AI search visibility.

The problem

When someone asks ChatGPT or another assistant for a recommendation, the answer names a handful of businesses. Most businesses have no idea whether they are in that handful.

The tools that promise an answer usually check each question once and report the result as fact. AI answers change from one run to the next, so a single check proves very little. Buying decisions get made on numbers nobody can defend.

This is measurable. When we measured one market to build our sample report, a map widget with star ratings decided the winners in 21 of 21 ChatGPT answers, some of the market's most recognized names never appeared at all, and the two assistants we measured barely agreed on anything.

What the audit measures

• How often assistants name you. Measured across many phrasings of the questions your customers actually ask, sampled repeatedly, and reported separately for each assistant.
• How you compare. The same measurement for competitors you name and for a comparison group of similar businesses measured the same way.
• Where the answers come from. The directories, review sites, and data sources assistants read when they answer. This is where the results become explainable.
• What to do about it. Specific findings tied to your public data footprint. The listings, sources, and pages that explain your results, with the evidence behind each one.

How we report

• Ranges, not scores. Because answers vary, every number comes with its margin of error. We do not publish blended scores or single-number ratings.
• The full question set, published. Every report includes every question we asked and how many times we asked it. You can check our work. That is the point.
• Honest limits. Some assistant behavior cannot be observed from the outside. Every report includes a section on what the measurement cannot tell you.
• Measurement, not a retainer. The audit is a fixed-price deliverable. We are not selling an optimization subscription on the back of it, so the findings have no reason to flatter anyone.

This work is sometimes called answer engine optimization, or AEO. We start earlier than that. Before anyone optimizes anything, you deserve a defensible picture of where you stand.

See exactly what you get: read a sample report built from a real measurement

Common questions

• What does an audit cost? A fixed price agreed before work starts. Scope drives the price: how many markets, competitors, and assistants you want measured. There is no retainer and no subscription.
• How long does it take? The measurement itself runs over multiple days, because AI answers vary and a single day of sampling is not enough. Most audits go from kickoff to delivered report in a few weeks.
• Which assistants do you measure? ChatGPT and other assistants that can be observed the way a real customer would use them. Some assistants cannot be measured honestly from the outside. We exclude those and say so in the report.
• Do you guarantee better visibility? No. Nobody controls what an AI assistant says, and anyone who guarantees a position is selling something they do not control. We guarantee the measurement: how it was run, how many samples, and what the numbers can and cannot support.
• Will you fix what the audit finds? The findings are specific enough to act on, and you can take them to anyone, including an agency you already work with. If you want Recursive Systems involved afterward, that is a separate conversation. The audit never depends on it.
• Why not just use an AI visibility tool? Tools are useful for watching trends. Most check each question a small number of times and report the result as a single score. If you need a number you can defend to a partner or a board, you need repeated sampling, published questions, and stated margins of error.
• What if the results are bad? Then you will know, and you will know why. The report ties results to the sources assistants actually read, which is where improvement starts. We do not soften findings.