The $0 AI Visibility Audit: Check What Every Major LLM Is Saying About Your Brand Right Now
A structured 20-prompt audit across ChatGPT, Gemini, Perplexity, and Claude that any marketer can run today. Includes scoring rubric, pattern analysis, and what to do with the results.
Before you start: understand what you are measuring
This audit gives you a directional snapshot of your AI visibility — not a statistically precise measurement. You are manually testing four platforms with four prompts each (16-20 queries total), running each query once. A single-run audit cannot give you statistical confidence in your mention rate. What it can give you is a clear picture of which engines are ignoring you, which intent types your competitors dominate, and whether the AI narrative about your brand matches your intended positioning.
For each engine, open a fresh incognito window before running queries. This minimizes personalization bias from your own search history. Do not use your logged-in account if you are a power user of any of these platforms — your session history will skew results.
This audit earns trust before it sells
The five query types to test per engine
Audit query types — replace [Brand] and [Category]
Type 1: Entity definition
“What is [Brand]?”
Tests: Entity accuracy, brand description, correct categorization
Type 2: Alternatives query
“Best alternatives to [Brand]”
Tests: How the AI frames your brand when positioned against alternatives
Type 3: Comparison query
“[Brand] vs [Main Competitor]”
Tests: Competitive positioning, whether framing favors you or competitor
Type 4: Category query
“Who are the top [Category] companies?”
Tests: Whether you appear in category leader lists at all
Type 5: Use case query
“[Specific use case] — what tool should I use?”
Tests: Whether you appear for intent-specific queries in your niche
The scoring rubric
For each query run, score the result across three dimensions. Track results in a spreadsheet with one row per query, one column per dimension.
Presence score (0-2)
2 — Primary mention: your brand is the recommended answer
1 — Listed mention: your brand appears in a list or comparison
0 — Absent: your brand does not appear
Accuracy score (0-2)
2 — Accurate: description, pricing, and positioning match current reality
1 — Partially accurate: some elements are correct, some outdated or missing
0 — Inaccurate: contains factual errors or significant misrepresentation
Framing score (0-2)
2 — Positive: framed as preferred choice, strong, or recommended
1 — Neutral: mentioned without strong positive or negative framing
0 — Negative: mentioned with caveats, limitations, or as secondary option
ChatGPT: specific prompts and what to watch for
Use ChatGPT with Browse enabled (the globe icon in the message input). This tests live citation behavior, not just training data. Key signals to watch: whether ChatGPT browses your website as a source, which competitor pages it retrieves for comparison queries, and whether it shows your brand in the “Searched the web for” sources.
Additional ChatGPT-specific prompt: “I need to solve [problem your product solves]. What are the best options available in 2026?” — This tests whether you appear in solution-oriented queries that match your product's core use case.
Perplexity: specific prompts and what to watch for
In Perplexity, look at the sources panel on the right side of each response. This shows you exactly which pages Perplexity retrieved to generate the answer. If your website appears in the sources, you have a retrieval advantage. If competitor pages dominate the sources, that explains why competitor brands appear in the response.
If you have Pro access, use Focus mode set to “Web” for fair comparison testing. The “Academic” and “Reddit” focus modes will produce systematically different results that do not reflect general AI search behavior.
Gemini: specific prompts and what to watch for
Gemini is more conservative about recommending specific brands than ChatGPT or Perplexity — it often hedges with “results may vary” language. Watch for whether your brand appears in the knowledge graph panel (the sidebar information) when you query your brand directly. The knowledge graph appearance indicates Google has strong entity confidence in your brand.
Claude: specific prompts and what to watch for
Claude with web search behaves differently from ChatGPT and Perplexity — it tends to be more skeptical of single-source information and more likely to note limitations or context. Watch for whether Claude presents your brand with confidence or adds hedging language (“according to the company,” “claims to,” “reportedly”) — these hedges indicate low entity confidence or insufficient third-party validation.
Interpreting your results
Pattern-to-diagnosis mapping
Absent across all engines on Type 4 (category query)
Diagnosis: Entity authority problem — you are not recognized as a category player
Fix: Build cross-platform entity presence, add Organization and DefinedTerm schema
Present on Type 1 but absent on Types 4 and 5
Diagnosis: Entity known, but not associated with specific use cases
Fix: Create use-case-specific content with HowTo schema and FAQPage for each use case
Present but with Framing score 0 (negative framing)
Diagnosis: Narrative drift or competitor comparison content dominating
Fix: Audit citation sources, create structured counter-narrative content
Present on ChatGPT, absent on Gemini
Diagnosis: Weak Google Knowledge Graph presence
Fix: Strengthen entity signals in Google-readable formats (schema, Wikipedia, Google Business Profile)
Accuracy score 0 (factual errors)
Diagnosis: Outdated training data or wrong source being cited
Fix: Update structured data with current facts, add dateModified schema, create fresh FAQ content
Fix priorities based on your audit score
Maximum possible score across 20 queries (4 engines × 5 query types) across 3 dimensions is 120 points. Interpret your total:
90-120: Strong baseline. Focus on moving from Tier 2 (shortlisted) to Tier 1 (primary recommendation) citations. 60-89: Moderate visibility. Prioritize schema implementation and cross-platform entity presence. 30-59: Significant gaps. Start with entity definition (Organization schema + DefinedTerm), then FAQPage schema for comparison queries. 0-29: Critical gaps. Run a full RankAsAnswer audit to get a complete diagnostic before investing in fixes.
Automating this audit at scale