A new analysis from Burson modelled believability across seven AI platforms and found that being mentioned in AI answers is not the same as being convincing. Concrete claims and business audiences scored higher, abstract leadership talk scored lower.
Showing up in the answer is table stakes. Sounding believable when you get there is the actual work.
Everyone is chasing brand mentions in AI answers. A new analysis says the mention is only half the job. Communications agency Burson modelled how believable AI answers are, not just whether a brand shows up, and called the result the Credibility Paradox.
Burson asked seven AI platforms questions about 85 companies, then ran the answers through a model it built called Decipher, producing more than 55,000 believability scores. Concrete claims about products and workplace culture scored higher than abstract talk about governance or leadership. Answers aimed at business audiences rated about 10% more credible than others.
The believability scores Burson generated across seven AI platforms to test whether AI mentions actually convince. Source: Search Engine Journal, June 2026.
Why it matters
The race to get cited by AI has turned into another vanity metric. Brands are counting mentions the way they used to count rankings, and a mention you cannot trust does not move a customer any more than a ranking nobody clicks. The Australian businesses pouring effort into AI visibility need to ask the second question, not just the first.
The pattern in the data is useful. Concrete beats abstract. Specifics about what you sell and how you work land harder than leadership platitudes. That is the same thing good marketing always rewarded, now measured inside the machine.
What to do about it
A mention with no credibility is noise. The work is being the answer people believe, not just the one that appears.