Written by a search visibility strategist with 7+ years of hands-on experience running AI visibility audits across dozens of client accounts. This guide explains the actual testing methodology used in real audits, not a simplified version meant to sell a tool.
Most business owners find out whether ChatGPT knows their brand by accident — a customer mentions it, or someone tries typing a question out of curiosity and gets an underwhelming answer. That’s not a real test, and it’s not how LLM experts actually approach the question. Testing brand visibility across AI platforms requires a structured, repeatable methodology, because a single prompt typed once tells you almost nothing reliable. This guide walks through exactly how LLM experts test whether a business is genuinely known and recommended by ChatGPT and Claude, and how you can apply the same thinking to check your own visibility.
Why a Single Prompt Isn’t a Real Test
The most common mistake business owners make is typing “What is [my brand]?” into ChatGPT, getting a reasonable answer, and concluding they’re covered. This approach has two serious flaws. First, AI responses are non-deterministic — the same prompt can produce noticeably different answers across separate sessions, so a single result is closer to a snapshot than a measurement. Second, and more importantly, asking an AI model to describe a brand you already named tells you nothing about whether that brand gets recommended to someone who doesn’t know it exists yet, which is the scenario that actually drives new customer discovery.
LLM experts test differently: with repeated, varied, unbranded prompts that mirror how real potential customers actually ask AI assistants for recommendations.
The Core Testing Methodology LLM Experts Use
1. Building a Balanced Prompt Set, Not a Single Question
A credible audit starts with a structured set of prompts, typically split across a few categories: direct brand-recognition questions (“What is [brand]?”), unbranded category questions (“What’s the best [category] for [use case]?”), and comparison questions (“[Brand] vs. [competitor]”). Unbranded, category-level prompts matter most, since they reflect how a new customer would discover a business without already knowing its name.
2. Running Each Prompt Multiple Times
Because AI responses vary between sessions, LLM experts run each prompt several times in separate sessions rather than treating a single response as conclusive. A brand that appears in three out of five runs of the same question tells a very different story than one that appears in zero or five.
3. Testing Across Multiple AI Platforms Separately
ChatGPT, Claude, Gemini, and Perplexity retrieve and cite information differently, so a brand visible in one platform can be completely absent from another. LLM experts run the same prompt set across each platform independently rather than assuming results from one represent visibility everywhere. Industry guidance from Ahrefs’ research on tracking ChatGPT visibility reinforces this point directly, noting that popularity-weighted, prompt-level tracking across platforms reveals visibility gaps that a single spot-check would miss entirely. You can read more in Ahrefs’ guide to tracking ChatGPT visibility, which covers this methodology in depth.
4. Scoring Responses Against Consistent Metrics
Rather than eyeballing whether a brand “seems” visible, a proper audit scores each response against a small, consistent set of metrics — whether the brand is mentioned at all, where it ranks relative to competitors mentioned in the same answer, the sentiment or framing of the mention, and whether a source or citation is provided alongside it.
5. Tracking Citation Sources, Not Just Mentions
When an AI model does mention a brand, LLM experts examine which sources it’s pulling from — a review platform, a comparison article, the brand’s own website, or a news mention. This reveals which third-party sources are shaping AI perception of the brand, which becomes a direct action item: strengthening presence on the sources that already influence the answer.
6. Benchmarking Against Named Competitors
Visibility only means something in context. LLM experts typically test the same prompt set against two or three direct competitors, producing a genuine comparison rather than an isolated data point. Finding that a brand appears in a fraction of the mentions a competitor receives is a far more actionable insight than a single “yes, it mentioned you” result.
7. Re-Testing on a Regular Cadence
AI models update frequently, and citation patterns shift as they do. A one-time audit provides a useful baseline, but LLM experts re-run the same prompt set on a recurring schedule to track whether visibility is improving, holding steady, or declining over time.
What a Typical AI Visibility Audit Reveals
- Whether a brand is mentioned at all for relevant unbranded category queries
- How consistently the brand appears across repeated runs of the same prompt
- How visibility compares to two or three named competitors
- Which platforms show strong visibility versus which show none
- Which third-party sources are most frequently cited alongside or instead of the brand
- Whether existing mentions are accurate, outdated, or need correction
A Simple Version You Can Try Yourself
While a full professional audit uses a larger, scored prompt set across multiple platforms, you can get a useful first read on your own:
- Write 5 to 10 unbranded, category-level questions a potential customer might ask (“best [category] in [city/niche],” “[category] for [specific use case]”).
- Run each question in ChatGPT and Claude separately, 2 to 3 times each, in fresh sessions.
- Note whether your brand appears, where it ranks relative to any competitors mentioned, and what source (if any) is cited.
- Repeat the same test with your two closest competitors’ names swapped in for comparison.
This won’t match the depth of a full audit, but it will quickly tell you whether you’re starting from zero visibility or have a foundation to build on.
Frequently Asked Questions
How do LLM experts actually check if ChatGPT knows my brand?
They run a structured, repeated set of prompts — including unbranded, category-level questions — across multiple sessions and AI platforms, then score the responses for mention frequency, ranking relative to competitors, and citation sources, rather than relying on a single question and answer.
Why does asking ChatGPT “What is [my brand]?” not tell me much?
That question only reveals how the AI describes a brand once it’s already been named. It doesn’t show whether the brand gets recommended to someone who doesn’t know it exists, which is the scenario that actually matters for new customer discovery.
Why do I need to run the same prompt multiple times?
AI responses are non-deterministic, meaning the same question can produce different answers across separate sessions. Running each prompt several times reveals a consistency pattern rather than a single, potentially misleading snapshot.
Is being visible in ChatGPT enough, or do I need to check other platforms too? Different AI platforms retrieve and cite information differently, so visibility in ChatGPT doesn’t guarantee visibility in Claude, Gemini, or Perplexity. A thorough audit tests each platform separately.
How often should I re-check my brand’s AI visibility?
AI models and citation patterns change frequently, so a recurring cadence — commonly monthly or quarterly — gives a more reliable picture of whether visibility is improving than a single one-time check.
What should I do if the audit shows my brand isn’t mentioned at all?
Look at which sources the AI is citing for competitors in the same category, and focus on strengthening your presence there, alongside improving your own site’s structured data, content clarity, and entity consistency.
Knowing whether ChatGPT and Claude actually know your brand isn’t something you can determine from a single curious search. It requires the same discipline LLM experts apply in a real audit: a structured, repeated, cross-platform prompt set, scored consistently and benchmarked against competitors. Whether you run a simplified version yourself or bring in an expert for a full audit, the goal is the same — replacing a guess about your AI visibility with an actual, measurable answer you can act on.