An AI channel audit tells you where you stand in the conversations B2B buyers are having with ChatGPT, Claude, Gemini, and Perplexity before they ever visit your website. You need a baseline measurement of your Share of LLM before you build anything, optimize anything, or allocate budget to AEO. This is the entry point for companies ready to compete in the AI Demand Channel.
The audit process below takes most teams three to five hours spread over a week. You will end with a spreadsheet showing exactly where you appear, where competitors appear instead, and which gaps cost you pipeline.
Build Your Query Set
Start with 30 to 50 queries across three categories. Pain-first queries sound like “how to reduce customer churn without hiring more support” or “why sales forecasts are always wrong.” Category queries sound like “best revenue intelligence platforms for Series B SaaS companies.” Comparison queries sound like “[your company] vs [competitor]” or “[competitor] alternatives.”
Mirror actual buyer language, not your internal marketing language. If your pitch deck says “unified customer intelligence platform” but buyers search for “tools to track account health,” use the buyer version. Pull language from sales call transcripts, Gong snippets, support tickets, and closed-lost interviews.
Include queries where your company is not named. These reveal your dark funnel presence. A CFO researching “how to justify marketing spend to the board” will never type your brand name, but if you are absent from that answer, you lose the deal before it starts.
This query set becomes the foundation of your ongoing query library. See How to Build a Query Library for AEO for the full process of expanding and maintaining it over time.
Run the Queries Across Four LLMs
Test ChatGPT, Claude, Gemini, and Perplexity as your baseline set. These four cover most B2B buyer AI research behavior today. For each query, record four data points: inclusion (yes or no), position when mentioned (first, second, third, etc.), citation accuracy (are the claims about you correct?), and sentiment (positive, neutral, or negative framing).
Run each query twice with a few days between runs. LLM responses vary based on training updates, retrieval patterns, and context windows. Single-run data is unreliable. If you appear in one run but not the other, count that as inconsistent presence, which is almost as bad as no presence.
Use a simple spreadsheet: one row per query, columns for each LLM, four sub-columns per LLM for the data points. You do not need specialized tools for this. Manual querying gives you qualitative insight that automated tools miss, like how your product gets framed or which competitor gets mentioned alongside you.
Track citation accuracy closely. If an LLM says you offer a feature you deprecated two years ago or describes your pricing model incorrectly, that is a content freshness problem or a citation authority gap. Both are fixable, but you need to know which problem you are solving.
Benchmark Against Two Competitors
Choose your two closest direct competitors and run the same query set for them. Not the market leader. Not the startup that just raised a big round. The two companies buyers compare you to most often in late-stage deals.
You are not trying to win every query. You are trying to understand where you are systematically absent relative to competitors. If they appear in 70% of category queries and you appear in 20%, you have an authority problem. If you both appear but they are always listed first, you have a citation strength problem.
Note the query types where competitors appear and you do not. These are your highest-priority gaps. Note where you appear but competitors do not. These are your current advantages to defend. A competitor absent from pain-first queries but dominant in category queries tells you they have strong analyst relations but weak thought leadership content. You can outflank them at the top of the funnel.
Map Gaps by Query Category
Separate your results by query type: pain-first, category, comparison. Each gap type indicates a different problem with a different fix.
Pain-first gaps indicate missing top-of-funnel content. Buyers are researching problems and you are not present in the answer. These gaps are slower to fix because they require new content, not optimization of existing assets. But they have the highest long-term compounding value. A single well-cited article on “how to build a business case for customer success software” can influence dozens of deals before a buyer ever searches your category.
Category gaps indicate authority problems. AI systems know you exist but do not include you in category recommendations. This usually means weak third-party signals: no analyst mentions, thin review platform presence, few comparison articles from neutral sources. You cannot content your way out of this. You need analyst relations, customer reference programs, and review platform strategies.
Comparison gaps indicate either missing comparison content or low Citation Authority relative to competitors. If buyers search “[competitor] vs [you]” and the LLM does not mention you in the answer, you have a severe authority problem. If the LLM mentions you but frames the competitor more favorably, you need better comparison content and stronger citation sources backing your claims.
Mixing up these gap types wastes time. Pain-first gaps need content. Category gaps need third-party authority. Comparison gaps need both. The AEO Trifecta framework shows how these three citation layers work together, but the audit tells you which layer is weakest.
