In one sentence
An AI hallucination is when an AI assistant states something false as fact, like a wrong price, a discontinued product or a feature you don't offer, and shoppers may believe it.
Why AI assistants make things up
A large language model (LLM) writes the most likely next words, not verified facts. When it has thin, old or conflicting information about your brand, it fills the gap with something that sounds right.
The risk goes up when your own site is hard to read, your details differ from page to page, or the model is working from training data that predates your latest launch.
Common AI hallucinations about ecommerce brands
- Wrong prices: an old sale price, or a price from another retailer.
- Dead products: a model you discontinued, recommended as current.
- Made-up features: a waterproof rating, material or size you don't sell.
- Wrong policies: return windows, shipping costs or warranties that aren't yours.
- Mixed-up brands: your brand confused with one that has a similar name.
Each one costs you twice: the shopper who believes it, and the one who finds out it's wrong and blames you.
How to fix wrong information AI says about your brand
- Ask the assistants direct questions about your brand and products, and write down every error.
- Fix the source: make the correct fact clear and consistent on your product pages, policy pages and FAQ.
- Add structured data so price, stock and specs are machine-readable.
- Correct third-party pages that repeat the error, like old reviews, marketplace listings or directory profiles.
- Check again. Assistants that search live can update fast. Answers from training data change only when the model does.
AI hallucination FAQs
Can I report a wrong AI answer?
Yes. ChatGPT, Claude, Gemini and Perplexity all let users flag a bad answer. It helps, but it's slow and not guaranteed. Fixing your own pages works faster.
How do I find out if AI is getting my brand wrong?
Ask it. Try branded questions like "What's the return policy at [your store]?" and "Is [your product] waterproof?" across all four assistants, and check each claim against your site.
Will fixing my site fix the AI answer right away?
Not always. Assistants that search the web live can pick up the fix once your page is recrawled. Facts baked into training data stay until the next model update.
About the author
Jenn Starr, Co-founder, Selix
Jenn has spent 20+ years helping SaaS startups build their marketing and implementation playbooks. She has bootstrapped three marketing agencies and is a 500 Startups alum. At Selix, she leads the work of getting ecommerce brands named in AI answers.
Related terms