LLM Brand Audits: What the $197 AI Visibility Scan Actually Checks

"I ran the nine-query test from your last post. We appeared twice out of nine. I know that's bad. What I don't know is whether the problem is our schema, our G2 profile, our product description, or something else entirely. I don't want to guess and spend three months fixing the wrong thing."

That message came in the day after the post on the five reasons your SaaS is invisible to AI chatbots went up. It captures the exact question this post is written to answer. Knowing you have a gap is step one. Knowing which gap to close first, with actual data behind the ranking, is what makes the difference between fast progress and wasted quarters.

This is a complete breakdown of what the $197 LLMRadar Audit measures: the engines it runs, the queries it uses, the signals it checks, and the sections of the PDF report you receive. Nothing is held back. If you want to run this yourself manually, you can. If you want it done in 24 hours with the analysis already structured, the order link is at the bottom.

Which engines the audit covers, and why those three

The LLMRadar Audit runs queries across three AI engines: ChatGPT (GPT-4o), Perplexity AI, and Claude (Anthropic). These three account for the majority of AI-assisted B2B vendor research as of mid-2026.

Each engine surfaces your product through different mechanisms. A brand that appears in ChatGPT but not Perplexity has a different gap profile than one that appears in Perplexity but not Claude. The audit captures your performance on each engine separately so you can see where your effort is best spent.

The 40 queries: how they are structured

The audit does not run generic searches. Every query is written to match the language real buyers use at different stages of vendor evaluation.

Here is the query distribution by type:

• Category-level queries (10 queries): "What are the best tools for [your category]?" These surface the default shortlist AI engines produce when asked an open-ended recommendation question. Your share of voice here tells you your baseline citation rate against the full competitive set.
• Comparison queries (10 queries): "Compare [your category] tools" and "What are alternatives to [leading competitor]?" These reveal whether you appear when buyers are actively comparing options. Many brands appear in category queries but vanish from comparison queries, which is a strong signal that entity signals are thin.
• Use-case queries (10 queries): Prompts built around the specific outcomes your product delivers, using the vocabulary buyers actually use. A fulfillment tool gets queries like "I need something that automatically sends customers their file after Stripe payment." A mismatch between how you describe your product and how buyers ask about it shows up clearly here.
• ICP-targeted queries (10 queries): Queries that specify the buyer's role and context. "I'm a solo SaaS founder, no engineering team, what do you recommend for automated onboarding?" These queries test whether AI engines can match your product to your specific customer type or whether you only appear for generic category searches.

The queries are customized to your actual product and category before the audit runs. You provide your product URL and a one-paragraph description of what it does. The audit team maps your category vocabulary, identifies the top three competitor names, and writes the 40 queries from that foundation.

The five signals the audit checks

After the queries run, the audit analyzes five structural signals that determine citation rate. These are the same five signals that separate the 27% of B2B SaaS brands that get cited from the 73% that do not.

What the PDF report contains, section by section

The deliverable is a PDF report, typically 12 to 18 pages, sent to your email within 24 hours of order. Here is what is in each section.

Section 1: Share of voice scorecard. A table showing how many of the 40 queries your product appeared in, broken down by engine and query type. Your overall share of voice score (appearances divided by total queries) is compared against the category average and the top competitor's score. This gives you the headline number: you appear in X of 40 queries; the category leader appears in Y.

Section 2: Raw AI response excerpts. The actual text each engine produced for a representative sample of queries. You see exactly how ChatGPT, Perplexity, and Claude describe your product category, which competitors they name, and what language they use. When your product does appear, you see the exact framing. When it does not appear, you see which competitors were named instead.

Section 3: Framing analysis. When your product appears in AI responses, how is it described? The framing analysis compares the AI-generated description to your intended positioning. Common framing gaps: AI engines describe your product as a tool for a different ICP than you target, assign you to a broader or narrower category than you occupy, or emphasize features you consider secondary while ignoring your primary value. Framing gaps require vocabulary fixes, not engineering work.

Section 4: Signal gap analysis. A scoring table for each of the five signals above, rated as Strong, Adequate, Weak, or Absent. Each weak or absent signal includes a specific action: the exact schema field to add, the aggregator profile to update, the vocabulary phrase to add to your product page, or the community thread to engage with.

Section 5: Prioritized fix list. The actions ranked by expected impact on citation rate. Not every gap is equally urgent. A brand with absent schema and strong entity signals should fix schema first. A brand with strong schema but thin community mentions has a different priority order. The fix list is sequenced based on your actual gap profile, not a generic template.

What the audit does not cover

Three things are explicitly outside the scope of the $197 audit:

Implementation. The audit tells you exactly what to fix and in what order. It does not implement the fixes for you. Schema edits, aggregator profile updates, and vocabulary revisions are things your team can execute directly from the fix list. Most founders complete the top three fixes within a week of receiving the report.

Ongoing monitoring. AI citation rates change as models update and competitors make changes. The audit is a point-in-time measurement. It reflects where you stand on the day the queries ran. Re-auditing in 60 to 90 days after implementing fixes is the standard practice to measure lift.

Paid search or traditional SEO. The LLMRadar Audit is scoped entirely to AI citation signals. It does not assess your Google rankings, backlink profile, or paid acquisition performance. Those require a separate audit methodology.

How to read your share of voice score

The share of voice score ranges from 0 to 100, representing the percentage of 40 queries in which your product appeared in the AI response. Here is how to interpret it:

Most B2B SaaS products that have not done deliberate AI visibility work score in the 0 to 10 range. Moving from that range to 21 or above typically requires fixing schema, updating two aggregator profiles, and making three to five vocabulary adjustments to the product page. That is work that can be completed in a week by a non-technical founder.

The difference between doing this yourself and ordering the audit

You can replicate most of what the LLMRadar Audit covers manually. The process looks like this:

1. Write 40 queries across the four types described above, customized to your product and category.
2. Run them across ChatGPT, Perplexity, and Claude, one at a time, recording responses in a spreadsheet.
3. Score your appearance rate by engine and query type.
4. Check each of the five signals manually: validate your schema with Google's Rich Results Test, check your G2 and Capterra profiles, read your product page first 200 words for category declaration, search Reddit for your product name, and compare your vocabulary to your top competitor's product page.
5. Build a prioritized fix list based on the gap scores.

That process takes four to six hours if you are organized and have not done it before. The audit takes 24 hours from your side: 10 minutes to submit the order form, then waiting for the PDF. The difference is structured analysis versus raw data. The audit produces a scored gap profile with context on what each gap means for your specific category and competitive set. The manual process produces a list of things you noticed without a benchmark to interpret them against.

Either approach works. The manual path costs more time. The audit path costs $197 and delivers a structured report you can act on immediately.