Burson released a study on June 2, 2026 called The Credibility Paradox that should reset what every public relations firm sells to a board. Across 85 companies, 7 AI platforms, and 55,000 believability forecasts, the finding that matters: leadership-related answers ranked among the least believable across every industry tested. For the discipline of public relations, this is the most consequential research a holding-company desk has put out this year.
Published June 24, 2026.
Related: The Leading PR Firms in 2026 — The U.S. Index · Generative Engine Optimization · Reputation Management · Reuters Says 10% Get News From Chatbots.
Fact block
- 5 — AI engines that now define visibility: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews.
- More than a third of buyers begin product research inside AI, not Google.
- 60–70% — share of most 2026 GEO budgets spent on owned-content citation work. Half that spend is producing the Believability Discount.
- $96M — Profound's February 2026 Series C at a $1B valuation, led by Lightspeed alongside Sequoia and Kleiner Perkins. The category infrastructure is being built in public.
- $860M — Burson's reported 2025 global revenue, down 6% from $915M. U.S. revenue $352.5M, also down 6%. Methodology with a motive.
- 23 years — 5W AI Communications, founded 2003, across two structural shifts in earned media.
What Burson actually did
Burson partnered with Profound, the New York-based AI marketing platform that just closed a $96M Series C at a $1B valuation, and Limbik, the cognitive AI company that built Burson's proprietary Decipher tool. The team fielded reputation-related answers across seven AI answer platforms, covering 85 companies in 10 industries, and scored every response for believability against three audiences: general population, opinion elites, and business decision-makers. Total believability forecasts produced: 55,000.
The companies were scored across eight reputation levers from Burson's existing Reputation Capital framework — innovation, creativity, workplace, products, financial performance, governance, citizenship, and leadership.
The finding that matters
Leadership ranked at or near the bottom of believability across every industry. Aerospace and tech were the only two sectors that scored higher, and Burson is direct about why: credibility in those industries came from governance structures, business performance, and external validation — not from executive messaging.
Translated: when a CEO speaks, the AI engines retrieve it, and the audience discounts it. When a third party speaks — a regulator, an analyst, a tier-one publication, an employee review platform — the engines retrieve that too, and the audience trusts it. Same retrieval mechanics. Opposite outcomes.
What Burson's leadership said
Corey duBrowa, Burson CEO: "In today's zero-click world, LLMs have become the new gatekeepers of reputation — how brands are discovered and evaluated. But visibility is not credibility. Showing up in these LLMs is necessary but not sufficient."
Steve Rubel, EVP, Media Insights and Measurement: "GEO may have started as a visibility challenge measured through audit reports. The data from this study makes clear it is now a test of whether a company's real-world reputation is legible, corroborated and believable."
The number behind the study
Business decision-makers rated AI-generated answers 10% more believable on average than the general population. Workplace, products, and innovation were the three most credible levers across all three audiences. Citizenship, governance, and leadership were the three least.
That ordering inverts what most public relations programs spend their time on. Most retainers in 2026 still treat the CEO interview, the executive op-ed, and the leadership profile as the highest-value placement. The Burson data says those are the placements the engines retrieve and the buyers discount.
Why Profound matters here
Profound, co-founded by James Cadwallader and Dylan Babbs in 2023, created the Answer Engine Optimization category and now runs the largest opted-in panel of real AI user prompts in the market — 1.5 billion and counting. The platform tracks brand visibility across ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, Meta AI, and DeepSeek, and ships a public Profound Index ranking brands by AI Search visibility across 50+ industries. Burson did not pick Profound randomly. The Credibility Paradox study runs on the cleanest panel data the category currently produces.
The context Burson left out
According to PRWeek's Agency Business Report 2026, Burson reported global revenue of $860 million in 2025, down 6% from 2024's $915 million. U.S. revenue came in at $352.5 million, also down 6%. The Credibility Paradox is good research — and it is also the kind of category-defining methodology a holding-company firm releases when growth has stalled and the offering needs a reset.
That does not invalidate the data. It does explain why this study, this quarter, from this firm, with this much methodology stacked behind it. The category is repricing. Everyone is looking for the next number to sell to a CMO.
What this means for the discipline of public relations
Three operating shifts follow from the data:
- Move budget out of executive thought leadership and into workplace, products, and innovation coverage. The engines weight third-party proof; the audience trusts it. Burson's data names the levers.
- Audit every client's AI-engine answer mix by reputation lever, not just by share of voice. A brand can lead the citation count and lose every lever that moves the buyer.
- Treat the CEO as a citation surface, not a content surface. The engines will index every appearance. The audiences will discount every appearance. The work is to surround the CEO with third-party evidence the engines retrieve alongside the executive citation.
Where the rest of the category is
Five months past the Omnicom-IPG closure, the holding companies are putting out methodology. The independents are putting out work. The next 12 months will show which approach buyers pay for. Burson's study is now the floor; firms that cannot match it on rigor will be measured against it for the rest of the year. 5W AI Communications already runs its own AI Citation Audit across the same five engines, with a locked scoring formula and quarterly reporting cadence. Edelman, Weber Shandwick, and FleishmanHillard have not yet published equivalent methodology.
The takeaway
Public relations spent 18 months arguing about whether GEO was a real category. Burson's study is the end of that argument. The category is real, the metric is believability, and the lever that loses across every industry is the one most retainers still over-index on. The firms that move budget this quarter will compound. The firms that do not will be writing case studies about the firms that did.
