Why AI May Send Customers to Everyone but You

For twenty years, getting found online meant ranking on Google. That is changing fast. More of your customers now begin with an assistant. They ask ChatGPT for a recommendation, they read the AI summary at the top of a Google search, they ask Perplexity to compare three vendors. In each case the assistant does not hand back a page of links to work through. It names a short list of businesses. The risk is not that you rank a little lower. It is that the assistant confidently recommends two or three companies like yours and never mentions you, and you will not see it happen, because a recommendation that never included you leaves no trace in your analytics.

Here is the uncomfortable part: a site that looks perfect to a person can be nearly unreadable to the systems doing the recommending. The two do not see the same thing.

Google is one door of many

For twenty years there was effectively one front door, Google search, and one thing to optimize for. That era is ending. A customer today might ask ChatGPT, read Google's AI summary, compare vendors in Perplexity, lean on Microsoft Copilot inside the tools they already work in, or use the assistant built into their phone. Each of these is a separate entry point, with its own crawler and its own way of deciding whom to name, and Google is now only one of them. A business that optimizes only for Google's blue links, the way most still do, is guarding one door while customers walk through five others. The encouraging part is that the work which makes you readable to one assistant makes you readable to nearly all of them, because they need the same things: a page they can read in raw form, clear facts stated plainly, and permission to come in. Get those right once and you show up across the board.

What the assistant actually sees

When an AI system looks at your site, it does not open a browser and wait for everything to load the way a person does. It requests the raw page and reads what comes back. If your site assembles itself in the browser with JavaScript, the assistant may receive an almost empty shell. Your homepage, to the machine, is a blank page, and it cannot recommend what it cannot read.

Who is actually doing the reading

It helps to know what is on the other end. When someone sees the AI summary at the top of a Google search, that answer is written by Gemini, Google's AI model. Gemini does not crawl the web itself. It draws on Google's existing index, the one built by Googlebot, the same crawler that has always read your site, along with some live retrieval. Other assistants run their own crawlers doing the same job. The practical point for an owner is simple. A page has to be readable by these crawlers, and present in the index, before any AI can name you. Give the crawler a blank shell, or shut it out entirely, and the AI has nothing to work with and moves on to a competitor.

The facts it needs, in a form it can use

People infer your business from context. Machines need it stated. Your name, address, phone, hours, and services should exist in the page as structured data, not only as text baked into a graphic. When they do not, the assistant guesses, and it often guesses wrong, or it simply moves on to a competitor whose facts are clear.

What it looks like in practice

Picture an owner in your area asking an assistant, "who installs security cameras for a warehouse near me." The assistant does not return ten links to sort through. It names two or three companies, a sentence or two each, with a reason. To be one of those names, your site has to be readable to the crawler behind that assistant, has to state plainly what you do and where you do it, and has to be allowed in. Miss any one of those three and you are simply not in the answer, and the owner never learns you existed. Now multiply that by every assistant and every question a customer might ask about what you sell. The cost of being unreadable is not a lower ranking on a page nobody scrolls anymore. It is a slow, silent disappearance from the shortlist.

Whether it is even allowed in

Several settings decide this, and most owners have never seen any of them. The first is robots.txt, a file that can quietly tell crawlers to stay out, often left in place by a plugin nobody chose on purpose. The second is a snippet setting, a small instruction such as nosnippet that tells Google not to show a preview of your page. It reads as harmless, but it also removes you from the AI summaries, because those are built from previews. The third is your security provider, the service that protects your site from attacks, which may be turning assistants away at the door even when robots.txt invites them in. As of late 2026 the major providers split this into separate controls, and a setting changed for one reason can block AI for another. We have watched sites that welcome AI in writing get turned away in practice, and the owner had no idea.

There is one more source of confusion worth clearing up. Many owners have heard of a Google setting called Google-Extended and assume it governs whether Google's AI can use their site. It does not. Google-Extended only controls whether your content helps train Google's AI models. Whether Google's AI answers can find and name you is decided by the crawler and snippet settings above. It is easy to spend worry on the wrong lever while the real one goes unchecked.

Freshness

Assistants favor sites that look maintained. A site with no sign it has been touched in years reads as stale, and stale loses to current.

None of this requires rebuilding your website. Most of it is configuration: making the copy readable in the raw page, adding the structured facts, correcting the crawler and snippet settings, and confirming your own security layer is not the thing blocking you. It is the difference between being one of the three businesses an assistant names and being the one it never mentions. And because a recommendation you were left out of makes no sound, the only way to know where you stand is to look.