How to run an AI visibility audit for your B2B brand in one afternoon

An AI visibility audit is a systematic process of querying generative search engines with a structured set of buyer prompts to measure how often your brand is recommended, cited, or omitted from AI-generated answers. While comprehensive tracking requires enterprise tools, any marketing team can run a manual baseline audit in a single afternoon to uncover where they stand in the synthetic search landscape.

The purpose of a manual AI visibility audit

Traditional SEO tools rely on search volume and keyword rankings to estimate organic traffic. However, generative search engines operate on different principles. They do not merely index pages: they synthesize concepts, compare solutions, and recommend specific vendors based on training data and real-time retrieval systems. To understand your performance in these systems, you must measure your presence directly within their outputs.

Conducting a manual AI visibility audit allows you to see exactly what your prospects see when they ask AI for recommendations. This process introduces you to the core mechanics of generative engine optimization, giving your team a practical understanding of how large language models perceive your brand. By running this diagnostic, you establish a baseline that helps prioritize future content and technical optimizations.

Step 1: Designing your diagnostic prompt set

A reliable audit requires a standardized set of prompts. If you query engines haphazardly, you will collect inconsistent data. To build a diagnostic set, you need to replicate the actual query patterns of your target buyers across three main categories.

First, define 5 to 10 informational prompts. These are conceptual questions your buyers ask when trying to solve a problem. For example, if you sell cybersecurity software, an informational prompt might be: "What are the best practices for securing APIs in a multi-cloud environment?" The goal here is to see if the engines cite your educational content as a source.

Second, define 5 to 10 comparative prompts. These are queries where buyers weigh different solutions. An example is: "Compare the top API security platforms for enterprise companies." This helps you evaluate whether the engine includes your brand in competitive roundups and how it positions your strengths against competitors.

Third, define 5 to 10 recommendation prompts. These are direct, high-intent queries such as: "Which API security tool is best for a team using AWS and Kubernetes?" These prompts measure your direct recommendations in highly specific buyer scenarios.

Write these prompts down in a spreadsheet. You will use this exact same list across every engine you test to ensure consistency.

Step 2: Sampling across the five major engines

Once your prompt set is ready, you must test it across the platforms that command the largest user bases. Do not limit your test to a single model, as each system utilizes different retrieval mechanisms and training data.

Open separate, clean browser sessions for each of the following platforms:

• ChatGPT (GPT-4o): The current market leader for general queries.
• Claude (Claude 3.5 Sonnet): Known for nuanced synthesis and deep analysis.
• Gemini (Gemini 1.5 Pro): Heavily integrated with Google Search and real-time web data.
• Perplexity: A dedicated conversational answer engine that relies heavily on real-time citations.
• Microsoft Copilot: Highly integrated with Bing and enterprise workflows.

For each prompt in your spreadsheet, paste the query into each engine. It is critical to use fresh, incognito sessions or clear the chat history between queries when testing different topics. This prevents the conversation history from biasing subsequent answers, ensuring you receive a neutral response every time.

As you run the queries, copy the generated text and the cited URLs into your spreadsheet next to the corresponding prompt and engine column. This raw data forms the foundation of your analysis.

Step 3: Calculating your answer share and analyzing sentiment

With your data collected, you can now calculate your baseline performance metrics. The most critical metric to establish is your answer share, which measures the percentage of times your brand is recommended or cited out of the total possible opportunities across your prompt set.

To calculate your answer share for recommendation queries, count the number of times your brand is mentioned as a recommended solution, then divide that by the total number of recommendation queries multiplied by the number of engines tested. For example, if you ran 10 recommendation prompts across 5 engines (50 total queries) and your brand was recommended in 10 of those responses, your answer share for recommendations is 20 percent.

Next, analyze the citations. Note which domains are being cited as sources when the engines recommend your brand or your competitors. If the engines are constantly citing third-party review sites, industry publications, or specific competitor blogs, those domains are the sources of truth for the LLMs. This is a key difference to note when reviewing a GEO vs SEO comparison: instead of optimizing solely for search engine crawlers, you must ensure your brand is highly visible on the external sites that feed these AI models.

Finally, evaluate the sentiment and accuracy of the mentions. Does the AI accurately describe your product features? Does it categorize you correctly? If an engine recommends you but claims you lack a feature that you actually offer, you have identified a critical data gap that must be corrected through targeted content updates.

Scaling your AI visibility strategy

A manual afternoon audit provides a valuable snapshot of your current brand perception in AI engines. It exposes obvious gaps, reveals which competitors are favored by specific models, and helps your team understand the reality of generative search.

However, manual audits have clear limitations. AI models update their weights regularly, real-time search indexes change daily, and user locations alter search results. A single afternoon of manual querying cannot track these fluctuations over time or scale across hundreds of commercial search terms.

To build a repeatable, data-driven marketing strategy, brands eventually require automated tracking and deeper analysis. If you want to move beyond manual spreadsheets and receive a comprehensive, data-backed evaluation of your brand's footprint, you can request a professional AI visibility audit. This deeper analysis provides the precise data points and optimization roadmaps needed to systematically grow your answer share across all major LLMs.