How to Audit Brand Visibility on LLMs (And Why Most Businesses Have No Idea What AI Is Saying About Them)

TL;DR: AI platforms like ChatGPT, Perplexity, Gemini, and Claude are becoming primary discovery channels for buyers. If you have never tested what these tools say about your business, you are operating blind. This post walks you through a practical, repeatable audit process to find out whether your brand is showing up, how it is being described, and what to do when the answer is not what you hoped. The steps are manual, require no special tools, and take less than two hours the first time.

Sure you can buy some fancy software at $200/month, but this process will get you there too. 🙂 I have personally tried some and they do a good job, but realistically, I'm only checking this every other month or so.

There is a version of your business living inside ChatGPT right now. You did not write it. You did not approve it. And there is a reasonable chance it is wrong! When this AI thing all started, I have found multiple instances where AI is saying we provide something we just don't. That drove me nuts for 6 months. So I had to figure out how to

Maybe it positions you as a budget option when you are not. Maybe it names three competitors and leaves you out entirely. Maybe it describes a service you no longer offer. Or maybe, and this is more common than anyone wants to admit, it says nothing about you at all.

Most business owners have spent years thinking about how they appear on Google. They track rankings, check their map listing, read their reviews. That habit made sense because Google was the front door. It still matters. But the front door is getting a second entrance, and a growing number of customers are walking in through that one instead.

AI-sourced traffic grew 527% year over year between January and May 2025, according to Previsible's State of AI Discovery Report analyzing nearly two million LLM-driven sessions. McKinsey projects that by 2028, $750 billion in US revenue will flow through AI-powered search channels. And their research found that use of AI-powered search now spans all age groups, including a majority of baby boomers. This is not a young-person trend. It is a behavior shift happening across your entire customer base.

The businesses that figure this out early will not need to catch up later. The ones who wait are going to spend a lot of money trying to reclaim visibility they could have built for far less.

What "Brand Visibility on LLMs" Actually Means

This is not about whether ChatGPT has heard of you. It is about how you appear when a potential customer asks a question you should be answering.

When someone asks Perplexity "what is the best pediatric dentist in [your city]" or "who should I hire to run Google Ads for my restaurant," those platforms generate a response. They do not rank a list of links. They compose an answer, often naming specific businesses, describing their positioning, and sometimes citing sources. Your brand is either in that answer or it is not. And if it is in the answer, what it says about you may or may not reflect reality.

• Brand visibility on LLMs has four distinct dimensions, and an audit needs to measure all of them.
• Presence is whether your brand appears at all when relevant queries are run.
• Accuracy is whether the information the model surfaces about you is factually correct.
• Positioning is whether the model's description of your business matches how you want to be known.
• Competitive context is how you compare to the brands that appear alongside you.

Most businesses, if they have thought about this at all, have only checked presence. They typed their name into ChatGPT once, saw something, and moved on. That is not an audit. That is a glance.

Why AI Gets Your Brand Wrong (and Why That Is Your Problem to Fix)

LLMs do not pull information from a live database of truth. They synthesize from what they were trained on, what they can access via browsing, and what signals the broader web has established about your brand. If your digital presence is thin, inconsistent, or poorly structured, the model fills gaps with whatever it can find. Sometimes that is reasonable. Often it is not.

A 2024 research framework called GEO (Generative Engine Optimization), developed by researchers at Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi, established that visibility-improvement strategies can produce up to 40% lift in how frequently brands appear in AI-generated responses. The same research found substantial variance across platforms in how different models handle the same queries and sources. That means your visibility is not uniform. You may be appearing accurately in one tool and missing entirely in another.

Research analyzing citation patterns found that brand search volume is the number one predictor of whether an LLM cites your brand, with a 0.334 correlation. The businesses that show up most often in AI answers are not necessarily the best. They are the ones with the strongest and most consistent digital signals. That is a structural problem you can fix with the right approach.

The Audit Workflow: How to Do This Yourself

This process requires no paid tools. You need a spreadsheet, access to the four major platforms, and about 90 minutes the first time through. After that, a monthly refresh takes about 30 minutes.

Step 1: Set Up Your Tracking Spreadsheet

Before you run a single prompt, create a simple log. This is where most people skip ahead and then regret it because they have no baseline to compare against later.

Your spreadsheet should have the following columns:

• Date
• Platform (ChatGPT, Gemini, Perplexity, Claude)
• Prompt used
• Did your brand appear? (Yes / No)
• How was your brand described?
• Competitors mentioned alongside you
• Sources cited (Perplexity in particular surfaces these)
• Accuracy issues noted
• Positioning issues noted

Create one row per prompt per platform. If you run 10 prompts across 4 platforms, you will have 40 rows after your first session. That is your baseline.

