A new Burson study called The Credibility Paradox found that, in generative engine optimization (GEO), fact-based claims regarding a company's innovation, products and workplace culture are more believable than claims around more subjective qualities like leadership and governance.
Burson partnered with AI marketing platform Profound to field thousands of reputation-related answers across seven major AI answer platforms, evaluating 85 companies in 10 industries.
The firm examined eight building blocks of reputation capital: innovation, financial performance, creativity, leadership, governance, workplace and citizenship.
Using Burson's proprietary Decipher tool, developed with cognitive AI firm Limbik, responses were scored for believability across three audiences: the general population, opinion elites and business decision-makers. This process generated over 55,000 believability forecasts.
Steve Rubel, EVP of media insights and measurement at Burson, said interpretation is the most important factor when it comes to AI answer optimization.
Business decision-makers rated AI-generated answers as 10% more believable than the general population on average. Opinion elites and business decision-makers ranked innovation first while the general population placed workplace and products in the top two positions.
The top four most credible levers were found to be innovation, creativity, workplace and products, in that order.
Companies in the tech industry were given an average credibility ranking of 1.5, with 1 being the most credible and 10 being the least. The aero industry received an average of 1.6, losing credibility on their products, but slightly outperformed the tech industry in the workplace category.
The energy industry had by far the least credibility in AI responses, receiving an average rank of 10, coming in last on every dimension Burson measured.
The telecom industry received an average of 3.9, being carried by its score of 2 in products, but losing credibility on innovation and creativity with scores of 6 and 7, respectively.
“The main thing is that there is this paradox, where you can be visible in something that you don’t want to be visible in. Or you can claim to be there and mark that as success, but people might view it very skeptically,” said Rubel.
“Many people are treating the answers as a uniform thing,” he said. “The reality is that different audiences ask different questions in different kinds of ways, and they interpret them in different ways.”
“Workplace” ranked number three across all audiences. The evidence behind a company’s reputation as an employer and its workplace culture is often externally observable and easier for audiences to verify.
“What’s interesting about workplace is there is a large body of evidence on the internet,” Rubel said. “There are numerous rankings and reports that come out about places to work. Companies very often do a lot to showcase that they are a great place to work.”
Conversely, “leadership” ranked at or near the bottom in believability across all sectors. Burson labeled this “one of the clearest liabilities in AI-mediated reputation.”
“Leadership is the hardest believability test to pass on AI. I think the realization there for us was that it’s less tangible,” Rubel said. “If I say, ‘Who or what is a leader?’ to 10 different people, I would get 10 different answers.”
Rubel recommended tying the leadership criteria to the other seven elements, allowing them to demonstrate a sound and credible leadership environment.
“These are not containers, there’s overlap between all these things. I would say it's really about how you [develop] a total body of evidence,” said Rubel. “And importantly, that the CEOs and other executives of the companies are out there delivering that same evidence.”
Burson is owned by WPP. Like-for-like revenue in WPP’s PR arm, which largely consists of Burson, fell 2.6% in Q1 to £157 million ($196.25 million). That compares to a 6.6% decline in the same period last year.
According to PRWeek’s Agency Business Report 2026, Burson reported global revenue of $860 million in 2025, down 6% from 2024’s $915 million. In the U.S., the firm reported revenue of $352.5 million, also down 6% from the prior year.
