How we make a business easy for AI to recommend

Here is a quiet way a business vanishes from AI answers. Years ago, someone turned on bot protection, a security filter that decides which visiting programs get through to a website. It was the right call at the time. Scrapers were copying prices, and the filter stopped them. Nobody has thought about it since.

When someone asks ChatGPT or another AI tool for a recommendation today, those tools send out reading programs, software that opens your pages and carries what it finds back to the assistant. To an old security filter, they look like more bots to turn away. The site loads fine for every human, yet the front door stays locked to the very programs that would recommend the business. Nothing looks broken. No error ever reaches the owner. The business simply never comes up.

That is the most common shape of the problem: not a missing trick, but an unnoticed blocker. Making a business easy for AI to recommend is diagnosis work. Confirm the AI tools can get in, make the facts easy to read, publish the free extras, then measure honestly, in that order, because that is the order the evidence supports.

Why an AI visibility audit comes before any deliverable

If you are weighing whether to pay for generative engine optimization, the trade’s name for getting a business recommended by AI tools, the honest version of the work looks less like a product and more like a checkup. The first step is an AI visibility audit: we test how AI tools currently see and describe a business, what they say, what they miss, and how competitors appear. Everything after that depends on what the audit finds, because a locked front door makes every other improvement invisible.

The sequence is not a preference. Access comes first because it is the one step the AI companies spell out in their own documentation. The evidence gets thinner further down the list, so that is where the budget goes last.

Step one: confirm the AI tools can get in

Bot protection is the usual suspect, and the tricky part is that bad bots really do exist. Fakes even impersonate the good ones. That is exactly why the companies behind ChatGPT, Claude, and Perplexity each publish an official roster of their reading programs and the addresses they use. With it, a site can check its visitor records, verify the real programs, and let them through while still blocking impostors. Either the programs are getting through or they are not; no faith required. Which bot belongs to which company, and which file controls the door, is a guide of its own.

The same check covers two more things. Does the business show up in Bing, the search engine two of the biggest assistants lean on? And do your key pages show their text right away, without a click or a scroll? Text hidden behind a tab is hidden from these programs too.

Step two: make your facts easy to read

Your hours, service area, what you actually sell, and how to reach you should sit on the page as plain text. Not inside a photo of a price board, not in a PDF, not in a widget that only appears after a tap. A reading program cannot pull your Saturday hours out of an image.

This step also covers your listings, so that directories repeat the same facts your site states. Getting your Houston listings consistent is ordinary local search work, and which facts earn a mention in AI answers is a subject of its own. The audit’s job is narrower: confirm the facts exist as text a machine can quote.

Step three: publish the free extras

Websites can publish a small family of files addressed to machines. The best known is llms.txt, a plain text table of contents that lists your important pages for AI tools. Who actually reads each of these files is a question that guide answers; the audit asks something narrower. Does your site generate them as part of its build, the way ours does? On a well-built site the files are a byproduct, refreshed whenever the content changes, which is why this step is cheap and sits third.

Step four: measure honestly

Two things can actually be watched. The first is referral traffic, visits that arrive by clicking a link from somewhere else, which is how your visitor reports can show that a customer came from ChatGPT. The second is citations. Bing’s free tools for site owners now include an AI performance report that shows when a site is cited in AI answers across Microsoft’s products. It is currently the only free, first-party view any of the companies offers. Beyond those two, measurement is hand work: asking the assistants real customer questions on a schedule and recording when the business comes up, which is how we track mentions.

Now the calibration, because trust is earned or lost here. These figures come from studies of the web at large by SEO research companies; they are not our client numbers. Semrush measured AI tools driving about 0.14% of website visits, up 66% in a year, and Ahrefs measured about 0.25%. One measurement of a few hundred professional-services websites found the median business received about six AI-referred visits a month.

Anyone promising a flood of ChatGPT customers next quarter is contradicted by every measurement that exists. The honest case is different. This is a positioning move made now because it is cheap now: the visits are few and growing fast, and the businesses whose facts are readable when the channel matures will be the ones already getting recommended.

Where the budget goes first, and why

The spending order follows the evidence. Access checks and plain-text facts come first because they are the only steps the AI companies document themselves. The generated files come next because their cost is near zero. Just as important is what the budget does not lead with:

• We do not lead with a deliverable. The audit comes first, and what you pay for is the diagnosis and the fixes it surfaces.
• We do not promise AI traffic. The numbers above are small today, so we frame this as positioning and show you the measurements.
• We do not guarantee that any AI will recommend you. These systems change constantly, and nobody outside them controls what they say. What can be guaranteed is the work: the audit, the fixes, and steady tracking of mentions.

A useful test for any pitch: does it start with a diagnosis, or with a deliverable? The deliverable-first pitch skips the step where the locked front door gets found.

What a first conversation covers

Not a pitch deck. Five questions, in order. Is anything on your site turning away the reading programs? Does the business appear in Bing? Do your hours, services, and service area sit on your pages as plain text? Are the machine-readable extras being generated? And what will be measured each month so you can see whether it is working?

Most owners cannot answer any of the five, which is normal. That is what the audit is for, and it is where our generative engine optimization work always begins. The five questions above are the agenda for the first call, so bring your website address, not your budget.

Frequently asked questions

How do I know if AI tools can reach my website?

Check whether a security filter is turning them away. Bot protection added years ago to stop scrapers often blocks the programs AI tools send to read websites. The companies behind ChatGPT, Claude, and Perplexity publish official lists of the addresses their programs use, so the real ones can be verified in your site’s visitor records and allowed through.

How much traffic do AI tools actually send to websites?

The share is small and growing fast: studies by SEO research companies measured AI tools driving roughly 0.14% to 0.25% of website visits, up about 66% in a year, and one measurement of a few hundred professional-services websites found a median of about six AI-referred visits a month. That is why the honest framing is positioning. The businesses whose facts are readable now will be the ones getting recommended as the channel grows.

Can an agency guarantee my business will show up in ChatGPT?

No, and a guarantee is a warning sign. AI systems change constantly, and nobody outside those companies controls what they say. What an honest agency commits to is the work itself: an audit of what blocks you, fixes in a proven order, and regular testing that shows whether mentions are improving.

What does an AI visibility audit include?

It checks four things in order: whether anything on the site is blocking the programs AI tools send to read it, whether the business appears in Bing and states its facts as plain text on the page, whether the machine-readable extras are being generated, and what will be measured month to month so results can be judged honestly.