Step 2: Build Your Prompt List

This is where specificity matters. Generic prompts give you generic information. You want to test the actual queries your potential customers are running.

Build your prompt list in three categories.

Brand-direct prompts test what each model knows about you specifically:

• "What do you know about [your business name]?"
• "What services does [your business name] offer?"
• "Is [your business name] a good option for [your core service]?"

Category prompts test whether you appear when someone searches your service area:

• "What are the best [your service type] in [your city]?"
• "Who should I hire for [your core service] in [your region]?"
• "What companies do [your specialty] well?"

Comparison and evaluation prompts test whether you show up when buyers are in decision mode:

• "How does [your business name] compare to [competitor name]?"
• "What should I look for when hiring a [your type of business]?"
• "Which [your service category] companies are worth the money?"

You are looking for 10 to 15 prompts in total. This gives you enough data without making the process unmanageable.

Step 3: Run the Prompts Across All Four Platforms

Open ChatGPT, Perplexity, Gemini, and Claude. Run each prompt in a fresh conversation on each platform. Do not continue conversations between prompts because context from earlier answers can influence later ones.

A few platform-specific notes worth knowing:

For ChatGPT, run brand-direct prompts both with and without web browsing enabled. The version without browsing shows you what the model's base training contains about your brand. The version with browsing shows what it can find live. The difference between those two responses tells you something important.

For Perplexity, pay close attention to the source URLs cited. Those URLs tell you exactly which content the model is pulling from when it talks about your industry and your competitors. If your site is not among those sources, that is a content and authority problem you now have a specific diagnosis for.

For Gemini, note whether Google Business Profile data appears to be influencing responses. Google's AI surfaces local business information in ways other platforms do not, and your map listing, reviews, and business description all feed into what it says.

Step 4: Document Everything Immediately

Log your results in real time, not from memory afterward. The specific language matters. If ChatGPT describes you as "a full-service digital marketing agency focused on small businesses" but you specialize in healthcare and e-commerce growth, that distinction is meaningful and worth recording exactly.

When you note accuracy issues, be specific. Wrong service descriptions, outdated pricing references, incorrect founding dates, misattributed locations, these are all fixable if you document them clearly.

When you note positioning issues, compare the model's language to your actual positioning statement. If there is a gap, it usually means one of three things: your website does not articulate your positioning clearly enough, there is not enough third-party content reinforcing it, or competitors have simply done more to establish their positioning in sources the models trust.

Step 5: Calculate Your Presence Rate

Once you have run all your prompts, calculate a simple baseline metric.

Take the total number of times your brand appeared across all prompts and all platforms. Divide that by the total number of prompt-platform combinations you tested. If you ran 15 prompts across 4 platforms, your total combinations are 60. If your brand appeared in 18 of them, your presence rate is 30%.

This number on its own means very little. Its value is as a benchmark. Run the same audit in 30 days and again 60 days after that. If your presence rate is climbing, what you are doing is working. If it is flat or declining while competitors are gaining ground, you know where to redirect effort.

Step 6: Audit Competitor Appearances

This step is uncomfortable but necessary.

Go back through your log and note how often each of your main competitors appeared in the same prompts where you did or did not show up. How are they described? What language do the models use for them? Where are they cited?

This is not about feeling bad. It is about identifying the gap between their digital signal strength and yours. In many cases, a competitor appears more frequently not because they are better, but because they have more consistent content, stronger citation sources, or a cleaner entity profile across the web.

If a competitor is regularly described in favorable terms and you are not, look at where they are being cited. Perplexity's source URLs are particularly useful here. Those sources represent the content that is earning trust in the AI ecosystem. Understanding them tells you where to build yours.

What to Do With What You Find

The audit gives you a diagnosis. Here is how to act on the four most common findings.

If your brand is not appearing at all in category prompts, the problem is almost always content depth and digital signal. You do not have enough published, authoritative content establishing you as a credible voice in your space. The priority is creating structured, question-based content that directly answers what buyers search. Rakos builds this through the Search Everywhere Framework, which is designed specifically to establish visibility across both traditional and AI search simultaneously.

If your brand appears but with inaccurate information, you have a source quality problem. The model is pulling from outdated pages, third-party directories with wrong information, or thin content that does not clearly establish your current services. Audit your top-level web pages and your directory listings for consistency. Structured data (schema markup) helps models understand exactly who you are, what you do, and who you serve.

If your brand appears but with misaligned positioning, the issue is messaging clarity. Your website is not communicating your differentiators in language that is clear enough for an AI system to accurately summarize. This is often the same reason human visitors do not convert. Fixing it helps both audiences.

