I see it on small business sites all the time now: a little chat bubble in the corner that pops open before the page finishes loading. “Hi there! How can I help you today?” It’s the website equivalent of a salesperson who greets you at the door while you’re still holding it open.
Then you ask it something real, like whether they service your zip code, and it says “Great question!” and hands you a link to the homepage. You were already on the homepage.
The short version
Usually not on day one. A chatbot is a shortcut to information, so if the information isn’t written down anywhere on your site, the shortcut leads nowhere.
Here’s the pattern I run into: the site has five pages, the services page says “We offer a wide range of solutions,” and there’s no pricing, no service area, no hours that match Google. Then somebody bolts a chatbot onto it and wonders why it keeps saying “Please contact us for more details.” Of course it does. That’s the only thing the site actually says.
(To be fair, that bot is being perfectly honest. It just doesn’t know anything.)
If your phone rings with the same six questions every week, a bot can absolutely help. But those six answers need to exist in plain words first. Write them down, put them on the site, and you’ve already fixed half the problem without any AI at all.
You are. Not the software company, not “the AI,” you.
This isn’t a hypothetical. In 2024, a Canadian tribunal ruled against Air Canada after its website chatbot told a customer he could claim a bereavement discount after his trip, which the airline’s actual policy said he couldn’t. Air Canada argued the chatbot was basically its own separate thing. The tribunal didn’t buy it and said the company is responsible for all the information on its website, whether it comes from a static page or a chatbot. The American Bar Association’s summary of the ruling is worth two minutes of your time.
The amount was small. The lesson isn’t. If your bot promises a price, a warranty, a same-day appointment or a refund you don’t offer, “the robot said it, not me” is not a strategy. It’s a half-assed excuse with a fancy font.
So before any bot goes live, I want to know exactly what it’s allowed to talk about and what it should hand off. Pricing it can quote from a real page? Fine. Anything involving money, safety, legal stuff or medical stuff? That goes to a human, every time. Call it cover your ass engineering.
Because most of them were installed, not set up. Somebody pasted a script tag, picked a friendly avatar named Sunny, and called it a day.
The useless ones have a few things in common. They open uninvited and cover the button you were about to tap on your phone. They answer every question with a link. They can’t tell you anything the footer couldn’t. And they never, ever let you reach a person without typing “human” four times like you’re yelling into a drive-thru speaker.
The good ones are boring in the best way. They wait to be asked. They answer from your real content. When they don’t know, they say so and offer a phone number, a form, or a booking link. That’s it. Nobody’s ever left a review that said “the chatbot was so charming,” but plenty of people have bailed on a site because the bot was a pain in the ass.
Your answers. Write the stuff people actually ask, on the pages where they’d look for it.
That means a service area that names real towns, prices or at least honest ranges, how booking works, what happens after someone fills out the form, and the stuff that makes you different that isn’t “quality” or “customer service.” A solid FAQ where every answer stands on its own does more work than most bots ever will.
Bonus: the exact same clear, specific content that makes a chatbot useful is what Google’s AI answers and ChatGPT pull from when somebody asks about businesses like yours. So you’re not doing the work twice. You’re doing it once and getting paid three times.
Then, if you still want a bot, it has something real to stand on. That’s the order I follow when I build AI automation for small businesses: content first, rules second, chat bubble last.
Test it like a skeptical customer, not like the person who paid for it. Open your site on your phone and ask it the five questions your customers ask most. Ask one it shouldn’t answer, like “Can I get 50% off?” Ask for a human.
Then look at the transcripts, if your tool keeps them. Most owners never do. That’s where you find out the bot has been confidently telling people you’re open Sundays since March. Damn, Sunny.
If it’s getting more wrong than right, turn it off. Honestly. A page with no chatbot loses nothing. A page with a bad one loses the customer and makes you look like you don’t know your own business.
They can be, once your site already answers the common questions clearly. Without that content, a chatbot just repeats “contact us” in a friendlier voice. Start with the answers, then add the bot.
Yes. In the Air Canada case, a tribunal held the company responsible for its chatbot’s wrong answer about a discount. Treat anything your bot says as something you said yourself.
I’d skip it. Auto popups cover buttons on mobile and annoy people who were about to do the thing you wanted. A small, quiet bubble that waits to be clicked works better.
Look, I like AI. I build with it every week. I just don’t like a cheerful little robot making promises on your behalf while you’re out on a job. If you want a site that answers the real questions first (and a bot that knows when to shut up and pass the phone), tell me what you’re working on and I’ll tell you honestly whether you even need one.
