ChatGPT just told a potential customer that your business closed last year — and it's completely wrong. AI search engines like ChatGPT, Perplexity, Claude, and Google AI Overviews are rapidly becoming the default way people research products, services, and companies. But unlike a Google results page — which links to sources users can verify — AI-generated answers can conflate, misattribute, or outright invent information about your brand. For NZ businesses, that represents a reputation risk that most haven't even begun to address.
What AI hallucinations mean for your brand
In AI terms, a "hallucination" occurs when a language model generates information that sounds plausible but is factually incorrect. It might attribute a competitor's negative review to your business, claim you offer services you don't, state incorrect pricing, or — in extreme cases — fabricate entire narratives about your company. Because these outputs are delivered with the same confident, authoritative tone as accurate information, users rarely question them.
As OpenAI's own research documentation acknowledges, even advanced models can produce outputs that are "inaccurate, untruthful, and otherwise unreliable." For a Wellington electrician or an Auckland accounting firm, a single hallucinated detail in a widely used AI tool could mean lost leads, eroded trust, and competitors winning business that should have been yours — without you ever knowing it happened.
How AI search engines get things wrong about businesses
AI hallucinations about brands don't emerge from nowhere — they arise from specific, identifiable patterns in how these models synthesise information. Understanding these patterns is the first step toward protection:
- Source conflation — The model mixes details from multiple sources about different businesses in the same industry. Your landscaping company gets credited with a project completed by a competitor because both appeared in the same local business directory the model was trained on.
- Temporal confusion — The model references outdated information as if it were current. Your business changed its name, moved premises, or updated its service menu two years ago — but the AI model's training data cutoff means it still presents the old details as facts.
- Authority inversion — The model weights an unreliable source (an old forum post, a sarcastic review, a satirical article) over your official website because it lacks the contextual judgment to distinguish between them.
- Gap filling — When the model encounters a knowledge gap about your business, it generates plausible-sounding filler rather than admitting uncertainty — creating details about your services, pricing, or history that never existed.
According to reporting by Vox on AI accuracy challenges, the fundamental issue isn't that AI models occasionally get things wrong — it's that they're structurally incapable of knowing when they're wrong, making brand-side monitoring essential rather than optional.
The three-tier approach to protecting your brand in AI search
Brand protection in the age of generative search isn't a one-time fix — it's an ongoing discipline that operates across three interconnected layers:
- Tier 1 — Foundation: Structured, authoritative web presence. AI models prioritise content that is well-structured, semantically clear, and backed by consistent signals across multiple authoritative sources. This means your Google Business Profile must be complete and verified, your website needs clear schema markup (Organization, LocalBusiness, Service, FAQ), and your brand information must be consistent across every directory, social profile, and industry listing where you appear. Inconsistency creates ambiguity, and ambiguity creates space for hallucinations.
- Tier 2 — Amplification: Proactive content creation for AI citations. Generative engines cite sources that demonstrate expertise, authority, and trustworthiness — the same E-E-A-T principles that Google uses, adapted for AI consumption. Publishing detailed service pages, author-attributed blog content, case studies with verifiable data, and FAQ sections that directly answer common customer questions gives AI models high-quality material to cite. When your content is the best available source on a topic, the model is more likely to reference it accurately.
- Tier 3 — Monitoring: Systematic AI output surveillance. You need to regularly query major AI platforms about your brand, products, and key executives — and document what they say. This isn't vanity searching; it's reputation auditing. Changes in AI outputs — a new error, a shifted description, a suddenly missing service — are early warning signals that demand investigation.
How to monitor what AI says about your business
Practical AI brand monitoring doesn't require expensive enterprise tools — it requires discipline and a systematic approach. Start with a structured query list: your brand name, your brand name plus your city, your brand name plus each core service, your founder or key executive names, and your brand name plus competitor names (to catch conflation errors). Run these queries across ChatGPT, Perplexity, Claude, and Google (to check AI Overviews) on a monthly schedule.
Document results in a simple tracking spreadsheet: date, platform, query, output, and a flag column for any inaccuracies. Over time, this log becomes your evidence base — showing patterns, tracking corrections, and proving the business case for ongoing investment in AI brand management. Google's Search Console documentation provides additional tools for monitoring how your site appears in search, including AI-generated features, giving you a technical baseline to build from.
What to do when AI gets your brand wrong
Discovering a hallucination about your business is frustrating, but the response matters more than the discovery. Panic-publishing corrections or demanding takedowns (which aren't possible in the traditional sense for generated content) rarely helps. Instead, follow a structured correction workflow:
- Identify the root cause. Trace the likely source of the error. Did a directory listing have old information? Did a news article misreport something? Was there a review that the model misinterpreted? Fixing the source prevents recurrence.
- Flood the zone with accurate content. Publish content that directly addresses the specific claim the AI got wrong — not with a "correction" framing, but with substantive, authoritative information on the topic. If the AI claimed you don't service a particular area, publish a detailed location page. If it misstated your pricing, create transparent pricing content.
- Strengthen your structured data. Update your schema markup to explicitly state the facts the AI hallucinated about. If the model confused your founding date, add a foundingDate property to your Organization schema. If it mixed up your service area, reinforce your areaServed markup.
- Submit updated information to AI training data sources. Some platforms now offer mechanisms for businesses to submit factual information for incorporation into model training. While these processes are still evolving, participating early establishes your brand as proactive.
Why AI brand protection is an ongoing investment
This isn't a project with a completion date. AI models are continuously retrained, search interfaces evolve, and new platforms emerge. The brand that was accurately represented in ChatGPT last month might be misrepresented in Perplexity this month — and corrected again next quarter. The businesses treating AI reputation management as a recurring operational function, not a reactive cleanup exercise, are the ones building durable digital trust.
For NZ SMBs, this represents both a risk and an opportunity. The risk is that competitors who invest in AI visibility will capture the customers who ask AI search engines for recommendations — while businesses that ignore it become invisible or, worse, misrepresented. The opportunity is that the barrier to entry is still low: most NZ businesses haven't started, which means the ones who act now establish a presence in AI-generated answers that becomes progressively harder for latecomers to displace.
The bottom line
AI-generated search isn't a future scenario — it's the present reality for a growing segment of your potential customers. Every day, New Zealanders ask ChatGPT, Perplexity, and Google AI Overviews for business recommendations, service comparisons, and buying advice. If your brand appears in those answers — and appears accurately — you gain trust and traffic. If it appears incorrectly, you lose both, often without ever knowing why. The businesses that build systematic AI brand monitoring and protection into their digital strategy now won't just avoid reputation damage — they'll capture the competitive advantage of being the brand AI search engines cite correctly, consistently, and first.
At Disruptive, we help NZ businesses build comprehensive Generative Engine Optimisation strategies that ensure AI platforms represent your brand accurately and favourably — from structured data and content authority to ongoing monitoring and correction workflows. Explore our GEO services or get in touch to discuss how we can protect and amplify your brand in the age of AI search.