BREAKING The machines have opinions about your brand Ask ChatGPT, Claude, Gemini the same question - get three brands back NEW AI Visibility Score, 0-100, tracks how often you appear Hallucinations = confident lies told to every buyer Missing from the answer is the new page two BREAKING The machines have opinions about your brand Ask ChatGPT, Claude, Gemini the same question - get three brands back NEW AI Visibility Score, 0-100, tracks how often you appear Hallucinations = confident lies told to every buyer Missing from the answer is the new page two
The Field Guide

The AI Visibility Audit
(Step-by-Step)

Everyone is racing to optimize for AI. Almost no one has checked what AI already says about them. That is where this starts.

Abstract representation of AI visibility auditing across generative engines
The audit turns vague AI anxiety into a ranked list of fixes.
Share LinkedIn Twitter / X Facebook Instagram
What it is

Your brand has a second reputation. You didn't write it.

For twenty years the question was simple: where do you rank on Google? Now a buyer opens ChatGPT, Claude, or Gemini and types a question in plain language. The engine answers in a paragraph. It names some brands and skips others. It states facts with total confidence, and sometimes those facts are wrong. The AI Visibility Audit is the practice of reading that answer before your customer does - and doing something about it.

The method is deliberately unglamorous. You write a set of questions the way real buyers ask them. You run each question across every major engine. You record whether your brand shows up, where it sits, who appears instead of you, and which sources the machine trusted to build its reply. Then you sort the gaps by how much they cost you. The mental model is closer to financial reporting than to a product launch: measure, benchmark, act, repeat.

4
Engines that matter: ChatGPT, Claude, Gemini, Perplexity
15-30
Prompts in a working library
0-100
The AI Visibility Score range
90
Days between re-runs as models shift
Ask three chatbots the same question and you get three brands back. The gaps between them are the most honest market research you'll ever read. The audit, in one sentence
The Method

Six steps, one afternoon

None of this requires a data team. It requires patience and honesty about what you find.

STEP 01

Write the questions buyers actually ask

Build 15 to 30 prompts across four intents: branded (your company and products), category ("best tools for X"), problem-solution (a pain point, no brand named), and comparison (you versus a named rival). Pull the phrasing from real sales calls and support tickets, not a brainstorm.

STEP 02

Run every prompt across every engine

Put the same question to ChatGPT, Claude, and Gemini - add Perplexity if your buyers use it. For each answer, record four things: do you appear at all, where in the reply, which competitors take your place, and which sources the engine cited.

STEP 03

Measure consistency

Line the answers up side by side. A brand that shows up in ChatGPT and vanishes in Gemini has a distribution problem, not a messaging one. Track appearance rate and position over time, and re-run on a schedule so you see the trend, not a snapshot.

STEP 04

Hunt for hallucinations

Read each answer against what you know is true. Flag invented features, wrong founding dates, stale pricing, capabilities you don't offer. These are the highest-priority items: a confident error is repeated identically to every buyer who asks.

STEP 05

Find the missing knowledge and use cases

Note where the engine is simply thin - a product it can't describe, a use case where a competitor owns the answer. Every use case the AI can't name is a recommendation you'll never receive. This gap list is your content brief.

STEP 06

Prioritize the fixes

Rank by reach and risk. Correct hallucinations first, then fill missing product knowledge and contested use cases, then strengthen the high-authority sources the engines cite - so the true, corrected story is the one they learn next time.

Know your engines

Four machines, four foundations

They disagree because they're built on different ground. Understanding the source explains the answer.

ChatGPT
Leans on Bing for live retrieval
Gemini
Anchored to Google's Knowledge Graph
Perplexity
Live web crawling, citation-first
Claude
Training-data consensus
Why now

The scoreboard changed and most people didn't look up

Share of voice used to be an advertising metric. In the age of answer engines it is closer to a survival metric, because the reply a buyer reads is often the only research they do. If the model never names you, you were never in the running - and you won't see a rejection email to tell you so.

This is why hallucination detection now sits at the center of the practice. A wrong ranking on a search page is annoying. A wrong fact stated with authority inside an answer is a small, repeatable act of misinformation about your company, delivered on request, at scale. Catching it early is brand safety, not vanity.

A hallucination about your brand is a confident lie told to every buyer who asks. Why hallucinations rank first

The good news is that the fix loop is short. Because the engines learn from the open web and the sources they cite, improving the record - clearer product pages, stronger third-party mentions, accurate structured facts - eventually changes the answer. The audit tells you exactly which record to fix, and in what order.

Questions

The five most common ones

What is an AI Visibility Audit?

A structured check that runs realistic buyer-intent prompts across ChatGPT, Claude, and Gemini, records whether and where your brand appears, compares you to competitors, and flags any inaccurate or missing information.

Which questions should I ask the engines?

Build 15 to 30 prompts that mirror real buyers: branded questions about your company and products, category questions like "best tools for X", problem-solution prompts with no brand named, and head-to-head comparisons with named competitors.

How do I measure consistency?

Run the same prompts across each engine and compare. Track whether you appear, where you sit in the answer, and whether the facts match. Repeat on a schedule - quarterly or more often - to see the trend rather than a single snapshot.

How do I spot hallucinations?

Read each answer against known facts about your brand. Flag invented features, wrong founding dates, outdated pricing, or capabilities you don't offer. These confident errors are the highest-priority fixes because every buyer sees the same wrong answer.

How do I prioritize improvements?

Rank gaps by reach and risk. Fix hallucinations first, then fill missing product knowledge and use cases where competitors dominate, then strengthen the high-authority sources the engines cite - so the corrected story is the one they learn.

Read further

How to Measure Brand Visibility in ChatGPT, Gemini, Perplexity & Claude - Sanbi A 10-Step Framework for Generative Engine Optimization - Profound GEO Audit: Methodology to Measure Visibility on Generative AI - AI Labs Audit How to Audit Your Brand's AI Visibility (10-Step Guide) - LLM Pulse 7 Steps for Tracking Your ChatGPT Visibility - Ahrefs ChatGPT, Gemini, Claude Brand Mentions: Tracking Guide - MaxAEO