The AI Visibility Audit: What a Real Diagnosis Covers Before Any Fix Ships
Two companies both receive an "AI visibility report." One gets a score and a keyword list. The other gets a cross-model breakdown of exactly which queries each AI skips them on, which competitor gets recommended instead, and which URLs are being cited to justify that recommendation. Only one of those is an audit.
- 1.A rank report gives you a score and a mention rate; a real audit tells you which queries, on which models, and why those specific models skip you
- 2.Source attribution is the layer most AI visibility tools miss entirely: knowing which URLs AI cites when recommending competitors tells you exactly where to invest, not just what to write about
- 3.A complete audit covers both technical gaps (schema, entity resolution, crawlability) and content gaps; most tools only surface content
- 4.The query set must be built from your ICP buying scenarios, not a keyword tool; generic keywords do not map to how buyers actually query AI chatbots
- 5.A 25-query free snapshot and a 150-query full diagnostic are genuinely different instruments: this guide explains what each one is for
- 6.The audit output should deliver specific, prioritised actions with rationale, grouped by effort and expected impact, not a flat list of findings
The report came back a 34 out of 100. Mention rate of 6%. A list of queries where the brand did not appear. The marketing director forwarded it to her CEO with a note that said "we have work to do" and spent the next three weeks unable to answer the follow-up question: which work, exactly? ChatGPT skipped them on every solution research query. A competitor kept appearing in its place. The report confirmed the gap but explained nothing about its cause.
Compare that to a different output, delivered to a similar-sized B2B SaaS company the same month. That team learned that ChatGPT misclassifies their platform as a "monitoring tool" because the three URLs it cites when describing their category are all review aggregator summaries that predate their execution features. They learned that Perplexity mentions them in 13% of relevant queries because their G2 profile includes detailed CMS integration information. They learned that their Organization schema is missing a description field, so Gemini's entity graph treats the company as geographically ambiguous. They had a six-item action list by the end of the week.
Both documents were called "AI visibility reports." This guide explains what separated them.
Updated August 2026
What Is an AI Visibility Audit?
An AI visibility audit is a structured diagnosis of why specific AI models skip or misrepresent a brand across buying-relevant queries. It covers query coverage, cross-model benchmarking, competitor citation share, source attribution, and technical entity health, producing a prioritised action plan rather than a score alone.
What a Rank Report Gives You (and What It Leaves Out)
A basic AI visibility report delivers three things: an overall score, a list of queries where your brand was checked, and a mention or citation rate. Those numbers are real and worth knowing. The problem is not what they say but what they leave entirely unanswered.
Rank Report vs Real Audit: What Each Answers
A rank report tells you
- ✓Your overall AI Visibility Score
- ✓Your mention and citation rates
- ✓Which queries were checked
- ✓A general competitor comparison
A real audit also tells you
- ✓WHY each model skips you on specific queries
- ✓Which competitor is recommended instead, and from which sources
- ✓Which exact URLs AI cites to justify that recommendation
- ✓Whether your entity is correctly understood by each model
- ✓Technical gaps (schema, crawlability) vs content gaps
A score of 34 out of 100 tells you something is wrong. It does not tell you whether the problem is a crawlability block that stops AI bots from reading your site entirely, a schema gap that makes your entity ambiguous to Gemini, a content gap where your best-fit queries have no matching page, or a source gap where you are simply absent from the platforms AI reads most heavily in your category. Those are four very different problems with four very different fixes. A score cannot distinguish between them.
The other thing shallow reports miss: which queries were checked. A 6% mention rate across 25 generic category keywords tells a very different story than a 6% mention rate across 150 queries mapped to your actual ICP buying journey. The second number is actionable; the first is only a starting point.
The Five Layers of a Real AI Visibility Audit
A complete AI visibility audit works through five diagnostic layers, each answering a different question about why a brand is or is not being recommended. Most visibility gaps turn out to be caused by a combination of two or three of these layers working against each other simultaneously, which is why a score alone, without the layers underneath it, cannot point to a fix.
