6 min read
You ask ChatGPT to write a short blog post about caring for leather boots for your shoe repair shop. It comes back polished and confident, and it includes a tidy statistic: “According to a 2023 industry study, 68% of boot damage is caused by improper drying.” It sounds great. You almost hit publish. Then a small voice asks: which study?
There is a good chance that study does not exist. When an AI tool states something that sounds true but is made up, that is called an AI hallucination, and learning to spot one is one of the most valuable skills a small business owner can build. The good news is that it does not require any technical knowledge. This beginner’s guide explains what hallucinations are, why they happen, where they cause real trouble, and a simple five-minute habit to catch them.
What an AI hallucination actually is
An AI hallucination is an answer that is presented as fact but is false or invented. It can be a fake statistic, a quote nobody said, a book that was never written, a law that does not say what the AI claims, or a product feature that does not exist. The tricky part is that hallucinations look exactly like correct answers. Same confident tone, same clean formatting.
The term is widely used across the industry; the Wikipedia entry on AI hallucination is a good plain overview if you want the background. You do not need to memorize the definition. You only need to remember this: an AI answer can be fluent and wrong at the same time.
Why AI tools make things up
Chat tools like ChatGPT, Claude, and Gemini run on what are called large language models, which is a technical name for software trained on huge amounts of text to predict what words should come next. They are extremely good at producing text that sounds right. They do not look facts up in a reliable database unless they are specifically searching the web or reading a document you gave them.
OpenAI has explained part of the reason in its write-up on why language models hallucinate: the way these models are trained and tested tends to reward a confident guess over an honest “I don’t know.” So when the model does not have a solid answer, it often fills the gap with something plausible.
Newer models hallucinate less than older ones, and tools that search the web or work from your own documents are more grounded. But no current AI tool is hallucination free, and that will likely stay true for a while.
Where hallucinations hurt a small business most
A made-up fact in a brainstorm costs you nothing. A made-up fact that reaches a customer can cost you money and trust. These are the danger zones:
- Numbers and statistics in blog posts, ads, and sales pages.
- Policies and promises in customer replies, such as refund terms, delivery times, or warranty coverage.
- Legal, tax, and licensing questions, where a confident wrong answer can lead you to skip a real requirement.
- Product and tool details, like pricing, features, or compatibility, especially for anything released recently.
- Names, quotes, and sources, including fake citations that look like real articles.
There is a well-known real example of the second category. In 2024, a Canadian tribunal ruled against Air Canada after its website chatbot gave a traveler incorrect information about bereavement fares, and the airline was held responsible for what its chatbot said. The BBC’s coverage of the case is worth a read. The lesson for a solo owner is simple: if your AI says it to a customer, you said it.
The five-minute fact check habit
You do not need to distrust AI. You just need a routine that catches the risky parts before they leave your hands. Here is one that takes about five minutes per piece of content.
Step 1: Highlight every “checkable” claim
Read the AI’s draft once and mark anything that is a number, a date, a name, a quote, a price, a rule, or a “studies show.” Everything else (tone, structure, general advice) is low risk. Usually only a handful of sentences need checking.
Step 2: Ask the AI where each claim came from
Ask: “For each statistic and factual claim in your answer, give me the source and a link. If you are not sure of a source, say so.” A good tool will often admit it cannot point to one. That admission is your signal to cut or replace the claim.
Step 3: Open the link yourself
Even when the AI provides a link, click it. Confirm the page exists and that it actually says what the AI claims. Fake or mismatched citations are one of the most common hallucinations.
Step 4: Replace, soften, or delete
If you can find a real source, use it. If you cannot, either soften the claim into something you know is true (“many boot owners dry them too fast”) or delete it. A vaguer true sentence beats a precise false one every time.
Step 5: Keep your own facts handy
Most hallucinations in customer replies happen because the AI was guessing about your business. Write a one-page “facts sheet” with your prices, policies, hours, and service area, and paste it in (or save it as instructions in your AI tool) whenever you draft customer-facing text. When the AI has your facts in front of it, it has much less reason to invent them.
Simple ways to get fewer hallucinations in the first place
- Give it the source material. Ask the AI to work from a document, web page, or notes you provide, rather than from memory.
- Tell it that “I don’t know” is allowed. Add a line like “If you are not certain, say so instead of guessing.”
- Use search-enabled modes for recent facts. For anything about prices, new products, or recent news, use a mode that searches the web, and still check the links.
- Keep humans on the final step. AI drafts, you approve. This matters most for reviews, policies, and anything translated, as we explained in our guide to never publishing AI translation unreviewed.
Your first step today
Pick one piece of AI-written content you have already published or sent: a blog post, a product description, or a saved customer reply. Run the five-minute check on it. Most owners find at least one claim they cannot back up, and fixing it now is far cheaper than explaining it to a customer later.
Then make it routine. Hallucinations are not a reason to stop using AI; they are a reason to use it the way you would use a fast, eager new assistant: grateful for the draft, careful with the facts. Have you caught an AI tool making something up? Share what it invented in the comments; it will help other readers know what to watch for.
Related reading
- AI Translation Is Good Enough for Your Website’s First Draft. Never Publish It Unreviewed
- How to Answer Every Google Review in Five Minutes a Week With AI (Without Sounding Like a Robot)
- Can Clients Catch Your ChatGPT Writing Now? 5 Myths About OpenAI’s New Text Watermark
- What Do the New State AI Laws Mean for a Solo Business? A Plain-English Starter Guide
