How to Run an LLM Brand Audit: Find Out Whether AI Actually Recommends Your Business

Here's the uncomfortable question most founders aren't asking: when a potential customer asks ChatGPT, Gemini, or Perplexity for recommendations in your space, does your business show up?

If you don't know the answer, you're flying blind in the channel that's growing fastest.

Why this matters right now

G2's March 2026 survey of 1,076 B2B decision-makers found that 51% of B2B software buyers now start their research with an AI chatbot more often than with Google - up from 29% in April 2025. That's a 22-point jump in twelve months.

And AI platforms are brutally selective about who they recommend. SOCi's 2026 Local Visibility Index, analyzing 350,000+ locations across 2,751 brands, found that ChatGPT recommends just 1.2% of business locations compared to 35.9% appearance rates in Google's local 3-pack. That makes AI visibility roughly 30x more selective than traditional search.

If your entire digital strategy is built around Google rankings, you're optimizing for a channel that's losing share to one where you may not exist at all.

The brand audit: what you actually do

You don't need expensive tools to run a baseline LLM brand audit. Here's the process we use and recommend to clients.

Step 1: Build your prompt library (20–30 prompts)

Create prompts that mirror what your ideal buyer would actually ask an AI assistant. You need three categories:

Category prompts (non-brand): These are the most important. They test whether AI recommends you when nobody mentions your name.

• "What are the best [your category] companies in [your city/region]?"
• "Who should I hire to [the thing you do]?"
• "What should I look for when choosing a [your type of service]?"

Comparison prompts (soft-brand): These test whether you show up in competitive contexts.

• "How do I compare [service type] providers?"
• "Is it better to [approach A] or [approach B] for [outcome]?"
• "What questions should I ask a [your category] before hiring them?"

Direct brand prompts: These test whether AI knows you exist at all.

• "What does [your company] do?"
• "Tell me about [your company]'s approach to [your core service]."
• "Is [your company] good at [thing you're known for]?"

Moz's 300-prompt experiment (April 2026) found that prompts containing a brand name generated an average of 14.5 brand mentions per response, while non-brand prompts generated just 0.79. The gap is enormous - and it means the non-brand prompts are where you discover whether you have real AI visibility or none at all.

Step 2: Run prompts across all four major platforms

Test each prompt on:

• ChatGPT (relies heavily on training data + retrieval)
• Google Gemini (blends search index + AI generation)
• Perplexity (actively retrieves current web content)
• Claude (training data-heavy, entity verification-focused)

For each response, log:

1. Were you mentioned? (Yes/No)
2. What position? (1st, 2nd, 3rd, or not mentioned)
3. Was the description accurate?
4. Were competitors mentioned instead? Which ones?

Don't batch this in one session - models can exhibit recency bias within a conversation. Start fresh threads for each prompt.

Step 3: Score your visibility

Calculate three metrics from your logged results:

Mention rate: What percentage of your non-brand prompts resulted in your brand being mentioned? If this is under 20%, AI essentially doesn't know you exist in your category.

Position share: When you are mentioned, are you in position 1–3 (high trust, high click-through) or buried below competitors?

Accuracy rate: When AI does describe your business, is the information correct? Inaccurate mentions can be worse than no mentions - a model confidently describing outdated services or wrong specialties actively hurts you.

Step 4: Map the platform gaps

Each platform weights different signals:

Platform	Primary driver	What this means for you
Perplexity	Content freshness (content updated within 30 days gets 3.2x more citations)	Fresh, regularly updated content wins
ChatGPT	Brand search volume + referring domains	Brand awareness and backlink profile matter
Gemini	E-E-A-T signals + SERP position	Your Google organic rankings still influence this one
Claude	Entity verification + technical accuracy	Wikipedia, Wikidata, and structured knowledge sources

If you're visible on Gemini but invisible on ChatGPT, the fix is different than if you're absent everywhere.

What actually drives AI brand mentions

ConvertMate's analysis of 80 million+ citations across 10,000+ domains identified the factors that correlate most strongly with AI visibility:

1. Brand web mentions (0.664 correlation): How often your brand is referenced across news, publications, forums, and reviews. This is the single strongest predictor.
2. Brand search volume (0.334 correlation): How many people search for you by name. This proxies real-world brand awareness.
3. Content freshness (3.2x multiplier): Content updated within 30 days gets dramatically more citations across platforms.
4. Community presence (3.9x multiplier): Brands with significant Reddit presence average 7.0 citations vs. 1.8 for brands with minimal activity.
5. Structured data (67% improvement): Schema markup - Article, FAQ, HowTo, Product - makes your content more parseable by AI systems.

Notably, traditional domain authority metrics showed negative correlations with LLM citations in ConvertMate's research (Domain Rating: r = -0.18). LLMs don't care about your backlink profile the way Google's algorithm does. They care whether you're a recognized entity that real people reference in real conversations.

The fix priority list

Once you've run the audit, here's the order of operations for improving your AI visibility:

If AI doesn't know you exist at all:

• Get mentioned on industry publications, relevant listicles, and comparison pages (these are what models retrieve)
• Build out your entity presence: Google Knowledge Panel, Wikidata, Crunchbase, relevant directories
• Ensure your homepage and about page clearly state what you do in natural language (not marketing-speak)

If AI mentions you but inaccurately:

• Update your primary web properties with current, specific descriptions of your services
• Publish content that directly answers the questions AI is getting wrong about you
• Claim and update your profiles on review platforms and business directories

If AI mentions competitors but not you:

• Analyze what those competitors have that you don't: are they cited in the publications AI retrieves?
• Create original research, case studies, or data that AI models would naturally reference
• Engage authentically in community platforms (Reddit, relevant forums) where models learn about industry participants

If your visibility varies wildly across platforms:

• For Perplexity gaps: focus on content freshness - update key pages within 30-day cycles
• For ChatGPT gaps: focus on brand awareness and getting referenced by other sites
• For Gemini gaps: maintain your Google organic rankings (they still feed into Gemini's recommendations)
• For Claude gaps: strengthen structured data sources like Wikipedia and Wikidata

How often to re-audit

Run the full audit quarterly. Run a lightweight version (10 prompts, non-brand only) monthly. AI models update their knowledge and retrieval indices faster than you'd think - Perplexity reflects new content within hours, while ChatGPT's training data has a longer lag but its retrieval layer updates frequently.

Track month-over-month trends. A brand that's invisible today can become consistently mentioned within 60–90 days with focused effort.

The Apptitude take

We run this audit for our own business and for clients evaluating their AI strategy. The consistent pattern we see: companies that are well-known in their local market but have thin digital footprints outside their own website are almost universally invisible in LLM responses.

The fix isn't mysterious, but it does require a shift in thinking. Traditional SEO was about ranking your own pages. AI visibility is about ensuring your brand shows up as a known entity across the sources that AI models draw from - and that's a fundamentally different kind of work.

If you run this audit and discover you're invisible, you now have a concrete baseline and a clear set of priorities. If you need help translating audit results into an action plan - or want to understand how AI visibility fits into a broader digital strategy - that's exactly what our AI strategy engagements cover.

---

Sources:

• G2, "The Answer Economy: 2026 AI Search Insight Report" (March 2026 survey, 1,076 B2B decision-makers)
• SOCi, 2026 Local Visibility Index (350,000+ locations, 2,751 brands)
• Moz, "Brand Bias in Prompts: An Experiment" by Dr. Peter J. Meyers (April 7, 2026)
• ConvertMate, "AI Visibility Study 2026" (80M+ citations, 10,000+ domains)