Query Coverage: Buying Scenarios, Not Keyword Lists
The query set determines everything else. A real audit builds queries from your ICP buying journey: problem recognition questions a buyer asks before they know your category exists, solution research questions where they are evaluating tools, and vendor evaluation questions comparing you directly to competitors. Generic category keywords do not map to how buyers actually query AI chatbots, and a mention rate across the wrong queries is noise.
Cross-Model Benchmarking: How Each AI Describes You Differently
ChatGPT, Claude, Gemini, and Perplexity are trained differently, index different sources, and weight different signals. A brand can have a 13% mention rate on Perplexity and a 2% mention rate on ChatGPT for the same query set, and the causes are distinct. Cross-model benchmarking identifies which models are skipping you, in which query categories, and how each model characterises your product. That characterisation gap is often where the most urgent work is.
Competitor Citation Share: Who Gets Recommended Instead
A complete audit maps which competitors appear in the responses where you do not, and their recommendation rates by query category. If a competitor wins 28% of solution research queries while you win 2%, the audit should tell you which query types account for that gap, not just that one exists.
Source Attribution: Which URLs AI Is Actually Citing
This is the layer most AI visibility tools omit entirely. When AI recommends a competitor, it is pulling from specific URLs: a G2 review page, a Reddit thread, a third-party comparison article. A real audit identifies those exact sources and classifies them by whether you can influence them directly (your own pages), partially (review platforms), or indirectly (independent editorial). That classification is what turns a visibility gap into a clear next step.
Entity and Technical Diagnosis: What Is Structurally Broken
AI models form an entity graph of every brand they know. If your Organization schema is missing key properties, your brand description contradicts itself across your own pages, or your site blocks AI crawlers via robots.txt, models build a low-confidence or incomplete picture. Technical diagnosis checks schema completeness, entity consistency, crawlability for each major AI bot, and Bing index status (ChatGPT searches Bing, not Google).
The Source Attribution Question
Can your current AI visibility tool tell you exactly which URLs AI models are citing when they recommend your competitors? If the answer is no, you know what topics to write about but not where to invest. Source attribution closes the gap between "we need more content" and "we need a presence on this specific G2 comparison page that Perplexity is reading for every solution research query in our category."
What the Audit Output Should Tell You
The output of a complete audit is not a flat list of findings. A flat list of 40 technical and content issues with no prioritisation is a backlog, not a roadmap. The marketing director who receives it will spend a week deciding what to do first and another week convincing engineering to care about any of it.
A real audit output does three things a findings list does not:
Each action ranked by expected visibility impact and implementation effort, with the specific reasoning: not just "add FAQ schema" but "add FAQ schema because Perplexity cites FAQ content for 40% of your solution research queries and you have none."
Technical fixes (schema, crawl access, entity resolution) and content gaps (missing pages, weak existing pages, off-site source gaps) are different workstreams with different owners. A useful output keeps them separate.
For off-site gaps, the output should name which platforms to target and why: not just "build a presence on review sites" but "your competitors are winning from G2 comparison summaries; your profile needs integration detail, not more star ratings."
A score tells you where you are. A diagnosis tells you why you are there. The gap between those two things is the gap between a report and an audit.
The level of specificity also determines what happens after the report is delivered. Vague findings produce vague next steps. "Improve your AI visibility score" is not a task anyone can assign. "Add FAQ schema to your three highest-traffic product pages covering the integration questions Perplexity cites most often" is a task with a clear owner, a clear deliverable, and a clear connection to a measurable outcome.
How BrandViz.AI's Diagnostic Phase Works
BrandViz.AI builds the query set from your ICP, not from a keyword tool. When a new brand connects to the platform, the diagnostic constructs buying scenarios from your product category, your target segments, and the specific roles in your ICP who make or influence purchase decisions. A 50-person B2B SaaS company selling to marketing directors gets different queries than a GovTech vendor selling to procurement teams, even if both operate in the same broad "software" category.
The full diagnostic runs up to 150 of those scenarios across ChatGPT, Claude, Gemini, and Perplexity, collecting every response, recording which competitors are named, tracing every citation back to its source URL, and classifying each gap by type: technical, content, or off-site source. The result is a report that looks like the one described in the second example at the top of this piece. You can see the actual output format at the BrandViz.AI demo report.
