Security firm Eset says AI assistants such as ChatGPT are directing users to fake online stores, cryptocurrency scams, imitation login pages and malicious downloads, according to research reported by Inside Retail Australia on 16 September 2026. If you run a business, the uncomfortable part is not that scammers are gaming AI search. It is that the same mechanism they are gaming is the one that decides whether an AI assistant recommends you, a competitor, or a clone of your website pretending to be you.

What the research found

Eset’s analysis of web traffic and AI search outputs found that AI assistants, which summarise large volumes of web content rather than ranking pages the way a search engine does, sometimes pull from low-quality or compromised sources. Scammers exploit that by publishing keyword-optimised, AI-generated content designed to be picked up when the assistant goes looking for an answer.

Scott Leman, Eset’s New Zealand country manager, described the risk as one of misplaced trust: someone asks an AI tool for help, follows the link it gives them, and ends up on a fake login page, a scam website or a malicious download. “That misplaced trust creates a significant entry point for malicious actors,” he said. He also warned that the breathing space businesses used to have between a weakness appearing and it being exploited “is disappearing” as attacks accelerate.

This is not a one-off. In June, TechSpot reported that the scam-checking service Ask Silver had found ChatGPT shopping results linking to cloned versions of UK retailers Russell & Bromley and Dunelm, on look-alike domains. Ask Silver’s Anna Jones said scammers appeared to be exploiting the gap left when Russell & Bromley went into administration in January 2026. Louise Baxter of National Trading Standards put it simply: “Consumers are increasingly turning to AI tools for advice, but criminals are adapting quickly.”

Why this matters even if you have never been cloned

Traditional search built a reputation system over twenty years: domain age, links, user behaviour, and a long list of signals that made it expensive to fake authority. AI assistants are younger, and they answer differently. They retrieve a set of pages that look relevant, read them, and compose an answer. If the pages they retrieve are wrong, thin or fraudulent, the answer is too, and it is delivered in a confident voice that most people do not question.

For an Australian business that means three separate exposures:

1. Impersonation. A clone of your site, or a fake page using your brand name, can be surfaced to a customer who asked the assistant where to buy from you.
2. Misdescription. The assistant describes your business using whatever it found: an outdated directory listing, an old price, a review site, a competitor’s comparison page. If your own site does not state the facts plainly, something else will.
3. Absence. If your site is not retrievable, not clear, or not trusted, you simply do not appear in the answer, and the customer never knows you existed.

The scam story is the extreme end of a spectrum every business sits on. The fix for all three exposures is the same work.

Step one: find out what the assistants say about you

Before changing anything, measure. Ask ChatGPT, Perplexity, Google’s AI mode, Gemini and Copilot the questions a customer would ask: “best [your service] in [your city]”, “is [your brand] legitimate”, “[your brand] pricing”, “where do I buy [your product] in Australia”. Record the answers, the sources each one cites, and every link it gives. Do it monthly. You are looking for wrong facts, wrong links, competitors named ahead of you, and any domain you do not own using your name.

This is the AI equivalent of googling yourself, and most businesses have never done it.

Step two: make your own site the easiest source to trust

AI assistants are looking for the same things a careful human would: clear statements of what you do, who you serve, where you are, what it costs, and evidence that you are real. Practically:

• State the facts in plain text on your own pages. Services, locations, pricing (or how pricing works), contact details, ABN, the people behind the business. Text the assistant can read beats text hidden in images or buried in a PDF.
• Use structured data. Organisation, LocalBusiness, Product, Service and FAQ schema tell machines exactly what your pages mean. It is cheap to add and it removes ambiguity.
• Answer real questions on pages that exist to answer them. The queries people put to assistants are questions. A page that answers one well, with a clear heading and a direct first paragraph, is what gets retrieved.
• Let the crawlers in. Several AI assistants use their own crawlers, and some hosting and CDN providers block them by default. If your site returns an error to those crawlers, you have opted out of being recommended. Check this; it is a common, silent problem.
• Keep your listings consistent. Google Business Profile, Bing Places, industry directories and your social profiles should agree with your website on name, address, phone and services. Inconsistency reads as untrustworthy to a machine.

Step three: defend the brand

The impersonation risk deserves its own short list:

• Register the obvious look-alike domains for your brand, and set up alerts for new registrations that contain your name.
• Search for your brand name plus “official” and “login” across the assistants and search engines regularly.
• If you find a clone, report it to the assistant (ChatGPT has a reporting form), to the domain registrar, and to Scamwatch, and tell your customers through your own channels.
• Put a clear “how to know you are on our real site” note on your contact page. It helps customers and it gives the assistants a source to cite.

Step four: treat AI search as a channel, not a curiosity

The businesses that will be recommended by AI assistants in 2027 are the ones that treat this the way they treated search in 2010: as a measurable channel with its own content, technical requirements and monitoring. That means a baseline of what the assistants currently say, a plan to fix the gaps, and a monthly check that the answers are improving.

Eset’s warning is really a warning about trust: people believe what the assistant tells them. That is a threat when the assistant is wrong, and an enormous advantage when it is right about you.

How Digital One Agency helps

Digital One Agency’s AI search visibility work starts with exactly the audit described above: what ChatGPT, Perplexity, Claude, Gemini and Google’s AI mode say about your business today, which sources they rely on, and where you are absent or misdescribed. From there we fix the technical blockers, restructure your key pages so they answer the questions customers actually ask, add the structured data, and set up ongoing monitoring so you can see the answers change. If you want to know where you stand before committing to anything, the free AI audit is the place to start.

AI assistants are already answering questions about your business. Book a meeting with Digital One Agency and make sure they get the answer right.

Related services

Digital One Agency delivers AI development solutions, AI integration services for digital transformation, private AI for data that must stay under your control, and secure AI system development, backed by AI consulting services that produce a costed plan in 30 days.

Frequently asked questions

Is ChatGPT really sending people to scam websites?

According to research by security firm Eset reported by Inside Retail on 16 September 2026, AI assistants including ChatGPT have directed users to fake online stores, cryptocurrency scams, imitation login pages and malicious downloads. In June, TechSpot reported ChatGPT shopping results linking to cloned versions of two UK retailers on look-alike domains.

How do AI assistants decide which businesses to recommend?

They retrieve a set of web pages that look relevant to the question, read them, and compose an answer. Clear, factual, well-structured pages on your own site, consistent listings elsewhere, and crawler access all make it more likely the assistant retrieves and trusts your site rather than a directory, a competitor or a clone.

How can I check what ChatGPT says about my business?

Ask the assistants the questions a customer would ask: best provider in your city, whether your brand is legitimate, your pricing, where to buy from you. Record the answers, the sources cited and every link given, and repeat monthly. Look for wrong facts, wrong links, competitors named ahead of you and any domain using your name that you do not own.

Why might my website be invisible to AI search?

The most common silent cause is that your hosting or CDN blocks the crawlers AI assistants use, so your site returns an error to them. Other causes are key facts hidden in images or PDFs, no structured data, and pages that do not directly answer the questions people ask.

What should I do if I find a fake site using my brand?

Report it to the assistant that surfaced it, to the domain registrar and to Scamwatch, tell your customers through your own channels, and publish a clear note on your site explaining how to recognise your genuine website. Registering obvious look-alike domains in advance reduces the risk.