Setting up AI visibility tracking

The brand profile

Your brand profile tells Asheard who you are: brand name plus aliases — short forms, misspellings, the domain itself. Good aliases matter because AI assistants are inconsistent with names, and a mention only counts if we can recognise it.

Competitors

Add the competitors you want to benchmark against. Every AI answer is scanned for them too, so you always see share-of-voice, not just your own mentions. When engines keep recommending a brand you don't track, Asheard raises a "rising rival" alert suggesting you add it.

Prompts

Prompts are the questions you test — write them the way real customers ask: "best rank tracker for agencies", "what should I use to monitor AI mentions". Aim for a mix of category questions, comparison questions and problem questions. Avoid mentioning your own brand in the prompt: you want to know if AI recommends you unprompted. Each prompt gets an intent automatically; prompts that name your brand are marked branded and stay out of your headline numbers.

Engines and the answers budget

Pick which engines each project queries: ChatGPT, Gemini, Claude, Perplexity, Grok, Google AI Overviews, Bing Copilot, DeepSeek, Mistral. Google AI Overviews and Bing Copilot are search surfaces — an empty answer there means the engine showed no AI answer for that query, which is itself a signal. Your plan includes a number of AI answers per run — prompts × engines — and the picker shows a live calculation as you toggle engines, along with each engine's credit cost.

Answers per check

AI engines don't give the same answer every time. Under Brand settings, Answers per check makes every engine answer every prompt up to 5 times in one check, scheduled or Run now, so you see how often you are named instead of one sample. Each answer is charged like a check and refunded on its own if it fails; only the first answer can come from the 24-hour cache. In the matrix a cell then shows how many answers named you, for example 2/3: green when all did, amber when some did, red when none did. Every metric counts each answer. GEO Starter checks one answer; Pro and Agency allow up to 5.

Tagging prompts

Tag prompts by persona, product line, or funnel stage — the same tags you use for keywords. The tag filter on the AI visibility dashboard then scopes everything, charts and scores included, to just that segment: "how visible are we on enterprise questions" becomes a one-click answer.