The AI Search Visibility Checklist We Run on Client Sites

More of our clients' customers now find them through ChatGPT, Perplexity, Google AI Overviews and Gemini before they ever open a classic search results page. Because of that, "AI search visibility" is now a standard line item in how we set up every new website.

There is no before-and-after graph in this post. Almost no AI platform publishes citation analytics yet, so anyone showing you a clean "AI visibility score" chart is working from estimates. What we can share is the checklist we run through on client sites, the parts that turned out to matter, and the parts that turned out to be mostly noise.

What AI search visibility means

AI search visibility is whether an AI system (ChatGPT, Perplexity, Copilot, Gemini, Google's AI Overviews) can find your content, understand it correctly, and choose to cite or reference it when answering a relevant question. It works alongside classic SEO. The same crawlers that feed Google's index increasingly feed these AI answers, and the clear, well-structured content that ranks well also gets quoted well.

The hard part is measurement. None of the major AI platforms give site owners anything like Google Search Console, a dashboard that says "you were cited X times this month." So the work happens upstream. We get the technical and content fundamentals right, then check by hand whether the site starts showing up.

Step 1: Check which AI crawlers are allowed on the site

Before touching content, we check robots.txt. On most sites we inherit, nobody has reviewed it since launch, and that one file decides whether AI systems can see the site at all.

There are two kinds of AI bots, and mixing them up is the most common mistake we see:

• Training crawlers (GPTBot, ClaudeBot, Google-Extended) scrape content to train future models. Blocking these has no effect on whether you appear in today's AI answers.
• Search-time crawlers (OAI-SearchBot, PerplexityBot, Google's regular Googlebot which now also feeds AI Overviews) fetch pages live to answer a specific user question. Blocking these is what removes you from AI answers.

By default we recommend that client sites allow the search-time crawlers. Training crawlers are the client's call, and we explain what each one does so they can decide with full information instead of blocking everything because "AI" sounds risky. The most common way we see businesses opt out of AI search visibility by accident is a blanket block in robots.txt, copied from a template by someone who did not know which bots do what.

Step 2: Decide whether to add llms.txt

llms.txt is a proposed standard, a plain-text file that gives AI systems a structured summary of a site's main pages, similar in spirit to a sitemap. It is new enough that there is no confirmed evidence it directly influences whether a model cites you. We add it on client sites where setup is cheap (it usually is, once the page inventory is documented), and we tell clients plainly to treat it as a low-cost hygiene item with no guaranteed effect on visibility. Anyone promising ranking gains from llms.txt today is ahead of the evidence.

Step 3: Structure content so AI systems can quote it

Most of the work happens here, and it overlaps heavily with good SEO. AI systems tend to quote content that answers a question directly near the top of a section, where nobody has to piece the answer together from three paragraphs of framing.

What we change on client pages:

• Add a direct, complete answer in the first two sentences of each section, then expand with detail below it.
• Add schema markup (Organization, Service, FAQ, Article where relevant) so machines can parse what a page is about without guessing. We cover the technical side of this in more depth separately.
• Break out a short FAQ block on service and product pages, phrased the way a person would ask the question.
• Use specific numbers and named examples instead of vague claims. "Our support package is 10 hours" gets quoted, and "flexible support options" does not.

Step 4: Check internal links and site structure

AI systems, like search engines, use internal links to work out which pages on a site matter and how they relate. A service page that nothing links to, buried three clicks deep, is unlikely to get cited even if the content is well written. Part of our GEO setup is making sure every core service page can be reached from the homepage and from at least one relevant blog post, with anchor text that describes the destination instead of a generic "click here."

Step 5: Check AI answers manually

With no analytics dashboard for this yet, we check the unglamorous way. On a recurring schedule, we run the client's target questions through ChatGPT, Perplexity and Google's AI Overviews ourselves and note whether the site shows up, what gets quoted, and what a competitor's page says that ours does not. This does not scale to hundreds of keywords. Right now it is still the only reliable signal available, and it tells us exactly which page needs a rewrite.