ChatGPT, Perplexity, and Gemini demonstrably output outdated or incorrect information about brands — old product names, outdated positioning, wrong attribution to competitors. This isn’t malicious; it comes from how the models themselves work: training data has a cutoff date, current sources are missing or contradict each other, and brands with similar names or categories get confused. For the affected brand, this is a reputational risk that most simply aren’t aware of.
Why AI systems say wrong things in the first place
Four mechanisms are most commonly responsible for this in practice:
Training data cutoff
Large language models are trained at a fixed point in time. What happens after that — rebranding, new products, acquisitions — the model often doesn’t know.
Missing current sources
Without current, clearly structured sources, AI systems fall back on older or indirect information.
Confusion with competitors
Similar names, categories, or positioning are occasionally mixed up by AI systems.
Lack of entity clarity
If a brand isn’t clearly machine-readable — inconsistent information across website, social media, and press — the error rate increases.
How does a language model technically “hallucinate”?
Language models like ChatGPT generate text by predicting the statistically most likely next word at each step, based on patterns from huge amounts of text — not by looking up what’s “true” in a database. There’s no built-in mechanism that checks every statement against a reliable source before it’s output. RAG-based systems (Retrieval-Augmented Generation) do additionally look up current sources, but that only works as well as the available, findable sources themselves. If a clear, current source is missing, the model fills the gap with the most plausible pattern from its training data — which can be factually wrong without that being detectable from the wording.
An example
If you ask an AI something like “What products does [brand] offer?”, an unremarkable-sounding but outdated answer can come out of it:
Such errors rarely stand out immediately — until a customer asks why the product is named differently than expected.
Cause and countermeasure
Whoever doesn’t monitor what AI says about their own brand usually finds out about misinformation first from an uneasy customer.
Why this is a reputational risk
Unlike a false statement on your own website, an AI answer can’t simply be corrected — it’s generated anew with every query, based on what the model has “learned” and which sources it currently draws on. That makes the risk diffuse: nobody gets notified when an AI says something wrong about their own brand. Without active monitoring, this often goes unnoticed for months.
When does this become dangerous?
Not every small inaccuracy is business-critical. It becomes critical when misinformation shows up at moments that directly influence a decision: during preliminary research ahead of a tender, during due diligence by investors or partners, or when potential customers ask an AI to compare your brand with a competitor. In these moments, nobody has the time or reason to additionally verify the AI’s answer — it’s taken at face value.
How to roughly check this yourself
A first, free test takes only a few minutes: open an incognito window and ask ChatGPT, Perplexity, and Gemini the same one or two questions in sequence that a potential customer would realistically ask — for example, about your core products, your positioning, or a comparison with your closest competitor. Watch for three things: do the stated facts (names, numbers, offerings) match what’s currently on your website? Is your brand associated with the right category and competitive set? And does the description read like your current positioning, or like an older state? A single test doesn’t replace systematic monitoring, but it often already shows whether action is needed in principle.
What you can do about it
Three steps help concretely. First, measure what’s currently being said about the brand — that’s what the AI Visibility Audit does. Second, keep messaging, product pages, and entity signals consistent, for example as part of a Sharp Positioning project. Third, continuously watch for new misinformation creeping in — that’s what Measurable AI Visibility is for. More on the underlying basics in What Is AI Visibility? and GEO vs. SEO: What’s Changing.
Frequently Asked Questions
Why do AI systems output incorrect information about brands?
Usually for three reasons: the models’ training data cutoff lies in the past, current or clearly structured sources are missing, or the brand is confused with a competitor because entity signals are inconsistent.
Can I get incorrect AI statements about my brand corrected?
Not directly like a Wikipedia entry, but indirectly: through current, clearly structured, and consistent content on your own website and in third-party sources, which AI systems prefer as more current, more citable information.
How often should I check what AI says about my brand?
Monthly monitoring is sufficient for most brands to catch new misinformation or changes early, rather than learning about them by chance from customers.
Is the risk the same for small brands as for large ones?
If anything, it tends to be higher: large brands have more current sources and press coverage for AI systems to draw on. Smaller brands with fewer sources are more often confused with similarly named competitors.
