Your next customer is asking an AI, not Google

For a few days while building this, we stopped reading about the shift and ran a small test instead. We opened ChatGPT, then Claude, then Perplexity, and asked each one the way a customer would — not "list barbershops in the city," but "where can I get a solid fade on a Sunday afternoon near here?" Then we asked about places we actually know — a neighbourhood barber, a café down the road, a garage we'd trust with a car.

None of it looked like a search results page. There was no list of ten blue links to scan and judge for ourselves. Each assistant just answered — a short, confident paragraph naming a couple of places, with hours and a line about what they're known for. Sometimes it was right. Sometimes the hours were from an old listing, or it described a shop that had changed hands, or it recommended the two businesses that happened to have the most written about them online and quietly skipped everyone else.

The businesses it skipped have no idea it happened. That's the part that stuck with us.

The answer is the page now

For twenty years, "getting found" meant ranking. A customer typed something, got a list, and chose. You could be third instead of first and still get the click. The whole game was your position on a page of options.

An assistant doesn't hand back a page of options. It hands back an answer. It reads whatever it can — your website if you have one, third-party listings, reviews, structured data, and increasingly files meant specifically for it — and composes a reply. Your business is either in that reply or it isn't. There's no second page. There's barely a first one.

So the question quietly changed. It used to be "where do I rank?" Now it's "when an assistant answers for a customer, does it know I exist, and does it get me right?" Those are not the same question, and the second one has almost no tooling around it.

You can't see the ones you lose

Here's what makes this hard to act on: the loss is invisible.

Search at least gave you feedback. You could watch rankings move. You knew, roughly, when you were showing up and when you weren't. Bad news, but legible.

An AI answer gives you nothing. When an assistant recommends three businesses and yours isn't one of them, there's no impression count, no "you appeared for this query," no notification. The customer gets their answer and moves on. You never learn the conversation happened. Nothing in the reply says "a business like this one should be here, but we couldn't read it."

That's the real problem underneath the technical one. It isn't only that businesses aren't machine-readable — it's that they can't feel the cost of it.

“Nothing lands in your inbox to say: you lost a customer to a paragraph you'll never see.”

Why nobody's panicking yet

Three honest reasons. One, you can't see it, so it doesn't feel urgent. Two, it still reads as a tech-person thing — "AI is answering questions about your café" sounds like a headline, not like a Tuesday. And three, the one we have the most sympathy for: every small-business owner we've met has been sold a "you have to be on this new platform" story by someone with a monthly retainer, and most of those stories were noise. The skepticism is earned.

The trouble is that skepticism is a bad filter here, because this is one you can check yourself in five minutes — which is exactly what most of the retainer stories couldn't survive.

It's a reader problem, not a ranking problem

The instinct is to treat this like SEO: a new algorithm to game, a new set of keywords. We don't think that's right, and we've had to keep correcting ourselves on it while building.

SEO is about ranking pages for a search engine that sends a human to click. This is a different reader — an assistant acting on a customer's behalf, that wants facts it can trust and repeat. It doesn't want to be persuaded. It wants your hours, your services, your prices, your policies, plainly, from a source it can tell is actually yours, so it can answer without guessing. When it can't find that, it does the most human thing an LLM does: it fills the gap with something plausible.

So the fix isn't ranking higher. It's being readable, in the boring literal sense — publishing the facts about your business in formats an agent can parse, grounded in what's true, from a source an assistant can recognize as authoritative. Less "beat the algorithm," more "leave a clear note for whoever's asking about you."

That's what we're building Lodestar around — a machine-facing presence for a small business, so an assistant can find it, answer about it accurately, and pass along a lead. But you don't need our product, or any product, to see the problem for yourself. You need about five minutes.

Go read your own answer

Here's the test, and it costs nothing. Open ChatGPT, or Claude, or Perplexity — ideally the ones your customers actually use — and ask about your business the way a stranger would. Not by name, but by need: the thing you do, near where you are. See if you come up. Then ask directly, by name, and check every fact in the answer — hours, services, prices, whether it even describes you as the right kind of business.

You'll get one of three results. It doesn't mention you. It mentions you and gets something wrong. Or it's right — in which case notice what it read to get there, because that's the part worth protecting.

Whatever you find, you'll have looked at a page your customers can already see and you never had. That's the whole shift, in one afternoon: your business has a second front page now, written by a machine from whatever scraps it could find. Might as well read it.