When someone asks AI for a recommendation, are you in the answer?

Blocked crawlers are the common one, and they're almost never a decision anyone made. They arrive with a security plugin, a CDN default or a hosting provider's 'bot protection' setting, and they're a one-line fix once you know they're there.

The second most common is a site that renders entirely in JavaScript. Google will usually cope. Several AI crawlers won't — they fetch the HTML, find an empty shell, and move on to a competitor whose site is plain enough to read.

Each assistant uses named user agents you can allow or block in robots.txt. They do different jobs, which is why blanket-blocking has consequences people don't intend.

Indexability comes first because nothing else matters while a page says noindex or points its canonical somewhere else. Then live-search crawler access, pages written in clear complete passages, and schema markup so your business details are machine-readable rather than merely visible. An optional /llms.txt can provide a compact index for tools that choose to read it, but it is not scored or treated as a prerequisite.

None of this involves tricking anything. It's the same work as making a page genuinely clear, which is why it tends to improve conversion rate at the same time.

The honest position: major search providers do not promise to use it, and Google Search explicitly says it does not. It can still be a convenient, cheap summary for agents and tools that deliberately request it, so we maintain one without pretending it earns visibility by itself.

A missing /llms.txt is not a fault and the free checker does not score it. If one is published, it should be real plain text rather than an HTML error page returned with a 200 status.

A human reads your opening hours from a styled table. A crawler sees a table and has to infer. LocalBusiness schema removes the inference: this is the address, these are the hours, this is the phone number, this is the area served. Service and FAQPage markup do the same for what you sell and what you've already answered.

It's also the cheapest form of verification available. Assistants prefer sources they can corroborate, and structured data that agrees with your Business Profile and your directory listings is corroboration.

The questions are yours — the actual phrasings a customer would use, agreed at the start and kept stable so the month-to-month comparison means something. Changing the questions until one of them returns your name is how this gets faked, so the list doesn't move.

One honest limit, stated up front: Perplexity AI is the engine we can check reliably and repeatedly. ChatGPT requires a logged-in session and Google blocks automated checks. The report names the engine actually tested and never implies more than that.

The searches most at risk are the orienting ones: 'what does X cost', 'do I need a professional for this', 'what's the difference between A and B'. Those used to produce a click. Increasingly they produce an answer with two or three business names in it and no click at all.

That's the strategic reason to do this work now rather than in two years. It's cheap while few local competitors have done it, and the pages that get quoted by assistants are, in practice, the same pages that rank well — so the downside if AI search plateaus is close to zero.

It's a tier rather than a separate service because the work isn't separable — the technical fixes that make a page quotable are largely the fixes that make it rank. Buying GEO from someone who isn't also doing your SEO means paying twice for one job.

Every engagement starts the same way: a free check, then an audit in 48 hours. If the audit finds nothing worth fixing, it's free.