Tools

Build a prompt panel you can run again

This ai visibility prompt list builder creates twenty buyer-style questions from your business category, location, services, and name. Download the CSV and use the same panel whenever you check how an AI assistant answers questions about your market.

The builder does not ask an AI model anything. It creates a repeatable test set. Asking a new collection of questions every month makes any change impossible to interpret. Keep the panel stable, record the answers, and investigate the sources behind them.

The five parts of the panel

Discovery prompts ask how a buyer would choose a provider in the market. Service-fit prompts ask when a service is needed and what a complete scope looks like. Cost and process prompts focus on quotes, deliverables, and timelines. Brand prompts mention the business directly. Comparison and objection prompts test alternatives, tradeoffs, and reasons someone might switch.

That mix avoids a measurement trap. If every prompt includes your brand, the panel is measuring recognition of your name, not whether you are discoverable when someone describes a need. If none mention your brand, you miss whether the assistant can explain your actual offer and terms. Both views belong in one panel, with the categories kept visible in the CSV.

How to run a useful check

Use the same model, prompt text, and approximate test conditions each time. Save the date, model, response, cited sources, and whether the answer accurately represents the business. A model can answer the same question differently on another day or with different context, so one response is not a durable rank. Look for repeated patterns and factual errors instead of declaring a position.

The location field should match the market you actually serve. The services field should use what a buyer calls the work, not internal package names. The optional buyer description makes questions more specific; a homeowner or a clinic manager changes the decision the prompt is asking about without asserting that you have worked for either.

What to do with the answers

If an assistant gives a wrong business detail, first check the page it appears to have used. If the site itself is unclear or contradictory, fix the public source. If the answer cites an old third-party listing, correct that listing where you can. If the model gives no source, record the uncertainty rather than attributing the answer to your latest page edit.

The CSV is a starting panel, not a proprietary score. Add or remove prompts once you understand the buyer questions in your market, but save the old panel when you change it so later comparisons remain honest. A question that never resembles a real buying decision should not stay in the panel just to reach a round number.