Full Diagnostic at a Glance
150
Buying scenarios per report
4
AI models benchmarked
ICP-first
Query construction method
URL-level
Source attribution depth
The free snapshot at /get-started covers 25 buying scenarios on ChatGPT and delivers results in under 10 minutes: the right starting point for a team that wants a fast baseline before committing to a full analysis. It will show your mention rate, your top missed queries, and which competitors are appearing in your place. What it will not show is the cross-model picture, the source attribution layer, or the full technical diagnosis. Those require the 150-query paid diagnostic, which is designed to be the foundation for the build and ship phases that follow, not a standalone report.
The diagnostic phase is also not a one-time exercise. AI models update their training, sources gain and lose citation weight, and competitor content changes. BrandViz.AI reruns the query set on a regular cycle so the action plan stays current rather than reflecting a snapshot that is three months old by the time fixes ship. For more on how the diagnostic feeds into the automated execution phase, see the guide to what to look for in a GEO platform with no technical team.
Frequently Asked Questions
What is the difference between an AI visibility audit and a GEO audit?
The terms are used interchangeably. An AI visibility audit and a GEO (Generative Engine Optimization) audit describe the same diagnostic process: a structured analysis of how a brand is currently represented across AI answer engines, what is causing gaps, and what actions would close them. "GEO audit" tends to appear in contexts where the remediation work is also in scope; "AI visibility audit" is more common when the diagnostic phase is being discussed on its own.
How many queries does a meaningful AI visibility audit require?
A 25-query snapshot is enough to produce a directionally useful baseline: your mention rate, top missed queries, and primary competitor displacement. A full diagnostic requires 100 to 150 queries to cover the complete ICP buying journey across problem recognition, solution research, and vendor evaluation stages, across multiple market segments. Below 50 queries, cross-model and cross-segment comparisons become statistically unreliable. The right scope depends on whether you need a fast baseline or a complete foundation for an execution plan.
Why does source attribution matter so much in an AI visibility audit?
Source attribution tells you what AI is actually reading when it makes a recommendation. Without it, you know you are underperforming but not where to invest to change that. With it, you know whether the problem is your own pages (a content or technical fix), your review platform presence (a sourcing fix), or independent editorial coverage (a PR or partnership fix). Those three problem types require completely different workstreams. Treating them as the same "content" problem is one of the most common ways AI visibility budgets get wasted. For a deeper look at how off-site sources drive AI citations, see our guide on why aggregate AI citation studies mislead strategy.
Does an AI visibility audit cover both technical and content gaps?
A complete audit covers both, and the distinction matters. Technical gaps include blocked AI crawlers, missing or incomplete Organization schema, entity inconsistency across owned pages, and Bing indexing gaps (since ChatGPT uses Bing as its live search layer). Content gaps include missing pages for high-value query types, answer capsule structure problems on existing pages, and FAQ coverage gaps. Most AI visibility tools surface only content gaps because they are easier to generate from response data. Technical diagnosis requires a separate crawl and schema audit layer. BrandViz.AI runs both and groups the output accordingly, since a technical crawl block will defeat even excellent content.
How is BrandViz.AI's diagnostic different from a monitoring tool's report?
A monitoring tool tracks what is happening. A diagnostic explains why it is happening and what to do about it. Monitoring tools report your mention rate and score changes over time; BrandViz.AI traces every gap to its cause, identifies the specific sources your competitors are winning from, classifies each finding by effort and expected impact, and generates prioritised actions that ship directly to your CMS as drafts. The diagnostic is designed to feed a build-and-ship phase; filing it away means the gaps stay open. For the full picture of what happens after the diagnostic, the guide to why brands are invisible on AI chatbots walks through the root causes a complete audit uncovers.
The free snapshot at /get-started runs 25 of your buying scenarios through ChatGPT and returns your mention rate, top missed queries, and primary competitor displacement in under 10 minutes. It is the right first step if you want to see the methodology before committing to the full 150-query diagnostic. The full diagnostic is what produces the source attribution layer, the cross-model breakdown, and the prioritised action plan that feeds into the build and ship phases.