When AI Gets It Wrong, Who Corrects the Record?

Type a question into a search engine today, and there’s a good chance the answer arrives before the results do. You’ll see a synthesized summary, assembled by an AI model, sitting above or next to links to several websites. For an industry that has spent 70 years building its credibility on precise, verifiable facts, that shift raises an uncomfortable question: When an AI system gets something wrong about your company, who corrects it?

The honest answer, right now, is no one. AI regenerates its answer fresh each time someone asks, drawing on whatever it can find. If the record it finds is thin, outdated or contradictory, that becomes the story it tells to anyone who asks.

That risk is no longer theoretical. New research from the communications firm Burson, conducted in partnership with the analytics firm Profound, fielded thousands of reputation-related questions across seven major AI models and scored how 85 companies across 10 industries were portrayed, then tested how believable those AI-generated answers were to the general public, opinion elites and business decision-makers, producing more than 55,000 believability forecasts.

One finding stands out the most: business decision-makers, the people negotiating partnerships, funding and contracts, rated AI-generated company summaries as 10 percent more believable, on average, than the general public did.

Aerospace’s advantage and its blind spot

Aerospace comes into this moment better positioned than most industries Burson studied because claims that AI models could verify against hard evidence – including certification, safety data, performance results and order books – were consistently rated far more believable than claims about leadership or values, which are harder for any system to check. Aerospace has spent decades generating exactly the kind of documented, third party-validated proof, and it showed in the results of Burson’s report: the industry scored strongly on innovation, financial performance and workplace reputation, and even held up relatively well on leadership (typically one of the hardest categories for any industry to win with AI audiences).

That last point is also the warning. Leadership reputation ranked near the bottom across every industry Burson tested, and while aerospace outperformed most peers, it was not exempt. Executive visibility on its own didn’t move the needle; what helped was leadership credibility that could be traced back to governance and performance records rather than personality-driven messaging. In other words, aerospace’s current AI credibility looks less like something companies actively built for this moment and more like something they inherited from decades of documentation and verification. That inheritance won’t spend itself automatically in a company’s favor going forward.

That advantage is also more fragile than it looks because reliance upon AI summaries is rising faster than their accuracy. A BBC report last year found more than a third of UK adults said they completely trust AI to produce accurate summaries (and nearly half of those under 35).Yet, a separate BBC and European Broadcasting Union review of more than 3,000 AI-generated answers found 45 percent contained a significant issue, from wrong facts to fabricated sources. More people are taking these summaries at face value faster than the accuracy rate can support, and aerospace’s strong current standing inside that system isn’t the same as being protected from it.

Building the record before you need it

Maintaining an accurate reputation in an AI-mediated world requires demonstrating it, documenting it and validating it through third-parties on an ongoing basis—all proactively, before anyone questions it. For most companies, that starts with treating the public record (certification, safety milestones, financial disclosures, technical achievements) as something that needs to be current, consistent and easy for a model to find, rather than something scattered across old press releases and buried PDF archives. A record that’s accurate but stale is not as attractive as fresh content to a system that’s re-synthesizing an answer every time someone asks.

On leadership specifically, the fix isn’t more executive profile pieces. It’s tying executive visibility to the same evidence discipline the rest of the industry already practices, including statements and interviews that point back to governance decisions, safety outcomes and performance data, not aspiration. And because AI systems increasingly draw stakeholder impressions from the same well of public content, companies are better served treating earned media coverage, owned content and social media channels as one connected evidence ecosystem versus three separate workstreams competing for the communications and public relations budget.

Rather than making a claim, aerospace companies demonstrate, document and independently verify and certify their work. AI works the same way. If there isn’t a consistent, regularly updated and well-documented public record of the facts people are searching for, AI will draw its own conclusions, and they won’t always align with a company’s narrative.

Being surfaced by an AI model isn’t the same as being represented accurately by one, and there’s no mechanism to fix a bad answer after the fact. The best approach is to build the evidence trail before the question gets asked. Aerospace already knows how to build that kind of trail; it does so every day. The industry best positioned to win the AI reputation contest is the one that already lives by verifiable proof, as long as it treats its reputation with the same type of discipline it brings to its engineering.

—Steve Rubel, EVP of Media Insights and Measurement, Burson