GEO Fundamentals
Is your brand visible
in AI search?
You can find out in ten minutes. Ask the assistants your buyers use what your buyers ask them, and watch whether your name comes up. This is the self-check: the exact prompts to run, how to read what comes back, and where a manual pass stops being enough.
GEO Fundamentals · Last updated
For the marketer who wants a straight answer on where they stand in AI.
To check whether your brand is visible in AI search, ask ChatGPT, Claude, Gemini, and Perplexity the questions your buyers ask, in the phrasings a buyer would use, and note whether you're named, described accurately, or absent. One pass is a smoke test; Resonate Labs runs a fixed query set across all four engines for the full, stable read.
Ask what your buyers ask
Use real buyer phrasings, not your brand name. The point is to see what a buyer sees when they don't already know you.
Named, wrong, or absent
Three outcomes, three different problems. Read which one you're getting before you decide what to fix.
One run is a smoke test
Answers vary between runs and between engines. A single check flags a problem; it doesn't measure one.
Run the discovery prompts
Start where your buyers start, with the question, not your name. The goal is to see what an assistant tells someone who doesn't already know you exist, so type the phrasings a buyer would actually use and leave your brand out of the prompt.
- Category and use case: "best [category] for [use case]," "what tools do [role] use for [job]," "top [category] platforms for a [company size / segment]."
- Competitor-anchored: "top alternatives to [a competitor buyers know]," "who competes with [competitor]," "[competitor] vs other options."
- Shortlist-shaped: "who are the leading [category] vendors," "which [category] company is best for [specific requirement]."
Run each prompt in all four engines that matter for B2B, ChatGPT, Claude, Gemini, and Perplexity, and run each one separately. They draw on different sources: Ahrefs found only about 12% of the sources AI assistants cite overlap with Google's top ten, so roughly 88% of citation opportunities are platform-specific. Being named in one engine tells you little about the others. Platform divergence covers why each engine has to be checked, and optimized, on its own.
Read the results: named, wrong, or absent
Every answer lands in one of three buckets, and they point to different work. Sort what you saw before you react to it.
Named and accurate. You appear in the answer and what's said about you is right. This is the goal. Note which prompts and which engines produce it, because that's your baseline to hold and expand, and check that the framing matches how you'd position yourself, not just that your name is present.
Present but wrong. You're named, but the description is outdated, vague, or simply incorrect, an old category, a stale claim, a feature you dropped. This is a defensive problem, not a visibility one, and more content won't fix it on its own. It means the sources the model reads about you are telling the wrong story. Defensive GEO is how to find and correct it.
Absent. The answer names several vendors and you aren't one of them. Note who is: the competitors filling the shortlist you're missing from are a map of where the citations are going. Absence at the discovery stage is the most expensive outcome, because you're cut before anyone compares you. It's also the most common, and the rest of this cluster, starting with how B2B buyers research vendors with AI, is about closing it.
Why one check isn't a measurement
A manual pass tells you whether you have a problem. It can't tell you how big it is, because AI answers move. Ask the same question twice and the sources often change, even when the gist stays the same.
The volatility is measurable. In Google's AI Overviews, Ahrefs found consecutive answers to the same query stay almost identical in meaning, around 0.95 similarity, while swapping out nearly half their cited sources each time. Across platforms, Profound found that 40 to 60% of the domains cited in AI answers are different from one month to the next. So a single run is a snapshot of a moving target: catch a good day and you'll feel fine while losing most weeks; catch a bad one and you'll overreact to noise.
That's the difference between checking and measuring. Checking is a handful of prompts on one afternoon. Measuring is a fixed set of buyer-intent queries, run across the engines on a recurring cadence and scored against a baseline, so movement becomes signal instead of anecdote. The self-check is the right way to find out you have a problem; a repeatable read is how you size it and track whether it's closing. How we measure GEO results is that read, and the ongoing monitoring checklist is how to run it as a discipline.
If you're absent everywhere, check the plumbing
One result worth isolating: if you're missing from every engine on every prompt, the cause may sit below your content. Many AI crawlers don't run JavaScript, so a site that assembles itself in the browser can serve them a near-empty page, invisible to the model even while it ranks on Google. Before you conclude your content is the problem, confirm the engines can read your site at all. Why a site that ranks on Google can be invisible to AI walks through how to check crawler access and rendering, and the technical GEO checklist is the full self-audit of the foundation underneath.
Where Resonate Labs fits
The self-check is meant to be run without us. When you want the version that's stable enough to act on, that's what Resonate Labs does. A free AI Visibility Snapshot takes the manual pass off your plate and shows how AI engines describe your company today, and the full AI Visibility Crawl runs a fixed set of buyer-intent queries across ChatGPT, Claude, Gemini, and Perplexity, scoring where you're named, cited, or absent so a moving target becomes a number you can track.
From there the work is the same one this page points at: hold the prompts where you're named, correct the ones where you're described wrong, and close the ones where you're absent. The measurement points at exactly which is which, and the fixes are the pages and structure that move them. If you'd rather start with a conversation than a crawl, book a call.
Frequently asked questions
How do I check if my brand shows up in AI search?
Ask the assistants what your buyers ask them. In ChatGPT, Claude, Gemini, and Perplexity, run the phrasings a buyer would use, like "best [category] for [use case]," "top alternatives to [competitor]," and "who are the leading [category] vendors for [segment]." For each answer, note three things: whether you're named at all, whether what's said about you is accurate, and which competitors appear when you don't. Run each engine separately, because they draw on different sources and being named in one doesn't mean being named in another.
Why do I get different answers every time I ask the same question?
Because AI answers are volatile in their sources even when their meaning is stable. Ahrefs found that consecutive Google AI Overviews for the same query stay almost identical in meaning (about 0.95 similarity) while swapping out nearly half their cited sources each time. Across platforms, Profound found that 40 to 60% of the domains cited in AI answers change from one month to the next. That's why a single run is a smoke test, not a measurement: it tells you whether you have a problem, not the full or stable picture. A fixed query set, run repeatedly across the engines, is what turns a lucky or unlucky sample into a trend.
AI describes my company inaccurately. What should I do?
Being present but wrong is a different problem from being absent, and it needs defensive work rather than more visibility. Models assemble what they say about you from the sources they can read, so outdated positioning, a stale third-party profile, or thin first-party content can all surface as a confident but incorrect answer. The fix is to find where the wrong information comes from and publish clear, current, first-party content that gives the model a better source to pull from. Defensive GEO covers how to audit and close those gaps.
Is a manual check enough, or do I need a tool or a partner?
A manual check is the right first move, and often enough to confirm you have a problem. It falls short as a measurement because answers vary between runs and between engines, so a handful of prompts can't tell you how often you're absent or whether you're improving. A real read runs a fixed set of buyer-intent queries across ChatGPT, Claude, Gemini, and Perplexity on a recurring basis. A free AI Visibility Snapshot from Resonate Labs is the fast version of that first look, and the full AI Visibility Crawl is the complete, scored read across every engine.
Next step
Run the check without running it yourself.
You can run the prompts by hand, or start with a free AI Visibility Snapshot and get the read back done for you. The full AI Visibility Crawl turns the manual pass into a stable, scored picture across every engine:
- The questions your buyers ask AI, and whether you're named
- Where you're visible, cited, or absent across the four engines
- What the first 30 days would move