It started with a pizza tip.
In 2024, someone asked Google’s new AI Overview feature how to keep cheese from sliding off pizza. The answer, served up to millions of users at the top of the search results page, was confident and specific: add non-toxic glue to the sauce. The AI had lifted the tip from a years-old joke post on Reddit and delivered it as fact. Screenshots went viral. Late-night hosts had a field day. Google quietly patched the feature and moved on.
Google survived it. You might not be so lucky.
AI tools have become genuinely useful for small business owners. Writing faster, responding to customers, generating ideas, keeping up with the content calendar. Most of the time, they work. But “most of the time” is not good enough when your brand name is attached to the output. When AI gets it wrong, it is your reputation on the line, not the software company’s.
Here is what that looks like in practice, and what to do about it.
The Chatbot That Threw Its Own Company Under the Bus
In early 2024, DPD (one of the UK’s largest parcel delivery companies) deployed an AI chatbot to handle customer service inquiries. A customer, apparently curious about what the bot would do, asked it to write a poem about the company. The bot obliged. The poem described DPD as “the worst delivery firm in the world.” It also swore at the customer.
The exchange was screenshotted and posted online. It reached over 800,000 views in less than 24 hours. DPD disabled the feature the same day and blamed a system update.
Now imagine that is your local service business. Your landscaping company. Your salon. Your gym. The bot you set up to handle after-hours inquiries does not know your voice, does not understand context, and can be nudged in directions you never anticipated. One screenshot shared in a neighborhood Facebook group can do real damage to a business that took years to build.
The AI That Made Promises Your Business Cannot Keep
Air Canada’s AI chatbot once promised a passenger a bereavement discount that did not exist in company policy. When the passenger claimed the refund, Air Canada refused, arguing the bot was a “separate legal entity” not responsible for what it said. A court disagreed. The airline was held accountable for its chatbot’s mistake.
For small business owners, this case matters for a simple reason: your customers do not distinguish between you and the tool you are using. If your AI scheduling assistant tells a client their appointment is confirmed when it is not actually booked, or your AI customer service tool promises a discount you are not running, you own that mistake. Not the software company.
Klarna, the buy-now-pay-later company, went all-in on AI customer service, and then quietly walked it back. The bots handled volume fine. But complex issues, frustrated customers, and nuanced conversations? Those still needed humans. Klarna returned to a hybrid model. The lesson was not that AI does not work, it was that AI alone is not enough.
When the Content Looks Fine, But Actually Isn’t
Not all AI failures are dramatic. Some are quiet, and those are the ones to watch.
In May 2025, the Chicago Sun-Times published an AI-generated summer reading list that recommended books that do not exist. Complete titles, real authors’ names, invented content. No one checked. A librarian caught it and called it out publicly. The paper apologized for “insufficient editorial oversight.” The damage to credibility was real and immediate.
AI confidently invents. That is not a bug, it is how the technology works. Large language models predict the most plausible next word, not the most accurate one. Which means a blog post, product description, or social caption can read perfectly well and still contain a fabricated statistic, a wrong date, or a claim that cannot be verified because it was never true.
For small businesses, this is the sneaky one. A misspelled word is easy to catch. A confident-sounding wrong fact that slips through your review? That is what a competitor screenshots.
The Voice Problem
Coca-Cola has spent over a century building one of the most emotionally resonant brands in history. In late 2024, the company released a version of its iconic “Holidays Are Coming” Christmas ad produced with generative AI tools. Same trucks, same snow, same music. Viewers called it “soulless.” Some said it looked like a deepfake of a memory.
The content was not wrong. It was not offensive. It just felt hollow. And that hollowness is something audiences noticed immediately, even if they could not express exactly why.
Your brand voice is one of your most valuable assets. It is the reason a longtime customer recognizes your email before they see your name in the sender line. AI is very good at producing content that sounds like content. It is less good at sounding like you. Unedited AI output tends to flatten brand personality into a kind of generic competence. Helpful, clear, and entirely forgettable.
What This Actually Means for Your Business
None of this means you should avoid AI. It means you should use it the way a smart business owner uses any tool, knowing what it is good for and what it is not.
- Treat AI output as a first draft, not a finished product. Every piece of AI-generated content that carries your name should pass through a human editor before it goes live. That editor does not have to rewrite it. They just have to read it with fresh eyes and ask: is this accurate? Does this sound like us? Would I be comfortable if a competitor read this?
- Never let AI speak to your customers without guardrails. If you are using a chatbot or automated responder, define exactly what it can and cannot say. Set hard limits. Redirect anything outside its lane to a human. The DPD story is funny when it happens to someone else.
- Fact-check anything with a number in it. Statistics, dates, product specs, pricing. AI will invent these with full confidence. Any claim that could be verified (or disproven) by a reader should be independently confirmed before you publish it.
- Edit for voice, not just accuracy. Read the AI draft out loud. Does it sound like you? Does it have the warmth, the specificity, the personality that your customers recognize? If it reads like a press release from a company you have never heard of, it needs another pass.
- Use AI where it helps most. Drafting emails, brainstorming captions, outlining blog posts, answering FAQs. These are tasks where AI saves real time and the stakes of a minor error are low. Customer-facing content with factual claims, customer service conversations, and anything that represents a legal or financial commitment to your customer? Those need more oversight, not less.
Key Takeaways
- AI tools are genuinely useful, but they make confident mistakes, and your brand absorbs the consequences.
- The most common failures are hallucinated facts, off-brand tone, and customer-facing chatbots that go sideways.
- Courts have ruled that businesses are responsible for what their AI tools say to customers.
- The fix is not less AI, it is better oversight. Human review, clear guardrails, and a consistent editorial voice.
- AI is the assistant. You are still the brand.
FAQs About AI Usage for Small Businesses
Is it okay to use AI to write my business’s social media posts?
Yes, but only with editing. AI can draft captions quickly, but unedited AI content often sounds generic and can miss the specific voice your audience expects from you. Use it to get words on the page, then make it sound like you before hitting post.
What is the biggest risk of using AI for customer communication?
Making commitments you cannot keep. AI chatbots have promised refunds, discounts, and policies that did not exist, and courts have held businesses accountable for those promises. Any AI that interacts directly with customers needs clear boundaries and human backup for anything beyond simple FAQs.
How do I know if an AI tool is right for my business?
Start with low-stakes tasks such as internal brainstorming, first drafts of emails, and generating headline options. See how much editing the output actually needs before it sounds like you. If it is saving time without creating more work downstream, expand from there. If every AI draft needs a full rewrite, you are better off writing it yourself.
The Bottom Line
The businesses that get AI right are not the ones using it the most. They are the ones using it the most thoughtfully. AI is a powerful first draft machine. It is not a brand manager, a fact-checker, or a replacement for the judgment that comes from actually knowing your customers.
Keep a human in the loop. Read everything before it goes out. And remember: the goal is not to produce more content faster. It is to stay connected to the people who trust your brand.
Need help building an AI-assisted content strategy that still sounds like you? Let’s talk.