If competitors consistently outperform you, prioritize earning citations in sources those models trust. This includes industry publications, local news coverage, review platforms, podcasts, and educational content other sites link to. Digital PR and consistent content production are the two most reliable levers here.

How Often to Run This Audit

Monthly or bi- monthly is the right cadence for most businesses right now. The AI search landscape is shifting fast enough that a quarterly review will leave you behind. Monthly gives you a trend line without consuming unreasonable time.

Set a recurring calendar reminder. Keep your spreadsheet running across sessions so you can see the trend over time. After three to four months, patterns will emerge that tell you what is working and where the persistent gaps remain.

Frequently Asked Questions

Do I need paid tools to audit brand visibility on LLMs?

No. The manual process described in this post requires only a spreadsheet and free access to ChatGPT, Perplexity, Gemini, and Claude. Paid monitoring tools exist that can automate this process and provide ongoing dashboards, but for most small and mid-size businesses, the manual process produces accurate results and does not require additional budget to get started.

How long does an LLM brand audit take?

The first session typically takes 60 to 90 minutes, which includes building your prompt list, running prompts across four platforms, and logging results in your spreadsheet. Monthly refresh audits, using the same prompts and comparing results against your baseline, take roughly 20 to 30 minutes once the system is set up.

What if my brand does not appear at all in AI search responses?

This usually points to a content and authority gap. LLMs surface brands that have a clear, consistent presence across credible online sources, including well-structured web pages, third-party mentions, citations in publications, and strong review profiles. If you are absent, the priority is building that foundation, not finding a shortcut. A growth audit can help you identify the fastest path from invisible to present.

Can a business control what LLMs say about them?

Not directly. You cannot submit corrections to a language model the way you might update a business directory. However, you can influence what models learn by improving the quality, accuracy, and consistency of the content and sources they pull from. Better web content, stronger schema markup, earned mentions in credible publications, and consistent NAP data across the web all contribute to more accurate and favorable AI representations over time.

Is this the same as traditional SEO?

Related but different. Traditional SEO focuses on ranking pages for specific queries in search engines. LLM visibility optimization, sometimes called GEO (Generative Engine Optimization), focuses on ensuring AI systems accurately understand and represent your brand when synthesizing answers. Many of the same inputs matter, including content quality, site authority, and backlinks, but the output you are optimizing for is inclusion and accuracy in AI-generated responses, not a ranking position.

How do I know if AI search is actually sending customers to my business?

Check your GA4 referral traffic sources and look for traffic from domains like chat.openai.com, perplexity.ai, gemini.google.com, and claude.ai. AI-referred sessions grew 527% year over year in the first half of 2025, and while the absolute volume may still be small for most local businesses, the trajectory matters. Setting up source tracking now gives you data to act on as the volume grows.

What industries are most affected by LLM brand visibility right now?

Healthcare, legal, financial services, home services, and professional services are seeing significant AI search influence in buyer decisions. Consumers in these categories use AI to research providers, compare options, and evaluate credibility before making contact. If you operate in any of these spaces, an LLM audit is not optional at this point.

Most of what has changed about search in the last two years happened quietly. There was no announcement, no clear moment when the rules shifted. Just a slow, steady migration of buyer behavior toward platforms that compose answers instead of returning links.

The businesses that stay ahead of this are not necessarily the biggest or the best funded. They are the ones that decided to understand the environment they are operating in and adjust accordingly. Running an LLM brand audit is one of the most concrete, low-cost steps you can take right now to understand where you stand.

If you want a second set of eyes on your broader visibility strategy, including how your brand shows up across traditional search, AI search, and local platforms, book a growth audit call with Rakos Media. No pitch. Just an honest look at where the gaps are and what it would actually take to close them.

Sources

• 2025 State of AI Discovery Report, Previsible — 1.96 million LLM session analysis showing 527% year-over-year AI traffic growth
• New Front Door to the Internet, McKinsey — $750 billion AI-powered search revenue projection and 20-50% traditional traffic decline risk
• AI Discovery: What 2 Million LLM Sessions Tell Us, ALM Corp — Platform-level growth rates and AI-referred traffic trajectory data
• GEO: Generative Engine Optimization, Princeton / Georgia Tech / Allen Institute / IIT Delhi — Original research establishing the GEO framework and 40% visibility lift findings (ACM SIGKDD 2024)
• AI Search Statistics 2025-2026, Omnibound — Zero-click search rates, organic CTR decline, and AI Overview adoption data
• Future of AI Search, ZipTie.dev — McKinsey AI Discovery Survey data on consumer preference shifts toward AI search