"I Googled my own brand and ChatGPT doesn't mention us."
That came from a B2B SaaS founder we audited in May. His product ranked on page one of Google for three competitive keywords. He had real traffic. His competitors weren't outranking him in search. But when buyers typed his category into ChatGPT, his product didn't appear. One competitor showed up, consistently, in four of the five major AI tools.
He wanted to know why. We ran the full audit to find out. This is exactly what we found.
The test most founders get wrong
The instinct when you first check AI visibility is to search your own brand name. You type your product into ChatGPT, it returns something, and you conclude you're visible. That's the wrong test.
The branded query ("what is [your product]?") tells you whether the AI knows you exist. Most products pass this test. The model has read your homepage, found your domain in some index, and can produce a sentence about you.
The category query is the test that matters. That's when you type "best tools for [category]" or "what should I use to [buyer problem]" and see whether your product appears in the recommendation list. That's the query buyers actually run at the moment they're deciding what to buy.
The founder from the opening was passing the branded test and failing the category test in four of five models. His competitor was winning the category test in four of five. That gap is a buying decision gap, not a content gap.
What each tool actually sees
We ran both tests across all five major models. Here's how each one processes the information you've given it, and where the gaps typically appear.
ChatGPT blends training data with real-time web search when browsing is enabled. It builds its understanding of your product from everything on the web: your own content, your competitors' comparison pages, forum discussions, review sites. New products with thin third-party coverage tend to underperform here because ChatGPT doesn't have enough signal to recommend them. The fix isn't writing more content about yourself. It's getting your product into third-party discussions and comparison pages where ChatGPT can find it.
Claude weights factual accuracy and specificity. A product described as "the AI operations platform that powers autonomous revenue" gets less traction than one described as "a workflow automation tool for B2B SaaS sales teams, typically deployed for outreach sequencing and pipeline management." Claude treats vague positioning as a credibility signal. The vaguer you are, the harder it is for Claude to recommend you in a context where a specific buyer need is being discussed.
Perplexity is the most transparent of the five: it shows you its sources. Run a category query in Perplexity and look at the citations panel on the right. If your domain isn't listed, you're not being cited. Perplexity draws heavily from Bing, not just Google. If your SEO work has been Google-only, you may be well-ranked in Google but barely indexed in Bing, which means Perplexity is largely ignoring you even when buyers in your category are actively searching.
Google AI Overviews integrates AI-generated summaries directly into Google search results, above organic listings, for informational and category queries. Overviews pull from indexed content and the Knowledge Graph. Appearing in an AI Overview for a category query is one of the strongest AI visibility signals available right now. Getting there requires structured data (especially FAQPage schema), clear entity definition, and enough link authority to signal credibility to Google's systems. It's not instant, but it compounds.
Gemini uses Google's Knowledge Graph more heavily than any other model. Your Google Business Profile matters here. So does structured data across your site and any coverage in credible third-party sources. Gemini also reads comparison content differently than ChatGPT: a well-structured comparison post explaining how your product differs from a named alternative tends to get cited directly in Gemini's answers to category queries.
The three gaps that explain most failures
After running this audit across a dozen products, three patterns explain most of the category query failures. They're structural, not content-quality problems. Better writing doesn't fix them. Fixing the structure does.
Gap 1: No FAQPage structured data. FAQPage JSON-LD tells LLMs what questions your product answers and what a buyer should know about it. A product page with five buyer-intent question-and-answer pairs in structured data is treated as a more citable source than an identical page without it. This is true across all five models. If your product pages, pricing page, and top blog posts don't have FAQPage schema, that's where to start. A developer can implement it in under half a day.
Gap 2: No llms.txt file. Several models now actively read the llms.txt file at your domain root when building their understanding of what a site does. The format is simple: a description of your product, a list of key URLs, and the primary use case you serve. Most B2B SaaS products still don't have one. Publishing it takes under an hour and gives every model a clean, unambiguous signal about how to describe you. A missing llms.txt isn't catastrophic, but a well-written one is a free signal boost.
Gap 3: Inconsistent category language across the site. LLMs build their understanding of your product from the language you use to describe it. If your homepage calls it "an AI operations platform," your pricing page calls it "an autonomous workflow tool," and your blog calls it "an agent orchestration system," the model gets a confused signal and defaults to whatever category language it already associates with this type of product. That language may not match what your buyers are searching for. Pick one phrase. Use it consistently on every page. The models will pick it up within a few crawl cycles.
The founder from the opening had all three gaps. No FAQPage schema on any page. No llms.txt. Three different category descriptions across his site. His competitor, who appeared in four of five models, had fixed all three. Not better content. Not more content. Cleaner structure.
Is this a Google problem or something different?
"Is this a Google problem or something different?" is the most common question we hear after founders run this test for the first time.
It's different. Related, but different.
Google SEO optimizes for Google's crawler: keyword matching, page authority, link signals. AI search visibility optimizes for how language models describe and recommend your product in response to buyer queries. The mechanics overlap (solid content helps both), but the signals don't map directly.
You can rank on page one of Google and be invisible in four of five AI tools. You can have thin Google presence and still appear in Perplexity if your content directly answers the questions buyers type. The two systems are running in parallel, and most B2B SaaS products have optimized for one while ignoring the other.
The gap is fixable. The three structural changes above close most of it. But you can't fix what you haven't measured, and most founders are still measuring the wrong thing (brand queries instead of category queries).
What to do this week
If you haven't run the category query test across all five models, do it today. Open each tool, type in the problem your buyers are trying to solve, and note whether your product appears. Run it as a buyer, not as someone who already knows your brand name.
If you appear in zero or one of five, you have a structural gap. The three fixes above close it. Start with FAQPage schema because it moves the needle across all five models simultaneously.
If you want to skip the manual work and get a full picture of where you stand across all five models right now, the LLMRadar Audit does that. We run the full test, identify the specific gaps by model, and deliver a prioritized fix list. No generic advice, just the actual results from your product in each tool.
Most B2B SaaS products are invisible to at least three of the five major AI tools their buyers use every day. That's a fixable problem. The founders who fix it now build a moat before the gap becomes the default state of their market.