Your brand could rank on page one of Google and still be completely absent from the AI-generated answers that millions of buyers read first. An AI search visibility audit tells you exactly where you stand inside those answer surfaces, not just in classic blue-link rankings. Without this baseline, you're making content and optimization decisions with a significant blind spot. That blind spot grows larger every month as AI Overviews, ChatGPT, Perplexity, and Gemini handle more of the early research that leads to real purchasing decisions.
This post walks you through a structured, repeatable audit process built around four measurable layers: presence, placement, source attribution, and accuracy. It includes a 20-prompt starter set, a practical scoring rubric, and a monthly cadence you can run in-house or hand to an agency. If you've already invested in answer engine optimization services, this audit is how you prove those investments are moving the needle. If you haven't started yet, it's how you find out how much ground you need to cover.
Why Do Classic Rankings Audits Miss Half the Picture?
Traditional SEO audits measure keyword positions, crawlability, and backlinks, but they don't track whether your brand appears in AI-generated answers. Since AI Overviews and conversational AI tools now surface brand recommendations before users ever click a link, a ranking audit alone leaves a critical visibility gap unmeasured.
A conventional SEO audit is still essential. Crawl errors, Core Web Vitals, and link authority all influence whether AI systems trust and cite your pages. But the output of that audit, a list of keyword positions, tells you nothing about what Google's AI Overview says when someone types "best HVAC company near me." It also tells you nothing about what ChatGPT recommends when a prospect asks "which dental practice in Orlando accepts new patients." Those are the moments of intent that convert.
AI answer surfaces appear across the full buyer journey. Research from multiple SEO tooling companies consistently shows that AI Overviews appear on somewhere between 15 and 30 percent of queries, depending on the vertical and query type. In high-intent, local, and comparison-style searches, that share tends to be higher. Perplexity and ChatGPT handle millions of branded and category queries daily. If your audit process doesn't include a structured look at these surfaces, you're flying blind on a growing portion of your total addressable search audience.
The good news is that the same underlying signals that drive classic rankings, authoritative content, structured data, and trusted backlinks, also influence AI citations. The audit simply forces you to look at the output layer, the actual answer text, rather than stopping at the index layer.
What Are the Four Layers of an AI Search Visibility Audit?
The four audit layers are presence (is your brand mentioned at all), placement (are you recommended or just listed), source attribution (which of your pages or third-party sites is cited), and accuracy (are the facts the AI states about your brand actually correct). Each layer requires a different remediation strategy.
Presence is the binary foundation. You run a prompt and check whether your brand name appears anywhere in the response. A brand that never appears has a presence score of zero for that query. Aggregated across your 20-prompt set, presence gives you a percentage: your brand appeared in X out of 20 prompts across four platforms, equaling an 80-platform-prompt pool. This single number is your visibility baseline.
Placement goes deeper. An AI tool can mention your brand in a generic list of five competitors, or it can open its answer by specifically recommending you. Those two outcomes have very different commercial value. Score placement on a simple three-point scale: 0 for absent, 1 for mentioned in a list, 2 for featured or recommended as a top choice. A brand averaging 1.6 across its present prompts is in a meaningfully better position than one averaging 1.1.
Source attribution identifies which URL the AI system cites when it mentions you, if it cites one at all. ChatGPT and Gemini sometimes cite sources; Perplexity almost always does; Google AI Overviews cite specific pages. When your brand is cited from a review platform like Yelp or Healthgrades rather than your own website, that signals your own content isn't authoritative enough on that topic. When a competitor's comparison page cites you in an unfavorable context, that's a different problem entirely. Attribution tracking reveals which content assets are doing the work and which are being bypassed.
Accuracy is the layer most auditors skip. AI systems sometimes state incorrect details about brands: wrong service areas, outdated pricing language, incorrect ownership, or fabricated credentials. If an AI tells a prospective patient that your dental practice is closed on Saturdays when you're actually open, that error costs you appointments. Checking accuracy requires someone who knows your brand facts to read the AI responses carefully and flag discrepancies for a structured correction workflow.
6 Steps to Build Your 20-Prompt Starter Set
A prompt set is only useful if it mirrors real buyer language. Generic prompts produce generic results. Build yours around the specific questions your customers actually type or speak when they're close to a decision. Here's how to construct a 20-prompt set that gives you meaningful data.
- Identify your five highest-intent service categories. For an HVAC company, these might be AC repair, furnace installation, duct cleaning, maintenance contracts, and emergency service. Each category anchors four prompts total across your set, giving you coverage without redundancy.
- Write comparison and recommendation prompts. Phrases like "best [service] company in [city]," "who should I call for [problem] in [area]," and "[service] vs [service]: what do I need" mirror the queries that trigger AI Overviews and conversational AI answers most reliably.
- Include brand-direct prompts. Ask each platform directly about your brand: "Tell me about [Brand Name]," "Is [Brand Name] a good choice for [service]," and "What do customers say about [Brand Name]." These reveal what the AI believes about you specifically, including any accuracy issues.
- Add competitor-adjacent prompts. Prompts like "alternatives to [Competitor Name]" or "[Competitor Name] vs other options in [city]" show whether your brand appears as a recommended alternative. That's a common and high-value placement type.
