Public trust in AI rises as media sentiment turns more sceptical
CARMA finds AI trust rose to 30% as media trust signals fell. What the gap means for transparency, accountability, and AI messaging.
CARMA says public trust in AI rose from 26% in 2025 to 30% in 2026, even as media trust signals fell from 35% to 26%. The details were outlined in the company’s study write-up.
That gap matters for marketers because it suggests AI perception is no longer shaped primarily by headlines. It is increasingly shaped by lived product experience, and by whether organizations can explain how their AI actually works.
One strategic tension is easy to miss: brands often assume “more positive coverage” equals “more audience trust.” CARMA’s results point to the opposite risk. When audiences are forming opinions through direct use, media narratives can lag behind reality, and corporate messaging can get penalized for sounding like it is still selling novelty.
Table of contents
Jump to each section:
- Why the trust gap is widening
- Transparency beats reassurance in AI messaging
- Accountability expectations are diverging
- Visibility is being driven by actions, not campaigns
- What this means for marketers
Why the trust gap is widening
CARMA analyzed 6,006 articles from 500 tier-one media outlets across 45 markets (Jan to Jun 2026) and surveyed 6,300 people across 19 countries. The top-line change is small in absolute terms, but directionally important: public trust increased to 30% while media trust signals declined to 26%.
A useful way to interpret this is that AI is shifting from “topic” to “infrastructure.” When a technology becomes a layer inside everyday tools, audiences do not wait for media framing to decide what they think. They decide based on whether the outputs feel dependable, and whether the system behaves predictably in real use.
Memorable observation: Once AI becomes a daily workflow layer, reputation is built through repetition, not press.
The other notable shift is emotional. For the first time in CARMA’s tracking, trust overtook excitement as the public’s leading emotional response to AI (33% vs 29%). That is what “normalization” looks like: less novelty, more evaluation.
ContentGrip
Transparency beats reassurance in AI messaging
CARMA found that audiences associate trust with transparency about how AI works and how decisions are made. Media coverage, by contrast, places more weight on safety, reliability, and accountability.
That distinction matters because it reframes what “credible AI communication” should look like. A brand can say it has guardrails, testing, and safety processes, and still fail to build trust if people cannot understand why a model produced an outcome or what data and rules shaped it.
CARMA’s survey also signals a messaging penalty for vague AI positioning:
- 59% of respondents said companies exaggerate their use of AI and rely on buzzwords.
- 44% said they pay less attention to messages that mention AI.
Meanwhile, messages framed around productivity, efficiency, accuracy, speed, and scalability were viewed as more meaningful.
Memorable observation: In AI messaging, “safe” is table stakes. “Explainable” is the differentiator audiences are asking for.
A practical implication follows: marketers should expect “AI-powered” to function less like a benefit and more like a claim that must be substantiated, especially when audiences already feel saturated with AI language.
Accountability expectations are diverging
CARMA highlights a clear split on who should be responsible when AI causes harm or makes mistakes:
- Media coverage assigns the largest share of responsibility to governments and regulators (35%).
- The public places AI companies and developers first (33%), with governments and regulators last (10%).
This is not just a policy argument. It is a brand risk map.
If audiences believe developers and AI companies hold primary responsibility, then brand teams using AI systems inherit some of that accountability in the public mind, even if the legal liability is complex. “We comply” can sound like deflection if users want to know who owns the decision, who fixes errors, and what recourse exists when outcomes go wrong.
Memorable observation: Accountability is becoming a product feature, not a legal footnote.
CARMA also notes AI governance is becoming more tied to geopolitics and national competitiveness, including technological sovereignty, export controls, semiconductor competition, and national AI infrastructure. Even for marketing teams, this can surface in procurement questions, vendor selection, and how cross-border AI capabilities are discussed.
Visibility is being driven by actions, not campaigns
CARMA’s analysis of media attention shows OpenAI remained the most visible AI company, with 21% share of voice. The biggest mover was Anthropic, rising from 4% in 2025 to 17% in the first half of 2026, overtaking Google for second place in global coverage.
The reason CARMA cites is instructive: visibility was attributed to Anthropic’s public clashes with the US government over military AI and export controls, not a communications campaign. The episode also drew attention to its AI safety positioning by highlighting how actions aligned with stated principles.
That pattern is a reminder for brand teams: in AI, narrative is increasingly audited against behavior. Thought leadership without operational proof will struggle, while tangible decisions, constraints, and governance stances can shape perception faster than messaging.
CARMA also found CEOs generate the most positive coverage as AI advocates, but academics are the most trusted voices among audiences. The more interesting question is what that does to brand credibility. It suggests “executive optimism” is not the trust engine; perceived independence and expertise are.
What this means for marketers
Marketers are operating in a new perception environment where trust can rise even as media scrutiny hardens, and where audiences punish empty AI signaling more than they reward AI labels.
- Treat AI claims like performance claims, not brand adjectives
If 44% pay less attention when AI is mentioned, “AI-powered” can reduce message effectiveness unless it is paired with concrete, user-relevant outcomes like accuracy, speed, or efficiency. - Build transparency into the story, not just safety
CARMA’s split between audience and media trust drivers implies reassurance alone will not be enough. Brands should be prepared to explain decision pathways, controls, and limits in plain language. - Plan for accountability expectations that point upstream
The public placing responsibility on AI companies and developers means partners and platforms matter to your reputation. Vendor choices and governance posture can become part of brand trust. - Assume skepticism toward corporate AI messaging is the default
With 59% believing companies exaggerate AI use, the burden of proof has flipped. Under-claiming and demonstrating may outperform big positioning statements. - Anchor credibility in independent expertise when possible
Since academics are the most trusted voices among audiences, credibility signals may come more from demonstrated rigor and expert validation than from executive advocacy.
The broader implication is that AI communication is moving from persuasion to proof. Not because audiences dislike AI, but because they now have enough exposure to judge it.
As AI becomes embedded in everyday products, “trust” stops being a sentiment and starts looking like a user experience metric. Brands that can show how systems behave, where they fail, and who owns the outcome will earn resilience in a more skeptical media environment.
Over time, the advantage will not go to the loudest AI storyteller. It will go to the organization whose AI story matches what people repeatedly experience.
