Key concepts

• Workspace Your team. A workspace holds the projects, the members, the plan and the pool of credits shared by everyone in it. You can belong to several workspaces and switch between them from the Workspaces list at the bottom of the sidebar.
• Project One brand and its website. Everything else (products, audits, roadmap, content) lives inside a project. You switch projects from the selector at the top of the sidebar.
• Brand structure The map of what the project sells, in three levels: brand, category and product. Poliris builds it from your website and you can edit it. A project usually has one brand; you can add more.
• Product An offer as a buyer would ask an AI assistant about it, for example “trail running shoes” rather than each model and size. Variants are grouped under one product. AI visibility is always measured per product.
• Site A website attached to the project. The technical audit runs on sites. The project website is added automatically; you can add others, including sites you do not manage, such as a competitor.

• Trading zone The country or region where you measure visibility. AI answers change from one country to another, so every product is audited per zone. The settings tab calls them Trade Areas.
• Language The language the questions are written in. It is chosen per zone. The interface itself is available in English and French.
• Topic A theme your brand is judged on: quality, value for money, customer support, sustainability, innovation and so on. Topics come from a fixed list of about thirty. Each product follows three to five of them.
• Prompt A question sent to the AI assistants, written the way a customer would ask it and in the zone language. Prompts never contain your brand name, so the answer shows whether the assistant thinks of you on its own. Each prompt is linked to one or more topics.
• Model (LLM) An AI assistant Poliris questions. Free Models are the public web interfaces (ChatGPT, Gemini, Perplexity, Google AI Mode, Google AI Overview): Poliris asks the question like a visitor would and reads the answer. API models are queried through the provider’s official API (OpenAI, Google, Anthropic, xAI, Mistral) and give access to specific versions.
• Analysis run One execution of an audit: every selected prompt sent to every selected model, for one product in one zone. 10 prompts and 2 models give 20 answers. Poliris keeps every run, so you can compare them over time.
• Competitors The brands that appear in the answers. Poliris extracts them, cleans the list and follows your brand plus up to five competitors in all rankings.

Actions that query AI models or crawl pages use credits from the workspace pool. Audits and crawls show an estimate before they start. See Credits and plans for the details.