to compare the question with the agreed panel. If it matches a question already being tracked, the screenshot becomes one more observation beside the scheduled runs. If it is new and reflects how real buyers ask, propose adding it at the next review rather than slipping it into this month's figures.

 

Step three is to rerun the question several times on the same platform and record each result. AI answers vary from one run to the next. Arfadia, a Jakarta-based agency operating since 2008, found in its own test that the companies named for repeated English buyer questions overlapped only partially, which is why one answer cannot stand for a position.

 

Step four is to read for accuracy, not just presence. Where the client does appear, check whether its services, location and description are correct. A confident wrong description can be a more urgent problem than an absence, and it deserves its own column in the log.

 

Step five is the reply. Thank the client for flagging it, explain in a sentence or two that answers vary between runs and platforms, describe what was logged and say when it will be reviewed. Avoid words such as score or rank, because a single screenshot supports neither.

 

Throughout, keep the counting rules unchanged. Do not merge linked citations with bare mentions to make a month look better, and do not blend platforms into one figure. A practitioner guide on AI visibility measurement from an Australian search agency explains these rules in more depth.

 

Handled this way, the screenshot becomes genuinely useful. It may reveal a buyer question the panel missed or a description that needs fixing. What it should not become is a headline figure in the next report. Over time, a folder of logged screenshots can also show which worries come up repeatedly, which is useful material for the next planning conversation with the client.