Two AI engines name the same brand in 81% and 43% of their answers

What happened

• Treyci, an AI visibility measurement platform, published an analysis of more than 1,200 scored answers on September 3, 2026.
• In one enterprise software category, one engine mentioned the tracked brands in 81% of its answers and another mentioned them in 43%, in the same month and with the same queries.
• The stated methodology runs about 100 purchase-intent questions per category (price comparisons, alternatives, recommendations by team size), repeating them on ChatGPT, Perplexity, Gemini and Grok.
• In a separate crawl of 100 enterprise software companies, 41 had published an llms.txt file for AI crawlers.

Why it matters

• A marketing team that reports brand visibility to its board by checking a single engine delivered a figure another sample would have cut in half.
• For Chilean small businesses that started appearing in AI answers without really knowing why, the finding organizes the conversation: visibility isn’t one number; it’s one number per engine and per month.
• The llms.txt file is being adopted faster than anyone can check whether it works. Forty-one companies published it; none showed that it changed anything.

The number

81% versus 43%. The same brands, the same questions, the same month, and almost a twofold difference between one engine and another.

Context

Treyci sells the monitoring tool that produced the finding and didn’t identify the two engines, the software category, the query set or the scoring criteria. With that as published, nobody can tell whether the gap comes from the engines’ behavior, from how the questions were worded or from the brands the company chose to track. The same day, another vendor documented the rise of paid advertising inside those answers.

What’s next

• Treyci didn’t announce publication of its query set or its scoring rubric.
• No timelines announced for replicating the measurement in other categories.

Bottom line

For twenty years the debate was about where a brand appeared within a single list of results. Now the debate is about which engine it appears in, and the answer changes depending on which one you ask.

Sources

• New Measurement Data: AI Engines Disagree by 2× on Which Brands to Recommend — press release on GlobeNewswire, September 3, 2026
• Treyci: AI Engines Disagree 2x on Which Brands to Recommend — MarTech Cube

Edited by Rodrigo Cornejo. How we select and verify the facts, in who writes.