- Include problem-first prompts. Start with the customer's situation, not the service label: "My air conditioner is making a grinding noise, who can fix it in [city]?" Problem-first prompts tend to trigger more recommendation-style AI answers than category-first prompts.
- Run every prompt on all four platforms. Google AI Overviews, ChatGPT (GPT-4o), Perplexity, and Gemini each have different training data, citation behaviors, and answer formats. A brand that dominates Perplexity but is absent from AI Overviews has a very different remediation need than one that is the reverse.
- Document the exact prompt text and date-stamp every run. AI answers are non-deterministic, meaning the same prompt can return different results on different days or even in different sessions. The audit value comes from running the same prompt set monthly and watching the trend lines, not from treating any single result as definitive.
How Should You Score and Interpret Your Results?
Score each prompt across all four platforms on presence (0 or 1), placement (0, 1, or 2), and accuracy (pass or fail). Aggregate these into a monthly scorecard. A brand scoring above 70 percent on presence and averaging 1.5 or higher on placement is in a strong position. Below 40 percent presence means urgent action is needed.
Here's a practical rubric. For each platform-prompt combination, record: Brand Present (1 point), Placement Tier (0, 1, or 2 points), Source Attribution Type (own site, third-party positive, third-party neutral, third-party negative, or no source), and Accuracy Status (pass or flag). A single platform-prompt combination can score a maximum of 3 points. Across 20 prompts and 4 platforms, your maximum raw score is 240 points.
Divide your actual score by 240 to get a visibility index. Scores in the 65-to-80 percent range indicate solid but improvable visibility. Below 40 percent signals that your brand is largely invisible to AI answer surfaces. That's the situation most local service businesses find themselves in when they run this audit for the first time. Above 80 percent suggests you're well-cited, but accuracy and placement quality still deserve close review. Even well-cited brands can be described inaccurately or positioned as a second choice.
To understand what drives your score, cross-reference source attribution with your content inventory. If Perplexity cites a three-year-old press release on a directory site every time your brand appears, that directory is outperforming your own pages on topical authority signals. That's a content gap you can close with targeted cluster pages, updated service content, and properly implemented schema.org structured data reference on your core service pages. If accuracy failures cluster around specific facts like service area, hours, or certifications, those facts need to appear more prominently and consistently across your own site and in your entity data across the web.
Track your visibility index month over month, not week over week. Month-to-month comparison smooths out the natural non-determinism of AI answers and gives you a signal that's actually meaningful. If your index rises from 38 percent to 51 percent over three months, something in your content or citation profile is working. If it stays flat despite new content, your strategy needs adjustment.
How Does SCALZ.AI Approach AI Visibility Audits for Clients?
Our team runs a structured monthly prompt-and-score process across all four major AI surfaces, cross-referenced with a client's existing content map and schema implementation. We treat the first audit as a baseline, not a verdict, and we don't draw optimization conclusions until we have at least two monthly runs for comparison.
In practice, we build each client's prompt set by combining their own search query data from Google Search Console with manual keyword research focused on buyer-intent phrasing in their service categories and geography. We then run each prompt in fresh browser sessions or API calls to reduce session-memory contamination. That contamination can cause platforms like ChatGPT to answer differently based on prior context in the same conversation. It's an often-overlooked methodological detail that skews results if ignored.
We also track source attribution at the URL level, not just the domain level. Knowing that ChatGPT cites a specific service page versus a location page versus an off-site review profile tells our content team exactly where to focus the next production cycle. We tie AI audit findings directly to the work outlined in our broader AEO audit process, which examines structured data completeness, FAQ and how-to schema coverage, and entity consistency across the web.
An honest limitation: this process takes meaningful time. Running 20 prompts across 4 platforms, documenting results, scoring them, and cross-referencing with content data is a two-to-four hour process per monthly cycle. For businesses in very low-competition local markets where AI Overviews rarely trigger for their query types, the return on that time investment is lower. We've had clients in small rural markets where the AI surfaces simply didn't appear for their core queries often enough to make monthly auditing worthwhile. In those cases, a quarterly cadence and a focus on classic SEO signals is the more efficient path. If you want a fuller picture of what metrics actually matter and how to track them, the post on how to measure AEO covers the broader measurement framework in detail.
For clients in competitive urban markets or multi-location brands, the monthly cadence consistently reveals actionable patterns within two to three cycles. The patterns we see most often: brand absent from Perplexity but present in AI Overviews (usually a citation-source problem), brand present but placed in generic lists rather than featured recommendations (usually a topical authority depth problem), and accuracy failures clustered around service area or hours (usually a structured data and NAP consistency problem).
Running an AI search visibility audit is not a one-time event. The answer surfaces change as models update, as competitors publish new content, and as your own site evolves. Treat the audit as a monthly instrument, the way you treat rank tracking, and it becomes genuinely predictive. The brands that establish this habit now, while the methodology is still being formalized across the industry, will have a data advantage that compounds. Start with your 20-prompt set, establish your baseline score, and let the trend tell you where to invest. For a broader context on Google Search Central on AI Overviews, Google's own documentation is worth bookmarking as the feature continues to evolve.
