The AI bouncer problem

Your customers are already talking to AI about you, right now, and you have no idea what it is saying.

This is happening today. People do not search the way they used to. For twenty years, the move was simple. You had a question about a business, you Googled it, you got ten blue links, you clicked around, and made up your own mind. You saw the business's own website. You saw its reviews. You judged for yourself with your own eyes.

Now, a huge and growing number of people just ask an AI. Is this place any good? What is their return policy? Are they cheaper than the other ones? Should I trust them? And the AI answers. Right away. Confidently. In a calm, friendly, authoritative voice that sounds like it knows exactly what it is talking about. Whether or not it actually does.

Google rolled out AI Overviews to more than 1.5 billion people a month, and that little AI summary now sits at the very top of the page, above all the links, answering the question before you ever click anything. ChatGPT has hundreds of millions of users every week, and a growing share of them treat it like a search engine. What I keep seeing, over and over, is the same thing, that the AI answer at the top is eating the clicks that used to go to actual websites. People read the summary and they stop. They do not scroll. They do not click. They take what the AI said as the answer and they move on with their day.

You are being described to potential customers in conversations you will never see, by a machine you do not control, working off whatever scraps about you happen to be floating around the internet. Old listings. A stale third-party directory. A review from four years ago. A competitor's comparison page. The AI scoops all of it up, blends it together, and serves a confident little verdict about you to somebody who is deciding right now whether to give you their money. And if the information it is working from is thin or old or just plain wrong, the AI does not stop and say, "You know, I am really not sure about this one." That is not what these systems do by default. They fill the gap with their best guess, and they say it with total confidence, and the customer has no reason to doubt it.

A restaurant changes its hours. The owner updates the website because, of course, they do, that is the responsible thing. But they do not update the dozen other places online where their old hours are listed. Google's listing. An old Yelp entry. Some aggregator nobody has thought about in years. Somebody asks their AI assistant, "Is this place open tonight?" The assistant pulls the stale data, and it says, "No, they close at six." The customer shrugs and goes somewhere else. Meanwhile, the restaurant is sitting there fully open, lights on, tables empty, staff standing around, bleeding money to a wrong answer it never gave and never even heard. There is no angry review to respond to. No complaint to fix. Just a slow, quiet leak of business that the owner cannot see and cannot explain.

This is the part that makes it genuinely unsettling. With a bad review, at least you know. You can read it. You can respond. You can fix what caused it. But when an AI quietly gives a wrong answer about you to a customer who then never shows up, there is no trace. No notification. No feedback. The customer does not know they got bad information. You do not know you lost them. The transaction that never happened leaves no evidence behind. It is the most invisible kind of lost business there is.

And it compounds. Because the more people rely on AI to do their first round of judging, the more your fate gets decided before a human ever lays eyes on your actual website or your actual reviews or your actual face. The AI is the bouncer now. It decides whether you even make the shortlist. If it has a bad or outdated impression of you, you get filtered out before the customer ever consciously considers you. You do not lose the comparison. You never make it into the comparison at all.

A year from now, managing what the AIs say about you will be as basic and non-optional as having a phone number. There is going to be a whole scramble, the way there was a scramble for Google rankings fifteen years ago, except this time it is about making sure the machines have an accurate, current, honest version of your business to work from. The businesses that get ahead of it will feed the world a clear, straight story about who they are and what they offer, and every assistant that mentions them will get it right. The businesses that ignore it will get represented by guesses, by stale scraps, by whatever the internet happened to remember about them, and a guess is not on your side. A guess does not care if you are open. A guess does not know you lowered your prices or expanded your hours or fixed the thing everyone used to complain about.

The single most useful thing you can do about all of this is also the most boring, and the trade is lopsided in your favor. It costs a few hours of honest attention, no budget and no agency, and it keeps paying you back every time an AI is asked about you, around the clock, for as long as the facts stay true. Keep your own website accurate, complete, and current. Not clever. Not optimized to death. Just true, thorough, and kept up to date.

Here is why that matters more than it used to, and here is the catch. When these systems go looking for an answer about a business, they would love to use that business's own website as the source, because in principle it is the horse's mouth, the one voice that should know best. But they do not hand over that authority automatically. They give it to the site that earns it, the one that looks current, accurate, rich, and genuinely authoritative on the exact question being asked. Make your own pages the most complete and up-to-date account of your business, and you win that trust, and the AI quotes you. Leave them thin, or stale, or silent, and the AI does not wait around. It reads the gap as permission to go hunting through the old listings, the four-year-old review, the competitor's comparison page, and it fills it with a guess. The horse's mouth only counts if the horse actually speaks up, clearly and completely, and keeps speaking as things change. That is the part you have to earn. So write down the things people actually ask about. Your hours. Your return policy. What you charge. What makes you better than the alternative. Whatever used to draw complaints, and how you put it right. Put it on your own pages, in plain words, and keep it current. That is not marketing. That is just handing the machines the real version of you so they stop inventing one.

And once you have done that work, there is a genuine bonus waiting, and it compounds. The very same accurate, complete, current content that earns you authority with the outside AIs is exactly what your own Guru runs on. The richer and fresher you keep it, the smarter and more capable your Guru becomes, because it is digesting a fuller, more current picture of your business with every update. You are not maintaining two separate things. You are feeding one source that serves both. So the discipline that gets the world's machines to describe you correctly also directly upgrades the experience your prospects and clients get when they talk to you, across web, email, phone, SMS, and WhatsApp. They ask harder questions and get better answers. Fewer dead ends, fewer "I am not sure," more of the right answer on the first try, delivered with the patience and accuracy of your best employee on their best day. This is what Curacao Department Stores does in production: it answers more than 100,000 interactions a month off its own current information at 97%+ accuracy, validated by their own Customer Service team led by SVP Joseph Jiron. (It is the same reason your AI is now your front desk, and the same discipline behind an AI that is willing to say "I don't know".)

So the whole game now comes down to one choice. Either you tell the world's machines who you are, accurately and on purpose, and let that same honest record serve your customers everywhere they reach you, or you let the machines make it up. And they will make it up. They are making it up right now.