When someone asks an assistant “who does this in my area, and are they any good,” the answer is assembled, not looked up. The model builds a description of your business out of two things: what your own site publishes in a form a machine can read, and what the rest of the web says about you. You control the first completely. You shape the second over time. Most businesses have never deliberately set either, so the answer a machine gives about them is an accident.
This is a reference on how that answer gets made, what moves it, and what does not. To see the machine-readable half of it measured on your own site, the practice publishes a free AI visibility check that reads your pages the way a crawler does and shows you the business a machine can assemble from them.
Where the answer actually comes from
An AI assistant is not reading your website live in most cases. It is drawing on an index built by crawlers, the same kind of index that powers ordinary search, plus whatever it can fetch in the moment a question is asked. That means the description of your business has two sources, and they matter in different ways.
The first is your own site, as a machine sees it. Not the site a person sees in a browser, with images and scripts and a phone number in a styled header, but the raw text and the structured data underneath. If the machine-readable version plainly says what you do, where you are, and how to reach you, the assistant has clean facts to work from. If those facts live only in a picture, or only appear after JavaScript runs, the machine does not have them, because most AI crawlers do not render pages or read images the way a person does.
The second source is everything off your site: your Google Business Profile, your reviews, directory listings, news, and any reference source that mentions you. This is where an assistant corroborates or contradicts what your site claims, and where a small local business is most often simply absent. You do not control this half. You influence it, by being consistent everywhere your name appears and by earning the mentions that reference sources pick up.
Why the answer is so often wrong
The instinct is to assume a wrong answer is the model’s fault. Usually it is not. The most common cause of a bad AI description is a plain fact the site never states where a machine can read it.
A business whose homepage is a striking full-screen image with the tagline baked into the graphic has told a person everything and a machine nothing. A restaurant whose menu and address load through a booking widget has hidden its two most important facts behind code a crawler will not run. A firm whose services are described only in a slideshow has no plain sentence anywhere that says what it does. In each case the machine does exactly what a machine does when a fact is missing: it guesses, or it stays silent, and a competitor who stated the fact plainly gets named instead.
The second common cause is a self-inflicted block. In the wave of concern about AI training, a great many sites added a rule to their robots.txt file to keep the training crawlers out. That is a defensible choice. But the same rule, written broadly, often blocks the search and answer crawlers as well, the ones that populate the indexes ChatGPT search, Perplexity, and Google’s AI answers actually cite. A business can mean to opt out of feeding a model and, in the same line, opt out of being recommended by one. The two are separate decisions and deserve to be made separately.
Generative engine optimization, without the mystique
The work of making a business legible to AI has acquired a name, generative engine optimization, and a shorthand, GEO. A parallel term, answer engine optimization, covers the same ground. The names are useful for finding the topic. They are not evidence of a new discipline.
Strip the vocabulary away and the work is familiar. State clearly what the business is and does. Put the facts a machine needs into structured data, the standardized format search engines have consumed for a decade. Keep your name, address, and phone identical everywhere. Make sure the content exists in the page itself, not only after a browser assembles it. Let the right crawlers in. Every one of those is ordinary technical and editorial quality, and every one of them helped your search visibility long before an assistant was the reader. A vendor selling GEO as a separate mystery, priced as a monthly subscription to a black box, is usually selling the SEO fundamentals a business already needed, relabeled for the moment.1
What you control, and what you only influence
The honest division is worth stating plainly, because it decides where effort is worth spending.
- Fully in your control: the machine-readable version of your own site.
- The plain statement of what you do and where. The structured data that names the business, its type, its address, and its contact details. Whether the content is in the page or trapped behind scripts. Which crawlers are allowed. This is the half that determines whether a machine can describe you at all, and it is entirely yours to set. It is also the half almost no one has set on purpose.
- Yours to influence over time: the corroborating web.
- Your Google Business Profile, claimed and accurate. Consistent listings across the directories that feed the assistants. Reviews. Mentions in sources a model treats as reliable. You cannot dictate these, but a business that is consistent and genuinely present accumulates them, and a business that is careless stays a blur.
- Not yours at all: the model’s judgment.
- How a given assistant weighs its sources, what it chooses to say on any particular day, and how that changes as the systems change. Chasing the output directly is a trap. The durable move is to get the inputs right and let the answer follow, the same discipline that has always separated marketing that compounds from marketing that reacts.
The one thing that matters most
If a business does only one thing, it should be this: say plainly, in text a machine can read, what it is, what it does, and where. Not in a slogan. Not in an image. In a sentence, early on the page, and again in structured data.
Everything else in this reference is real, but it orbits that center. An assistant asked to describe a business is, underneath, trying to answer three questions, who, what, and where. A site that answers those three questions unmistakably, in a form a machine can read, has done the majority of the work, because it has removed the need for the machine to guess. This is also, not by coincidence, what makes a site clear to the human who lands on it. The practice has argued for years that clarity is the whole job; AI has only added a second reader who is even less forgiving of vagueness than the first.
To see this measured on your own pages, run the AI visibility check. It reads your site the way a machine does, scores what it finds against the signals above, and shows you the business a machine can state from your pages, with a blank wherever your site is silent.
Frequently asked questions
How do I get my business to show up in AI search?
Make the facts a machine needs easy to read: state plainly what the business does and where, in your page text and in structured data; keep your name, address, and phone consistent; and let the AI crawlers reach the site instead of blocking them by accident. AI search draws on the same signals ordinary search does, so most of the work is the same work.
What is generative engine optimization?
Generative engine optimization, or GEO, is the practice of making a business legible to AI assistants so they cite and describe it accurately. In substance it is not a new discipline. It is the existing work of clear content, structured data, and technical fetchability, aimed at a new set of readers. Treat any vendor selling it as a separate mystery service with caution.
Can I control what AI says about my business?
You can control the inputs on your own site completely, and you influence the rest. What you publish, how you structure it, and whether crawlers can reach it are yours to set. What the wider web says about you, your reviews, and your presence in reference sources, you can shape over time but not dictate. An assistant assembles its answer from both.
Why does AI get my business wrong?
Usually because the site never states the fact plainly where a machine can read it. If your services live only in an image, your address only in a footer graphic, or your whole page is assembled by JavaScript that crawlers do not run, the machine fills the gap with a guess. Wrong AI answers are far more often a silence on your side than a mistake on theirs.
Does blocking AI crawlers hurt my visibility?
It depends which ones. Blocking model-training crawlers is a legitimate choice with little downside for visibility. But the same robots.txt rule often blocks the search and answer crawlers too, which is what removes a business from AI answers. Many sites that meant to opt out of training opted out of being cited too, without intending to. Separate the two before you block anything.
Is AI visibility different from SEO?
Mostly it is the same foundation viewed through a new lens. AI answers are built on the same crawled, indexed, structured web that search results are. The differences are at the margin: answer engines lean harder on machine-readable facts and plain statements of what a business is, and they read a page as text rather than rendering it. Good SEO gets you most of the way; the rest is clarity.
Binary Glyph is a brand and marketing practice in Toledo, Ohio. Making a business legible to both machines and people is the ordinary work of the practice, not a separate product; most engagements begin with a fixed-fee marketing audit that documents where a business is clear and where it is not, credited toward a six-month Foundation Engagement. If an AI assistant is describing your business wrongly or not at all, begin a conversation →
This reference describes how current AI systems generally find and use information. The specifics change as the systems do; the underlying discipline, stating your facts plainly and making them machine-readable, does not.