Insights · AI visibility · from our 2026 research
When someone asks ChatGPT, Claude, Gemini or Perplexity “who’s the best [your trade] in [your town]?”, the assistant doesn’t search your website. It searches the web the way a researcher would, picks a handful of sources it trusts, and names the businesses those sources name. In our 2026 AI Visibility Report we put 422 buyer questions to four engines for 108 businesses. 41 of them were never cited once. Here is why, in the order we saw it most.
What we found
A robots.txt that blocks GPTBot, ClaudeBot or PerplexityBot, a noindex tag left over from a redesign, or text that only appears after JavaScript runs. It happens to good businesses more than you’d think: our own site shipped a launch with a noindex tag on every page and nobody noticed for two weeks. Fix: run the free check; it reads your site the way a crawler does and tells you in ten seconds.
For local questions, engines lean on directories and round-ups: “10 best X in South Carolina” lists, Yelp, LinkedIn, Reddit threads. In our tests LinkedIn, Yelp and Reddit were among the ten most-cited sources. If those pages don’t mention you, the engine has no way to. Fix: find the three sources the engines actually cite for your question (we show you in the full check) and get listed, reviewed or mentioned there.
Engines cite pages that answer the question asked. “Who can build an AI agent for emailed orders?” gets answered by a page about AI agents for emailed orders, not by a homepage that says “innovative solutions.” Most sites have a services page; few have a page per question. Fix: write down the five questions a buyer asks before they call you, and give each one a page that opens with a two-sentence answer.
No author, no date, no credentials, no named clients. Engines (and Google’s own guidance) favour first-hand experience: a named person, specific numbers, original data, real case studies. Commodity copy gets skipped. Fix: put a visible byline and an updated date on every guide and case study, and quote a real result with a number in it.
Your name is one thing on your site, another on Google, a third on LinkedIn; the address differs; there’s no structured data saying “this is a business, here, doing this.” Engines match entities, and an entity that doesn’t resolve cleanly is easy to leave out. Fix: one consistent name, address and phone everywhere, Organization and LocalBusiness markup on the site, and a current Google Business Profile.
The reporting shift
Pick ten questions your buyers actually ask. Once a month, put them to ChatGPT, Claude, Gemini and Perplexity and count three things: how often you’re cited as a source, how often you’re named, and how often you’re recommended. That’s the number. Rankings and sessions tell you about Google in 2019; this tells you about the assistant your next customer is talking to.
Most of it, yes: the robots check, the bylines, the question pages. The part that’s hard alone is knowing which sources each engine trusts for your questions, because it changes by engine and by intent. The full check runs the questions for you and shows who got cited instead.
The free instant check reads your website the way an AI crawler does in about ten seconds. The full check puts your buyers’ questions to four engines and shows who they name.
