If you are still obsessing over your position for a keyword in a localized Google search, you are fighting a war that ended three years ago. Today, the battleground is RAG (Retrieval-Augmented Generation) and live web retrieval. If a Large Language Model (LLM) doesn’t know your brand exists, you aren’t just invisible in search; you’re excluded from the synthetic conversation entirely.
I’ve spent the last decade auditing sites that were once "industry-leading"—a term I despise because it lacks proof—and watching them drop off the map. When I audit these sites, the first thing I ask is: "What would I screenshot to prove this changed?" Most SEOs have no answer because they aren’t tracking entity recognition. They’re tracking blue links. Let’s change that.
Why is traditional rank tracking dead?
Traditional SEO is based on indexing and ranking a document. AI visibility, however, is based on entity extraction and knowledge graphs. When a user asks a question, models like ChatGPT don't look for a "top-ranking page"; they query a vectorized index of facts and entities. If your brand is not explicitly linked to the concepts (the entities) you want to own, the model will hallucinate a competitor into your spot.
Unlike standard Google results, AI retrieval is unpredictable and non-linear. You need a way to track whether a model identifies your brand when it retrieves information on a specific topic. This isn't about volume; it’s about attribution.
What is the simplest spreadsheet structure for tracking AI mentions?
To track this effectively, you need to treat your spreadsheet like a database of prompts and responses. Don’t overcomplicate it. You need a record of what you asked, where you asked it, and whether your brand showed up. If you aren't tracking your first mentioned brand, you are ignoring the most critical signal of authority in a response.
Below is the minimal viable architecture for an AI mentions tracker.
Prompt List Rows Topic/Entity ChatGPT Column Perplexity Column Gemini Column First Mentioned Brand "Best tools for X" Industry Software Mentioned (Yes/No) Mentioned (Yes/No) Mentioned (Yes/No) [Brand Name] "What is the alternative to Y?" Comparison Mentioned (Yes/No) Mentioned (Yes/No) Mentioned (Yes/No) [Brand Name]Keep a separate tab for the raw output logs. When you see a result, take a screenshot. If a competitor appears instead of you, that is your evidence. Compare the HTML of your page against theirs. Did they use specific schema that you didn't? Did they have an @id link that clarified their relationship to the entity? That is the data that matters.
How does Schema.org and @id linking fix visibility?
LLMs rely on structured data to disambiguate entities. If you have a company called "Echo," how does the AI know it’s your software and not a sound effect? You must define your brand as a node in the Knowledge Graph.
The simplest way to do this is through Schema.org @id linking. You need to ensure that every page on your site—especially your About, Product, and Service pages—includes JSON-LD that links back to your primary business identity.
Example of what your code should look like:
"@context": "https://schema.org", "@type": "Organization", "@id": "https://yourdomain.com/#organization", "name": "Your Brand", "url": "https://yourdomain.com", "sameAs": ["https://www.linkedin.com/company/yourbrand"]By defining the @id, you are creating a digital fingerprint. When a web crawler retrieves your page for a RAG-based search, it sees that specific ID and links it to your entity profile. Without this, you are just a string of text on a page, and the AI is guessing your relevance.
How do you audit your schema effectively?
Don't tell me your schema is "fine." I’ve seen hundreds of sites with schema that passes validation in basic tools but fails to link entities. Use the Google Rich Results Test religiously, but don't stop there. Look at the "Data" tab in the test. If your passage ranking for ai visibility entities aren't appearing as distinct, interconnected nodes, the AI is likely ignoring them.
I also keep a running list of bots in my robots.txt file. If you see a user-agent that shouldn't be scraping your high-value entity pages, block it. You don't want low-quality LLMs hallucinating data about your brand based on incomplete scrapes of your footer or sidebar.
Can you actually track AI referral traffic in GA4?
Tracking AI referral traffic https://stateofseo.com/what-does-recommendation-position-mean-in-ai-answers/ in Google Analytics 4 (GA4) is notoriously messy. Most AI platforms don't send a clean "referrer" header. They often strip it or pass it as "Direct."
However, you can get a directional signal. Look at your traffic segments for platforms that *do* pass headers (like some versions of Perplexity) and look for anomalies in your direct traffic spikes following a mention in an AI response. If you see a spike in traffic to a specific landing page that you’ve been testing in your prompt list, you’ve likely found your correlation.
It’s not perfect, but it’s better than staring at organic search rankings that are increasingly irrelevant to the actual decision-making process of the user.
Entities to watch: How Four Dots and FAII.ai demonstrate the shift
To see how this works in practice, look at brands like Four Dots or FAII.ai. They aren't just publishing "content"; they are optimizing their web presence for entity extraction. When you query these brands in an AI model, they appear consistently because they’ve clearly mapped their services to the entities the models are trained on.
If you search for them, you’ll notice the AI doesn't just list them; it describes their specific value proposition, often pulling directly from their structured data or well-defined entity pages. This is the difference between having a website and being an entity in the Knowledge Graph.
Final checklist for your AI visibility audit
Before you close your spreadsheet, ensure you’ve completed these steps to verify your brand’s presence:
Audit the Prompt List Rows: Are you testing long-tail queries that potential customers actually ask? Verify @id Linking: Is every page on your site pointing to a canonical Organization schema? Run the Rich Results Test: Do you see your entity connected to your services, or is it just sitting in a silo? Screenshot the Loss: When a competitor appears in your platform columns, screenshot the exact response. Compare their schema to yours. Check Robots.txt: Are you blocking unwanted scrapers, or are you accidentally blocking the retrieval-augmented bots that need to see your structured data?Stop guessing. If you aren't tracking the specific brand mention as a row in your sheet, you aren't doing SEO in the age of AI. You're just clicking "publish" and hoping for the best.