AI learns from huge amounts of human writing. That writing includes people's assumptions and unfairness, and AI can repeat them. This guide covers how bias shows up in small-business work and how to check for it. It also covers a related honesty rule: never use AI to fake reviews or make claims you can't back up.
How bias shows up
"Bias" here means AI treating people or groups unfairly, often in ways that are hard to see. Some examples from small-business work:
- Hiring. Asked to write a job ad, AI may use wording that quietly discourages some people from applying, like "young, energetic team." Asked to rank resumes, it may favor certain names, schools or work histories for reasons that have nothing to do with the job.
- Tenant screening. A summary of applicants might lean on details unrelated to whether someone will pay rent and care for the property. Housing decisions are covered by fair housing laws.
- Tone toward customers. AI drafts may sound warmer to some customers than others, or make assumptions about who someone is based on their name, neighborhood or writing style.
- Images and examples. Ask for "a picture of a contractor" or "an example nurse" and you may get the same stereotype every time.
How to check for it
You don't need special software. You need a few habits:
- Swap the details. Run the same prompt with a different name, age or neighborhood. If the answer changes in ways it shouldn't, that's bias.
- Read for assumptions. Does the draft assume someone's gender, age, family situation or background?
- Stick to written criteria. For hiring and tenants, decide your criteria first, in writing. Apply them the same way to everyone, and don't let AI invent new ones.
- Keep a human decision. People make decisions about people.
- When in doubt, ask. Employment and housing have specific laws. Check with a qualified professional.
Can this go into AI? A salon owner wants AI to rewrite a job ad for a stylist so it sounds more welcoming. No applicant names or details are included.
Answer: Yes — it's your own business text with no personal data. Then read the result for wording that could discourage some applicants, such as references to age or "fitting in."
No fake reviews, ever
AI makes it easy to write a dozen glowing reviews in a minute. Don't.
The U.S. Federal Trade Commission (FTC), the federal agency that protects consumers from unfair business practices, has a rule on consumer reviews and testimonials. It bans fake reviews, including reviews written by AI and reviews that pretend to come from customers who don't exist or never used your business. Beyond the law, fake reviews destroy trust when they're found out.
What you can do with AI:
- Draft a polite message asking real customers to leave an honest review.
- Draft replies to real reviews, good and bad. Remove names before you paste, then personalize the reply yourself.
- Summarize themes from your real reviews to see what to improve.
What you can't do: write reviews or testimonials "from customers," change a real customer's words to say something they didn't, or post AI-written reviews under any name. If you're unsure about a review practice, check with a qualified professional.
Honest marketing claims
The same goes for the rest of your marketing. AI will happily write "the best-rated plumber in town" or "trusted by thousands." If you can't prove it, don't publish it.
- Remove claims like "best," "number one" or "fastest" unless you can back them up.
- Check every number, award and qualification the AI included. It may have invented them.
- Don't let AI describe services you don't offer or results you can't promise.
Key points
- AI can repeat unfair assumptions in hiring, tenant screening and customer communication. Check by swapping details and reading for assumptions.
- Use written criteria and keep people in charge of decisions about people.
- Never use AI to write fake reviews or testimonials. The FTC's rule on consumer reviews bans fake reviews, including AI-generated ones.
- Only publish marketing claims you can prove.
Fair wording and honest reviews
Prompts for checking drafts for bias and for asking for and answering real reviews.
- Review this job ad for wording that could discourage applicants because of age, gender, family situation or background. List each phrase and suggest a neutral alternative. [job ad]
- Rewrite this customer message so it is equally warm and respectful whoever the customer is. Don't make assumptions about their background. [message with names removed]
- Write a short, friendly message asking a customer who recently used our [service] to leave an honest review. Don't ask for a positive review or offer anything in return. Use [Customer Name] and [review link] as placeholders.
- Draft a calm, professional reply to this negative review. Thank them, acknowledge the issue, and invite them to contact [role] at [phone or email]. No arguments and no promises. [review with the reviewer's name removed]
- Summarize the common themes in these real customer reviews: what people like and what they want improved. [reviews with names removed]
- Check this marketing copy and list any claim (like 'best,' 'number one' or a number) that we would need proof to publish. [copy]