AI Visibility Audit: How to Measure If AI Models Cite You

Most businesses have no idea whether ChatGPT, Perplexity, Claude, or Gemini mention them at all. That’s a problem. When a potential customer asks an AI assistant to recommend a marketing agency in Boston, your company either shows up or it doesn’t. There’s no page-two consolation prize.

An AI visibility audit gives you a clear answer. It’s a structured, repeatable process that tracks whether AI models cite your brand, link to your site, or recommend your competitors instead. We’ve been running these audits at CROSS Digital since early 2025, and the data they produce is consistently surprising, even for companies with strong traditional SEO.

What an AI Visibility Audit Actually Measures

The core method is simple enough to run on a Tuesday afternoon. You create a fixed set of 15 prompts that mirror what your ideal customers would ask an AI assistant. Then you run every prompt across four AI engines: ChatGPT, Perplexity, Claude, and Gemini. Do this monthly, same prompts, same engines, and you’ve got a trend line.

For each response, you’re tracking four things:

• Mentioned by name: Does the AI response include your brand name anywhere in its answer?
• Cited with a link: Does it actually link back to a page on your site?
• Which URL: If cited, which specific page earned the mention? Your homepage, a blog post, a service page?
• Competitor mentions: Who else shows up in the same response? This is your AI-era competitive landscape.

Fifteen prompts across four engines gives you 60 data points per month. That’s enough to spot patterns without drowning in spreadsheet noise.

CROSS’s Baseline: What We Found When We Audited Ourselves

We ran our own audit before recommending this to clients. The results were humbling.

CROSS Digital appeared in just 1 out of 15 prompts across the four AI engines, a 6.7% mention rate. Meanwhile, Brick Marketing, a competitor, appeared in 7 or more mentions across the same prompt set. They weren’t doing anything exotic. Their content was simply structured in ways that AI models could easily parse and cite.

That gap between 6.7% and 46%+ became our roadmap. Every optimization we made in the following months was designed to close it. And tracking the audit monthly let us see exactly which changes moved the needle.

The 6-Factor Citation Scorecard

After analyzing hundreds of AI responses, we’ve identified six factors that determine whether your content gets cited. Use this scorecard to evaluate any page on your site.

Factor	What It Means	Pass Test
Quotable Atoms	Your content contains self-contained statements an AI can extract and cite without needing surrounding context.	Can you pull a single sentence that fully answers a question on its own?
Direct Answers First	The page leads with a clear, concise answer before providing background or elaboration.	Does the first paragraph contain a direct answer, not a preamble?
Entity Richness	Content includes specific names, places, dates, numbers, and named concepts that AI models use for retrieval.	Does the page mention 5+ named entities relevant to its topic?
Original Information	The page offers data, research, frameworks, or insights not available elsewhere on the web.	Could you remove this page and the internet would lose something unique?
Machine-Readable Structure	Proper use of headings (H2/H3), schema markup, tables, and lists that AI crawlers can parse cleanly.	Does the page use FAQ schema, HowTo schema, or structured tables?
Trust Signals	Author bios, citations to sources, HTTPS, domain authority, and consistent NAP for local businesses.	Can a crawler verify who wrote this and why they’re credible?

Score each page on a pass/fail basis for all six factors. Pages that pass four or more tend to show up in AI responses. Pages that pass two or fewer almost never do.

How to Run Your Own AI Visibility Audit (DIY Steps)

You don’t need expensive software to start. Here’s the process we use, stripped to its essentials.

Step 1: Build your prompt list.

Write 15 prompts that your target customers would realistically type into an AI assistant. Mix informational queries (“what’s the best way to…”), comparison queries (“top 5 agencies for…”), and local queries (“who does X in [your city]”). Don’t use your brand name in any prompt. You’re testing whether the AI already knows you.

Step 2: Run each prompt across all four engines.

Open ChatGPT, Perplexity, Claude, and Gemini. Paste each prompt one at a time. Copy the full response into a spreadsheet. This takes about 90 minutes for 15 prompts across four engines.

Step 3: Score every response.

For each of the 60 responses, record: were you mentioned by name (yes/no), were you linked (yes/no), which URL was cited, and which competitors appeared. Calculate your mention rate as a percentage of the 60 total responses.

Step 4: Run the 6-Factor Scorecard on your top 5 pages.

Take the five pages you’d most want AI models to cite. Score them against the six factors above. The gaps you find are your optimization priorities.

Step 5: Repeat monthly.

Same prompts, same engines, same spreadsheet. The month-over-month trend is where the real value lives. One snapshot tells you where you stand. Three months of data tells you whether your efforts are working.

What to Do With Your Results

If your mention rate is below 10%, you’ve got foundational work to do. Start with the complete GEO guide to understand the full optimization framework. If you’re between 10% and 30%, you’re in a good position to make targeted improvements. Our AI search optimization service is built for exactly this stage, identifying the highest-impact changes and implementing them systematically.

Above 30%, you’re ahead of most businesses. At that point, the audit becomes a monitoring tool rather than a diagnostic one. You’re watching for drops, tracking competitor movement, and fine-tuning.

Frequently Asked Questions

Related guides

The Complete Guide to Generative Engine Optimization (GEO)

How to Get Your Business Mentioned in ChatGPT

AI Search Optimization Services