A few months ago, we explored an idea in these very pages that has since become impossible to ignore: digital searches are becoming conversations. The habit we’ve repeated almost without thinking for more than two decades—typing a question into Google and scrolling through a list of results—is rapidly disappearing.
Consumers are no longer just searching for information. They’re turning to artificial intelligence for direct answers. And this isn’t simply a hunch. According to Forrester’s 2026 State of Business Buying Report, 94% of B2B buyers already use generative AI tools during the purchasing process. In other words, due diligence no longer begins on your website or with a Google search. It begins in a conversation with an AI platform you don’t control.
Once we accept that search has become a conversation, a crucial strategic question emerges—one every marketing and communications director should be asking: Who is shaping that conversation?
Put another way, when a customer, investor, or analyst asks ChatGPT, Gemini, or Perplexity about your company, what will the AI say? How will it describe your business? Will the information be accurate? And if someone asks about the leaders in your industry, will your competitors get the spotlight instead?
We’re entering a new era in which brand reputation is no longer measured solely by media coverage, sentiment across online communities and social networks, or traffic to your corporate website. It also depends on how effectively you help AI models understand your business and represent your brand. These large language models (LLMs) can now shape perceptions of your company in a matter of seconds.
Identity hijacking, outdated narratives, and hallucinations: How AI distorts your brand
For the past decade, marketing professionals have focused on driving traffic to their websites. SEO, SEM, email marketing—many of our tactics have shared the same objective: getting users to click a link and read our side of the story on a platform we control.
That control has fundamentally changed. LLMs don’t form opinions in the way people do. They generate responses based on patterns learned from training data and information available to them. When a brand’s digital footprint is unclear, outdated, or difficult for AI systems to interpret, three types of reputational distortion can emerge. We’re already seeing all three in the market.
- Identity hijacking: This happens when AI systems struggle to access or interpret a company’s official information. Without clear, structured information from authoritative sources, they may turn to social media discussions, online forums, and third-party review sites to fill the gaps. This is particularly problematic for consumer-facing businesses (B2C), where AI models may combine financial information with Reddit complaints or reviews from third-party platforms, creating a misleading picture of the company. In our recent IBEX 35 study, for example, we found that when asked certain questions about Unicaja Banco, AI responses prioritized references to phishing, online scams, and court rulings, overshadowing the core attributes of its banking business.
- Conceptual inertia: AI models can struggle to keep up with how a company evolves. When a large organisation changes its business model or strategic focus but fails to communicate that shift clearly in the digital channels AI relies on, it risks becoming trapped in an outdated narrative. The model continues to associate the brand with what it did a decade ago, rather than what it does today.We’ve seen this with Logista, where several AI models continue to overlook its strategic expansion into pharmaceutical logistics, instead reducing the company to its traditional association with tobacco distribution.
- Reputational hallucinations: When information is ambiguous or reliable corporate sources are lacking, LLMs may fill in the gaps with inaccurate claims. These errors can extend to a company’s core business activities and capabilities. Rovi is one example. Some models have confused the company’s role as a pharmaceutical manufacturer with that of a researcher. In some responses analyzed in our study, they even incorrectly attributed ownership of the Moderna COVID-19 vaccine patent to Rovi.
To understand the scale of these challenges, TEAM LEWIS Spain analyzed how major publicly listed companies perform in our IBEX 35 AI SCORE study. The findings point to a clear problem: being invisible to AI is a real reputational risk. If LLMs can’t find reliable information about your company—or misinterpret what they find—you risk being overlooked by the very people you need to reach.
Why PR matters more than ever in the age of AI
Given this landscape of clickless searches, many communications professionals might think that traditional public relations and media outreach have lost their strategic value. The reality is quite the opposite. The great paradox of generative artificial intelligence is that it has transformed corporate PR into one of the most important drivers of brand reputation.
To reduce errors and deliver credible answers, language models tend to favor sources they recognize as authoritative, reliable, and well established. This tendency can be described as a “prestige bias”: information from reputable, verifiable, and frequently cited sources may carry greater weight in shaping how AI describes a company.
When a respected media outlet publishes information about your business, that coverage can help reinforce the credibility of the information AI retrieves. Financial publications such as Expansión and Cinco Días, alongside international news agencies such as Reuters, can serve as important reference points for LLMs.
Scientific evidence and regulatory rigor can also help protect a brand’s reputation. Despite the reputational impact of the Gotham case, Grifols remains in second place in our study’s ranking. Its position as a global leader in plasma-derived medicines and blood products is supported by a substantial body of technical, scientific, and regulatory information. This provides AI models with a stronger foundation for accurately describing the company and its activities.
When online forums shape your reputation
But LLMs don’t rely on business news alone. This creates a particular challenge for B2C companies: when authoritative corporate information is scarce or a strong media narrative is missing, AI models may turn to user-generated content to fill the gaps.
In determining which sources to draw on, AI systems regularly encounter professional platforms such as LinkedIn and video-sharing platforms such as YouTube. But some of the biggest reputational risks can arise from online communities and discussion forums, including Reddit and Forocoches (one of Spain’s best-known online forums).
For a company with millions of customers—a bank, insurer, or telecommunications provider, for example—complaints, reviews, and negative comments across online forums can become signals that AI incorporates into its responses.
If a model can’t find enough reliable information in its training data or on the web, it may give disproportionate weight to the discussions it can access. The result? Official financial statements and corporate announcements can end up alongside complaints from dissatisfied customers, producing an account of the company that overemphasizes negative experiences and distorts its overall reputation.
From digital showcase to reliable source: How your website needs to evolve
Brands can no longer leave control of their narratives entirely in the hands of third parties. They need to rethink their owned digital channels, particularly because many corporate websites remain difficult for LLMs to interpret.
The solution has two parts: how you communicate and how you structure your information.
First, the way you communicate. Compelling brand storytelling is still important, but it isn’t enough on its own. AI models also need clear statements, verifiable facts, and information that makes relationships between people, products, services, and organisations easy to understand.
LLMs can benefit from question-and-answer formats that address real user queries directly, supported by concrete details such as figures, percentages, and dates. These specifics help establish context and make claims easier to verify. This doesn’t mean overhauling your entire website. It means identifying the pages and sections that matter most and making sure they communicate the right information clearly.
Second, the technical foundations. Your website needs to do more than look good. It must also make your information accessible and understandable to AI systems.
That means paying close attention to your company’s presence on reference platforms such as Wikidata and Wikipedia, where appropriate, and ensuring that your digital content uses structured data standards such as Schema.org and JSON-LD. These help machines identify and interpret information about your organisation, its products, and its activities, reducing the risk of confusion and inaccurate responses.
The benefits of getting this right are tangible. Our study found that Acciona Energía, Bankinter, CaixaBank, and Mapfre lead the proprietary channels category, in part because their websites are structured to help AI models identify and cite official documents, including PDFs, financial reports, and ESG reports. This gives models access to reliable primary sources rather than leaving them to rely on potentially misleading third-party discussions.
The conversation between brands and millions of users is already taking place in this new generative environment. The question is no longer whether your company is part of that conversation, but who is shaping its narrative: your corporate channels, the media, or the noise coming from third parties?
Brands that act now—strengthening media relationships, structuring their data, and publishing clear, verifiable information that AI can readily interpret and cite—will be better positioned to protect their generative reputation. Those that don’t risk leaving AI to fill in the gaps, potentially with a story they would never tell themselves.
Originally published in Spanish by MarketingDirecto.com.
