Every scan runs the same 25 queries against 4 independent AI engines. This page explains why those engines were chosen and how the queries are designed to model a real buyer journey.
The engine panel
AI visibility is not one thing. What GPT believes from its training data, what Gemini says inside Google's ecosystem, and what a web-grounded engine like Perplexity assembles from live search results are three different surfaces, and a brand can be strong on one and absent on another. The panel is built to cover both kinds of visibility: model knowledge and retrieval.
Direct model engines
| Engine | What it represents |
|---|---|
| OpenAI GPT | A live production OpenAI model answers from its own knowledge, the way ChatGPT answers when it does not browse. This captures what the most widely used model family actually believes about your space. |
| Google Gemini | A live Gemini model answers the same queries independently. Gemini powers Google's AI experiences, so its organic answer is a strong proxy for your Google AI footprint. |
| DeepSeek | DeepSeek represents the fast-growing open-model ecosystem that powers many AI products behind the scenes, and gives the panel a third independent training-data perspective. |
Direct engines answer purely from model knowledge: no browsing, no personalization, no chat history. Each engine leg is pinned to one specific vendor's model, and the engines never see each other's answers.
The web-grounded engine
The fourth leg works the way Perplexity and Google AI Overviews work. For each query we retrieve the top live web results, then have an AI model synthesize an answer from those sources only. This measures your retrieval visibility: whether the open web, as AI search engines see it right now, leads back to your brand.
The same retrieval step doubles as the source authority signal used in scoring: for every query, we record whether your own domain appears among the top live web sources. See How the score is calculated for how that feeds the score.
Query design: 25 queries, 5 intents
For your brand and keywords, the tool generates 25 queries: 5 in each of 5 intent categories, spanning the full buyer journey from research to decision. Generic one-size-fits-all prompts would overstate visibility for big brands and understate it for niche ones, so the queries are generated per brand while the category mix stays fixed for comparability.
| Intent | Journey stage | Example |
|---|---|---|
| Brand direct | Brand | "Is [brand] legit?" |
| Product category | Buying intent | "Best [category] tools in 2026" |
| Comparison | Buying intent | "[brand] alternatives" |
| Problem solving | Research | "How to improve [keyword]" |
| Long tail | Research | "[keyword] for small agencies" |
- Brand direct queries test whether AI engines know your brand at all and what they say when asked about it: reviews, trust, pricing.
- Product category queries are the highest-stakes ones. These are the "best X" questions where buyers discover options, and where being unnamed means losing the deal before it starts.
- Comparison queries catch the decision moment: versus questions and alternative lists where AI engines actively rank brands.
- Problem solving queries are the how-to questions your audience asks before they know they need a product.
- Long tail queries are niche, persona-specific questions where smaller brands can realistically win first.
With 25 queries across 4 engines, a full scan produces up to 100 independent AI answers. Each query and engine pair becomes one scored sample, which is the unit the whole scoring model is built on.