Showing up in AI answers isn't enough if audiences don't believe them

Brands may be winning the battle for visibility in AI-generated answers, but that visibility means little if audiences do not trust what they are reading. That's according to a new report from Burson, which argues that the next phase of generative engine optimisation (GEO) will be defined not by whether brands appear in AI-generated responses, but by whether those responses are believable to the audiences that matter most.

The report, titled "The credibility paradox", analysed more than 55,000 AI-generated responses across seven major AI answer platforms, examining how 85 companies were represented across eight reputation dimensions: innovation, creativity, workplace, products, financial performance, governance, citizenship and leadership. The findings suggest that while many organisations have focused on securing visibility in AI-generated answers, communicators now face a new challenge: ensuring that the narratives AI systems construct are credible.

Conducted in partnership with AI marketing platform Profound, the study evaluated AI-generated responses using Burson's proprietary Decipher tool, developed with cognitive AI company Limbik. The tool assessed the believability of responses among three audience groups: the general population, opinion elites and business decision-makers. One of the report's strongest findings was that AI rewards proof over positioning.

Don't miss: Does reputation have a price tag? Burson puts it at US$7.07 trillion globally

AI rewards proof, not positioning

According to the report, reputation attributes supported by observable and independently verifiable evidence consistently outperformed those built around more subjective or institutional claims. Innovation ranked as the most believable reputation lever overall, followed by creativity, workplace and products. Meanwhile, governance, citizenship and leadership occupied the bottom half of the rankings.

In tandem, it found that AI-generated responses grounded in product launches, customer experiences, awards, workplace reviews and third-party coverage were generally more believable than narratives centered on corporate intent, executive judgement or organisational values. In fact, the data showed a strong directional separation between performance and foundational levers, arguing that in AI environments, proof is no longer simply support for a message but the infrastructure that makes the message credible.

Interestingly, leadership emerged as the weakest reputation dimension across almost every industry and audience group studied. Executive visibility and CEO commentary rarely translated into believable AI-generated narratives unless they were reinforced by governance practices, business performance, employee experiences and third-party validation.

In fact, leadership ranked among the bottom two reputation dimensions across every sector examined, making it one of the clearest vulnerabilities in AI-mediated reputation management. The findings may challenge traditional approaches to executive thought leadership, with Burson arguing that communicators should focus less on executive visibility itself and more on building what it described as "credible, extractable and externally validated leadership proof".

Meanwhile, workplace reputation emerged as an unexpectedly powerful credibility driver. Workplace ranked first among the general population and remained in the top half across all audience groups. According to the report, workplace narratives perform strongly because they are supported by highly visible and independently verifiable signals, including employee reviews, labour reporting, talent rankings, workplace policies, hiring activity and earned media coverage.

Different audiences trust different signals 

The study also found that credibility varied significantly depending on who was consuming the AI-generated response. Business decision-makers rated AI-generated answers 10% more believable on average than members of the general population. While business decision-makers and opinion elites ranked innovation as the most believable reputation attribute, the general population placed workplace and products at the top of the list. This means organisations need a more audience-specific approach to GEO.

For example, an innovation-led narrative may resonate with business audiences who understand technical capabilities and industry context, but prove too abstract for broader audiences unless it is translated into tangible customer outcomes or visible product benefits.

Beyond audience differences, the report also highlighted significant variations between industries. Technology emerged as the most credible sector in AI-generated answers, maintaining a top-two position across all reputation dimensions. Aerospace also performed strongly, ranking within the top three across every category measured, including governance, leadership and financial performance.

At the other end of the spectrum, energy recorded the lowest credibility scores across every reputation dimension studied. Burson attributed these differences to what it calls "credibility burdens" and "translation burdens".

A credibility burden occurs when audiences understand a company's claims but require stronger proof before accepting them, particularly around areas such as governance, sustainability, transparency and social impact. A translation burden, meanwhile, affects sectors such as industrials, infrastructure and complex B2B businesses, where the challenge lies not in trust but in helping audiences understand the value being created. In these cases, companies need clearer explanations, stronger category education and more visible proof points rather than simply producing more content.

The report also warned against applying a uniform GEO strategy across global markets. While AI systems generally favour verifiable and performance-based proof, the information ecosystems they rely on differ significantly across regions. In markets such as China, Japan and South Korea, AI systems often draw more heavily from local platforms and closed digital ecosystems, reducing the influence of sources that may carry weight in Western markets. As a result, companies should spend less time optimising individual pieces of content for AI platforms and more time building locally relevant proof systems that AI can access and validate.

The report arrives as marketers, communications professionals and reputation specialists increasingly explore GEO strategies to improve visibility across AI platforms such as ChatGPT, Gemini and Claude. However, Burson argues that success in AI-mediated environments will depend less on securing "share of answer" and more on building what it calls "evidence ecosystems" – a consistent mix of earned, owned and social content reinforced by independent validation.

"As AI becomes an increasingly influential layer between companies and their stakeholders, it is shaping not only how brands are discovered, but also how they are understood and evaluated," said Red Surtida, APAC head of intelligence and transformation at Burson.

"The real opportunity for organisations is not simply to secure share of answer, but to ensure those answers are grounded in evidence, backed by credible sources, and believable to the audiences that matter most," added Surtida. 

Related articles:   
Indonesian CEOs move beyond AI experimentation as leadership structures evolve
APAC trust gap hits record high as income disparity doubles  
AI use rises, but so do trust demands from Singaporeans