How ADMATIC Tracks AI Visibility

    November 22, 2026|
    AI
    SEO-AEO-GEO
    Author:

    TL;DR

    • A single screenshot of a ChatGPT answer isn't AI visibility reporting. If your business name was in the question, a mention was guaranteed.
    • Track the questions customers actually ask, sourced from Google Ads search terms, Search Console and People Also Ask, with your brand name left out.
    • Weight each set of about ten questions towards the choosing stage: roughly six choosing, two comparing and two deciding.
    • Run the same questions across customer types and AI assistants every month. The average is what's reliable.
    • Being recommended is the goal. For one ADMATIC client, a key differentiator earning zero AI mentions became the top content priority.

    A lot of AI visibility reporting still starts with a screenshot. Someone asked ChatGPT about the business, got a mention, and passed it up the line. It's one answer on one day, and if the business name was in the question, a mention was guaranteed. Here's how ADMATIC tracks AI visibility instead, and how the results turn into a plan.

    Ask the Questions Your Customers Are Asking

    Start with how your customers actually ask, backed by research. The longer, more specific questions people used to type into Google appear to be moving to AI assistants. That makes your search data the best guide to their wording.

    • Google Ads search terms, Search Console queries and People Also Ask show the words and situations buyers use.
    • Rewrite those as full questions with the buyer's situation in them.
    • Weight them towards the moment someone chooses a provider.
    • Keep your own business name out of every question.

    The Types of Questions to Track for AI Visibility

    AI answers each stage of the buying journey differently, so ADMATIC groups questions by where the buyer is.

    Researching

    “What does boat insurance usually cover?” AI answers these with information and rarely names a business. We usually keep them in the content plan, not the tracking list.

    Deciding Whether They Need It

    “Do I need separate boat insurance, or does my home and contents policy cover my boat?” These show whether AI points buyers towards your kind of product or service, or towards something else.

    Comparing Options

    “Is a specialist marine insurer better than a general insurer for boat insurance?” These show whether AI puts you on the shortlist.

    Choosing a Provider

    “Who's the best insurer for a six metre trailer boat used for fishing?” This is where recommendations turn into enquiries.

    The Question Mix per Customer Type

    • About ten questions per customer type.
    • Roughly six at the choosing stage, two comparing and two deciding.
    • Add more deciding questions where buyers lean towards an alternative, such as cover they already have.
    • Add more comparing questions where competitors get searched by name.

    Measure AI Visibility Over Time, Not in One Answer

    AI can give a different answer to the same question on a different day, so a single result tells you very little. What works is a system: several types of customer, around ten questions each, asked in each assistant your buyers use, every month, for a long time. The average across all of those answers is what's reliable, because it cuts through the day-to-day noise.

    In our own client work, we often average five customer types, so maybe around 50 questions are tracked every month. The assistants don't behave alike. With one specialist insurance client, Perplexity missed the brand entirely on its motorcycle and prestige vehicle questions. Tracking one assistant alone would have given a misleading picture.

    To keep that system honest:

    • The first run is a pilot.
    • After that, the question set is locked and only changed at a quarterly review.
    • Questions are never run from someone's own logged-in account, where the assistant's memory can shape the answer.
    • When anything in the test changes, the report says so.

    Why Recommendations Are the Goal of AI Visibility

    Being mentioned and being cited are steps along the way. The goal is being recommended: AI telling a buyer to choose you, ahead of your competitors. Tracking the answers over time shows you:

    • who is being recommended for each type of customer
    • why the assistant recommends them
    • what content they've published that earns the recommendation
    • which sites the assistants cite when they recommend them

    Turning the Answers Into a Plan

    Those answers become the plan. They tell you what content you need to produce, where you need to be cited, and what on your website is technically holding you back.

    For that same insurance client, one of its genuine differentiators, agreed value cover, earned zero mentions across ChatGPT, Gemini and Perplexity. That single finding moved straight to the top of the content plan. The technical side surfaced too: the site's Cloudflare settings were blocking AI and search crawlers, which no amount of new content would have fixed.

    The Signals Around AI Visibility

    Branded Search in Search Console

    When AI puts your name in front of people, some go on to search for you on Google. A rising trend in searches for your name in Search Console is a sign it may be working. Search Console now also shows how often links to your site appear in Google's AI Overviews and AI Mode, once your site appears there often enough, though not how many clicks they send.

    GA4 Is Only a Small Part of the Picture

    Clicks from Google's AI Overviews are counted as ordinary organic traffic. Someone who hears your name from ChatGPT and searches for it a fortnight later shows up as a branded search, not an AI visit. Add “an AI assistant like ChatGPT” to your “how did you hear about us?” options to catch some of what GA4 misses.

    Server Logs, With a Caveat

    If you can see your server logs, you can watch AI bots visiting your site. Treat them as a variable signal. OpenAI, for example, runs separate bots for training its models, for its search index, and for fetching a page when someone asks ChatGPT a question. Visits triggered by a user's question are closest to your page being used in an answer, and even those don't confirm a citation.

    Start Tracking Your AI Visibility With Ten Questions

    Pick your most valuable type of buyer and write ten questions they'd actually ask. Run them in ChatGPT, Perplexity, Claude and Gemini in fresh chats while logged out. Note who gets recommended, who gets named first, and which sites get cited, then run the same questions next month.

    If you'd like a second pair of eyes, ADMATIC offers a free first look at your business's AI visibility.

    If you're working out how to measure your own AI visibility, reach out to an ADMATICian. We're always up for that conversation.

    Frequently Asked Questions

    How should a business track its AI visibility?

    Track AI visibility with a fixed set of questions customers actually ask, across several customer types and the AI assistants buyers use. Run them monthly in fresh, logged-out chats and record who gets recommended and which sites get cited. A single screenshot reflects one answer on one day. The average across many answers over time is what's reliable.

    Where do the questions for AI visibility tracking come from?

    The best sources are Google Ads search terms, Google Search Console queries and People Also Ask. These show the words and situations buyers use. Each one is rewritten as a full question that includes the buyer's situation. The business's own name is left out, because naming it guarantees a mention.

    What types of questions should be tracked for AI visibility?

    Questions fall into four stages: researching, deciding whether they need it, comparing options and choosing a provider. Researching questions rarely name a business, so they suit the content plan instead. ADMATIC's starting mix is about ten questions per customer type. Roughly six are choosing questions, two comparing and two deciding.

    Why is being recommended by AI more important than being mentioned?

    A mention or citation shows an AI assistant knows a business exists. A recommendation is the assistant telling a buyer to choose that business ahead of competitors. Recommendations at the choosing stage are where AI answers turn into enquiries. Tracking who gets recommended, and why, shows which content and citations to pursue.

    Can GA4 measure AI visibility?

    GA4 captures only a small part of the impact of AI visibility. Clicks from Google's AI Overviews are counted as ordinary organic traffic. Someone who hears about a business in ChatGPT and searches for it later appears as branded search. Adding an AI assistant option to “how did you hear about us?” forms helps fill the gap.

    Share this article: