How a score is calculated
AI answers move around. ARKMONT treats that as part of the measurement, and labels anything that did not come from a live engine.
Questions
A free check crawls the site, classifies it as SaaS, local, e-commerce, or services, and builds buyer questions for that type, in English or Spanish. Local businesses get city questions. Paid plans keep a list you can edit, up to the prompt cap on the plan, and add rival comparisons only when an engine named the rival.
Engines
We call the vendor APIs with web search where the API allows it: OpenAI web search for ChatGPT, Google grounding for Gemini, Perplexity search and Anthropic's web search tool for Claude. Grok answers from its own knowledge, without live search. The report names the model that answered. An engine without a key is skipped and labeled as having no API key. If every key is missing, all five run in simulation and the report says so.
Samples
The free score is a single sample, to keep the cost of the lead magnet low. The Monday job asks each active question three times. A manual run from the dashboard is one sample.
Mention and position
Mentioned means your brand name or domain appears in the recommendation list or the answer text. Position is the place in that list, starting at 1. If the prose names you but does not rank you, the mention counts and the position stays blank.
Score
The score is 0 to 100. Seventy points come from how often you were mentioned. Thirty points come from how close to first place those mentions sat. Errors are left out of the denominator.
Sources and competitors
Competitors are the other product names in the list. Sources are the URLs the engine returned. We do not check that every URL still resolves.