Executive summary
A business can sit on the first page of Google and still be left out of a ChatGPT or AI Overview answer. The two systems share some plumbing and differ in the rest, so a good Google result does not settle the AI question. What it does is hide it. Nothing in your rank tracker will turn red.
This piece expands a LumiRank Instagram carousel from 27 October 2026 into five signs that point to the gap. Each one rests on something a platform has published about its own behaviour, and each can be tested in minutes with a browser or a one-line command. None of them is a measurement of how often you are recommended; nobody can give you that number reliably, as one measurement is worth less than it looks. They are conditions that make being recommended harder or impossible.
One caution up front. Google states that there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show a snippet, and that inclusion is not guaranteed. Signs one and two below mostly concern ChatGPT and Claude. Signs three to five apply to both.
Sign 1: Your robots.txt or firewall turns away the AI search crawlers
OpenAI runs three separate crawlers. GPTBot collects content that may be used to train its models. OAI-SearchBot is "used to surface websites in search results in ChatGPT's search features." ChatGPT-User fetches a page when a person asks about it. OpenAI's documentation says each setting is independent, so blocking GPTBot to keep your content out of training does not remove you from ChatGPT search. Blocking OAI-SearchBot does. Opted-out sites can still show up as plain navigation links, and robots.txt changes can take around 24 hours to take effect.
Anthropic splits the same way. ClaudeBot is for training. Claude-SearchBot "navigates the web to improve search result quality for users," and Anthropic says that if you block it, your content isn't indexed for search optimization, which may reduce its visibility and accuracy in Claude's search results.
The usual failure is a blanket rule. Someone adds a block for every AI user-agent after reading a worried article about training data, and the search crawlers go with it. A second cause is one layer below robots.txt. Cloudflare announced on 1 July 2025 that it was changing its default to block AI crawlers unless they pay creators, so a site on a CDN or firewall may refuse these bots without a single line in its own files. Whether that applies to your zone depends on when and how it was set up. Check the dashboard.
Test it: open yourdomain.com/robots.txt and look for OAI-SearchBot, Claude-SearchBot, PerplexityBot or a wildcard Disallow: / under a group that names them. Then check your CDN's bot settings. Whether to allow training crawlers is a business decision. Allowing the search ones is what makes a citation possible. The full picture is on how to get recommended by ChatGPT.
Sign 2: The important details only exist after JavaScript runs
Google renders JavaScript. Its documentation says a headless Chromium renders the page once resources allow, and that a page may wait in the render queue for "a few seconds, but it can take longer." That is why a JavaScript-heavy site can rank fine.
Other crawlers are less patient. A December 2024 study by Vercel and MERJ, based on traffic to nextjs.org and Vercel's network, found that none of the major AI crawlers it examined rendered JavaScript. GPTBot, OAI-SearchBot, ChatGPT-User and ClaudeBot fetched JavaScript files but did not execute them. Gemini, which uses Googlebot's infrastructure, and Applebot were the exceptions. That study is almost two years old and the authors do not claim it covers every site. Treat it as the baseline to retest, not a standing fact.
For a local business the exposure is specific. Opening hours loaded by a widget, a service list inside a tabbed component, prices in an app-style booking tool, a phone number set by a script. A human sees all of it. A crawler that does not run scripts sees a shell.
Test it: run curl -s https://yourdomain.com/ | grep -i "your phone number" and repeat with a service name and your city. If a fact a customer needs is absent from that raw HTML, assume the simpler crawlers cannot read it either.
Sign 3: Your pages carry nosnippet, max-snippet:0 or noindex
This is the one sign that is a documented eligibility rule rather than a risk. Google's robots meta documentation says nosnippet also "prevents the content from being used as a direct input for AI Overviews and AI Mode," and that max-snippet limits how much of the content may be used. A page can rank on a nosnippet tag and simply never be quoted. noindex removes the page from results altogether.
These tags get added for reasons that were sensible once: a legal worry, a plugin default, a staging setting that shipped to production, an SEO setting someone toggled while chasing a different problem. Nobody revisits them because rankings look normal.
Test it: view the page source and search for robots. Also check the response header, since X-Robots-Tag can carry the same directive where the HTML shows nothing: curl -sI https://yourdomain.com/ | grep -i x-robots.
Sign 4: Your name, address and phone number disagree across the web
An assistant answering "who does this near me" has to decide that several records describe the same business. If your Google Business Profile says one thing, the directories say a second, and your site says a third, that decision gets harder. Suite 4 against #4, an old phone number on a data aggregator, a former trading name.
The strongest published evidence is in the local assistant data covered in Four assistants, four different local businesses. It reports that profile accuracy differs sharply between assistants grounded in Google Maps and assistants reconstructing a record from the open web. That study covers chains with 50 or more locations, not single-site businesses, and we could not re-check its figures against the publisher's own page, so we are pointing to the analysis rather than repeating the numbers.
Test it: write your name, address and phone number exactly as they appear on your site, then compare against your Business Profile and the first five directories that rank for your business name. Differences, even small ones, are the work. Our guide to fixing wrong business info on Google covers the Google side, and Local SEO in Toronto covers the rest.
Sign 5: Nobody outside your own site describes you
A site can be technically perfect and still be a lone voice. AI answers draw on many sources, and an assistant weighing a claim you make about yourself has little to check it against if no review site, directory, local press mention or customer discussion repeats it.
Ahrefs studied 75,000 brands in a report published 12 December 2025 and found branded web mentions correlated with AI visibility at roughly 0.66 to 0.71 depending on the platform, with YouTube mentions strongest at about 0.74. Its text describes link metrics as showing very weak correlations. Ahrefs states the usual caveat itself: correlation isn't causation, and the patterns do not mean improving these metrics will automatically lift AI visibility. A well-known brand attracts both mentions and citations, so the arrow may point either way. The argument is made at more length in SEO is becoming brand engineering.
Test it: search your business name in quotation marks, excluding your own domain, and count the independent pages that describe what you do and where. Then ask ChatGPT or Gemini a plain customer question for your service and city and see who is named. One answer proves little because these answers vary run to run. It does show you what a customer might be told.
The counter-argument
Some will say all of this is ordinary SEO under a new label, and Google's guidance supports part of that: for AI Overviews, it says no special optimization is needed. That is true for Google. It is less true for ChatGPT and Claude, which have their own crawlers and their own opt-out settings that Google's rules do not cover. A site can pass every Google check and fail sign one. A second objection is that these signs are not proof of anything. Correct. Having all five clear does not guarantee a recommendation, and having one present does not guarantee you are missing. They are the conditions worth ruling out before spending money on anything fancier.
What to do first
Start with signs one and three, because a wrong line in robots.txt or a stray nosnippet tag is a ten-minute fix with an immediate effect on what crawlers may do. Sign two comes next, since it can need a developer. Signs four and five are slower and more valuable, and they overlap with work a business should be doing for customers anyway.
If you would rather have someone run the checks, our AI visibility audit covers crawler access, rendering and entity consistency, technical SEO covers the fixes for the first three, and entity and knowledge architecture covers sign four. Nothing here, and no service we sell, can guarantee that an assistant will name you.
