What makes an AI assistant recommend one company and ignore another?

Signals is the internal research method used in Crafty Pixels' AI visibility engagements.

SEO reports could tell our clients where they ranked. They could not explain why an assistant recommended a competitor, why it trusted a directory instead of the company website, or why it recognised a brand only after the buyer mentioned its name. We needed to see the answers for ourselves, the sources behind them, and enough buyer questions to tell an isolated result from a pattern. That work became Signals. The software handles the repetitive collection; we conduct the investigation.

Crafty Pixels has spent years working inside complex marketing and revenue systems. That experience matters. An AI visibility finding is only useful when the people interpreting it understand the content, data, technology and internal constraints behind it.

A B2B company was easy for AI assistants to identify when buyers already knew its name — it appeared in seven of ten branded tests. Across thirty unbranded questions from buyers looking for the service it provides, the company was cited once. The assistants were relying on directories, specialist publishers and comparison articles; those sources controlled the shortlist before the company entered the conversation. The final recommendation was not 'produce more content' — it identified the missing pages, outside sources and entity evidence most likely to change the result.

A written intelligence brief for the people deciding what happens next. It contains the buyer questions we tested, the answers we observed and the evidence behind them. It shows which companies kept appearing, which sources shaped the recommendations and where the client's public story broke down. Each recommendation is tied to the finding that produced it. The evidence is collected by Signals. The conclusion comes from us.

When AI discovery has become a real commercial question. Perhaps competitors are appearing in answers where your company is absent. Perhaps leadership wants to know whether AI visibility deserves investment. Perhaps the marketing team has plenty of activity but cannot see which work would affect recommendations. The investigation gives the team a starting point based on observed buyer conversations rather than a generic checklist.

Signals is not offered as a standalone subscription. Clients receive its research and evidence as part of an investigation. The public scan on this page is a demonstration of the collection layer, not a substitute for one.

Every investigation is reviewed and delivered by a senior practitioner. That sets a natural limit on how many we can accept. We keep the limit because the interpretation is the valuable part of the work.

We review the company, the question and whether the findings can lead to meaningful action. Suitable engagements begin with a short working session before the investigation is scoped.

Some clients take the findings to their own teams. Others ask Crafty Pixels to make the changes, coordinate the work and repeat the investigation later to see what moved.