Score and Prioritize
Assign each gap a pipeline impact score: high, medium, or low. Base the score on query volume and buyer intent, not on how embarrassing the gap feels. A comparison query with high buyer intent gets fixed first even if the gap is narrow. A pain-first gap in a low-volume query can wait even if you have zero presence.
Comparison queries with high buyer intent are your fastest wins. A buyer searching “[you] vs [competitor]” is already aware of you. If the LLM answer is wrong or incomplete, you can often fix it with a single well-structured comparison page and a few authoritative citations.
Pain-first gaps take longer to fix but have the highest long-term value. A strong answer to “how to reduce churn in a usage-based pricing model” can influence deals for months or years. These gaps require original content, not optimization. Prioritize pain-first gaps that align with your product’s core value proposition and appear in multiple buyer personas’ research paths.
Category gaps often require work outside your content team. If LLMs do not include you in “best [category] platforms” answers, you need third-party validation: analyst coverage, review platform presence, mention in authoritative comparison articles. Track these gaps separately and route them to product marketing, analyst relations, or partnerships depending on the gap type.
For a full breakdown of what to measure and how to track progress over time, see 5 AEO Metrics to Track.
Audit Frequency and Ongoing Tracking
The AI Demand Channel is not static. LLM training cycles, algorithm updates, and new content from competitors all shift citation patterns. A query where you ranked first in March might not include you at all in May. A competitor’s new comparison page can displace you in Perplexity citations within days.
Run monthly audits at minimum. Use the same query set, the same LLMs, the same competitors. Track four metrics month-over-month: inclusion rate (what percentage of queries mention you), average position (where you appear when mentioned), citation accuracy (are claims about you correct), and competitor overlap (where you and competitors both appear).
Quarterly deep dives with full competitor benchmarking are best practice. Expand your query set, add new competitors, test additional LLMs. Look for patterns: are you losing ground in category queries but holding steady in pain-first queries? Are competitors citing new sources you have not seen before? Are there new query patterns emerging in your sales transcripts that are not in your library yet?
This audit reveals your Share of LLM baseline and your Citation Source Mix health. Both metrics should be tracked from the first audit run and compared month-over-month. A rising Share of LLM with a narrow Citation Source Mix means you are vulnerable. A stable Share of LLM with a diversifying Citation Source Mix means you are building durable presence.
Most B2B companies discover they have near-zero presence in the AI Demand Channel when they run their first audit. That is not a failure. It is a baseline. The companies that win in this channel are the ones that measure, prioritize, and fix gaps systematically. If you need help running your first audit or interpreting the results, A6 Group’s AI Channel Strategy practice works with B2B companies to build and execute full AEO programs from audit to optimization.
What is an AI channel audit and why do I need one?
An AI channel audit measures where your company appears in conversations buyers have with ChatGPT, Claude, Gemini, and Perplexity before visiting your website. You need a baseline measurement of your Share of LLM before building or optimizing anything, because the AI demand channel is now the primary B2B research destination. It reveals exactly where you appear, where competitors appear instead, and which gaps cost you pipeline.
How many queries should I include in my AI channel audit?
Start with 30 to 50 queries across three categories: pain-first queries (like ‘how to reduce customer churn’), category queries (like ‘best platforms for Series B SaaS’), and comparison queries (like ‘[your company] vs [competitor]’). Include queries where your company is not named to reveal dark funnel presence. This query set becomes the foundation of your ongoing library as you identify new buying patterns.
What data should I collect for each query across LLMs?
For each query on ChatGPT, Claude, Gemini, and Perplexity, record four data points: inclusion (yes or no), position when mentioned (first, second, third, etc.), citation accuracy (are claims about you correct?), and sentiment (positive, neutral, or negative framing). Run each query twice with a few days between runs since LLM responses vary based on training updates and retrieval patterns.
How should I source the queries for my audit?
Mirror actual buyer language, not your internal marketing language. Pull language from sales call transcripts, Gong snippets, support tickets, and closed-lost interviews. If buyers search for ‘tools to track account health’ but your pitch deck says ‘unified customer intelligence platform,’ use the buyer version. This ensures your audit reflects real research behavior.
How long does an AI channel audit take to complete?
Most teams complete the audit process in three to five hours spread over a week. Manual querying across four LLMs with two runs per query and careful data recording is achievable in this timeframe. You will end with a spreadsheet showing exactly where you appear versus competitors and which gaps impact your pipeline.