The buying journey now starts with an AI prompt
A few years ago, a marketing leader evaluating a new tool would type a query into Google, scan the top results, click two or three blue links, and decide. The funnel started with a search engine results page. Marketing teams optimised for it because that's where attention lived.
In 2026, a growing share of that same evaluation starts somewhere else. The buyer opens ChatGPT or Claude and types a question — "what's the best tool for X?", "which platform should I pick between A and B?", "is there a cheaper alternative to Y?". The AI assistant answers in plain language, names three or four products, and explains why each one might fit. The buyer reads the answer and moves to the next step — often a vendor's website, often skipping the search results page entirely.
That shift matters because the AI's answer is the new shortlist. If your brand is named, you're in the consideration set. If you're not named, the buyer may never know you exist for that question. There's no second-page recovery; there's just the answer.
Why this is different from search results
A search engine returns ten links and lets the reader judge. An AI assistant returns one synthesised answer and tells the reader what to think. The same query can produce very different visibility outcomes between the two.
A brand that ranks well in Google but is rarely mentioned by ChatGPT loses ground in AI-assisted buying without that loss showing up in any traditional analytics tool. Web traffic looks fine. Brand searches look fine. But the AI shortlist quietly excludes you, and you find out only when win rates start to slip.
This is why "AI visibility" is its own discipline. It's measured against the answers AI assistants actually give, not against the pages search crawlers index. The data sources, the levers, and the measurements are different — even when the strategic goal (be where buyers look) is the same.
How to find out where your brand stands
There are three rough levels of effort to answer the question "is my brand visible?", from cheapest to most defensible.
First, ask the AIs yourself. Open ChatGPT and Claude. Type the questions a buyer in your category would type. Read the answers. Note whether your brand appears, whether it's described accurately, and which competitors come up instead. This takes thirty minutes and gives you a feel for the answer. It also varies — ask the same question tomorrow and the wording may shift — so a one-off check is directional, not defensible.
Second, run the same questions multiple times across multiple providers. AI models give slightly different answers run to run, so a single ask isn't enough to decide whether you're systematically absent or just unlucky on one query. A defensible reading needs repeated runs on at least two providers, and a clear rule for when a brand counts as "mentioned".
Third, do that on a fixed question set so you can compare across time. The questions matter; if the set changes, the score moves for reasons that have nothing to do with your visibility. Lock the questions, run them on a schedule, and the number tells you something real. A paid analysis runs up to thirty questions on ChatGPT and Claude, three times per provider — up to 180 AI calls per report. A brand counts as mentioned for a question when it surfaces in at least two of the three runs. That threshold is the part most one-off checks miss.
Three questions to ask ChatGPT about your brand right now
If you want to take a quick reading before reading further, paste these three prompts into ChatGPT or Claude. Replace the bracketed text with your category and your brand.
1. "What are the best [your category] tools in 2026?" — Does your brand appear in the list? Where in the list? How are you described?
2. "[Your brand] vs [closest competitor] — which one should I pick?" — Does the AI know your product exists? Does it describe your positioning accurately or generically? What does it say is the difference?
3. "Affordable alternatives to [your closest competitor]" — Does your brand surface as an alternative? If not, who does, and why?
Run each question twice. Note the differences between runs. That variability is itself useful data — it tells you how stable the AI's view of your category is on a given week.
What to do next
If the three quick prompts above made you uncomfortable, you're not alone — most marketing leaders we talk to discover at least one question where they expected to surface and didn't.
The next step is to make the reading defensible. A one-time check on three questions is a directional signal; a fixed thirty-question set, run repeatedly across two providers, is something you can take to a leadership meeting and act on.
You can run a free Lumialo analysis on ten questions today, or read the methodology page first to see exactly how the score is computed before you commit. Either way, the goal is the same — turn the question "am I visible?" into a number you can track and improve.