Is your business invisible to ChatGPT?

How AI assistants decide who to recommend

When someone asks ChatGPT "what's the best CRM for solo realtors," ChatGPT doesn't show ten results. It picks one or two and tells the user to go with them. Same with Claude, Perplexity, Gemini, Google AI Overviews, and Microsoft Copilot. Six different AI assistants, six different ways of picking which businesses get recommended — and almost no small business owner knows what those rules are.

Here is what we have learned from auditing hundreds of small business websites for AI search visibility.

The signals every AI assistant looks for

AI assistants pull from a mix of crawled web pages, structured data, and licensed knowledge bases. They favor sites that make it obvious what the business does and who it serves. The five signals that matter most:

• Clear entity identity. Your business name, what you sell, where you operate, and how to contact you should all be stated in plain text — not buried in images or hidden behind JavaScript. Organization schema (a small block of structured data in your page) makes this even clearer.
• Content depth. Pages with under 500 words are treated as thin. AI assistants prefer pages that explain a topic thoroughly, with specific facts, numbers, and examples instead of vague claims.
• Citation-friendly structure. Clear headings (H1, H2, H3), bullet lists, FAQ sections, and HowTo steps make it easy for an AI assistant to grab a single passage and quote it as the answer. Sites without this structure get skipped.
• Open crawler access. If your robots.txt blocks GPTBot, ClaudeBot, PerplexityBot, or Google-Extended, those assistants literally cannot see your site. Many website builders (Wix, Squarespace, certain Shopify themes) block AI crawlers by default.
• Authority signals. Mentions on other sites, original research, dated content, named authors, and consistent branding across the web all tell AI assistants that you are a credible source. New sites with no outside mentions struggle here — this is the part that takes the longest to build.

What llms.txt is and why it matters

In September 2024 a developer at Answer.AI proposed a new web standard called llms.txt — a single text file at the root of your site that tells AI assistants which pages on your site are most important and what they are about. It is roughly the AI search equivalent of a sitemap. Reffed checks whether your site has an llms.txt file, whether it is properly formatted, and whether it covers your highest-value pages. As of mid-2026 only a small percentage of sites have one, which means an llms.txt is still an easy way to stand out.

How Reffed measures whether AI assistants would cite you

Beyond the technical checks, Reffed runs a Mention check on every audit. We identify what category your business is in (B2B SaaS for solo founders, dental practices in Calgary, AI automation agencies in Vancouver — whatever fits), then assess how likely a major AI assistant would be to cite your site for a typical category query versus larger, better-known competitors. The verdict comes back as likely, possible, or unlikely — with the specific reasons why. This is the same authority assessment we use internally to set realistic expectations with our own Reffed Watch customers.

Why this is urgent (but not panic-urgent)

Most small businesses still get the majority of their inbound from Google search, Google Maps, social media, and word of mouth. AI search visibility is a small fraction of total traffic today — but it is growing fast, and the businesses that get cited now are building the citation history that AI assistants will keep pointing at for years. Setting up the foundation while the field is uncrowded is far cheaper than trying to catch up after every competitor has done it. That is the bet Reffed Watch is built around: spend $29 a month for the next year, build the foundation, and be cited when the wave actually arrives.