Why these numbers hold up.

Keywords, not prompts

A keyword is the intent behind every way people phrase a search. We translate each keyword into the single question that carries that full intent, ask it once, and keep the keyword's Google search volume. Example: in Google "paracetamol" means "what is it and where do I buy it"; in ChatGPT it is only "what is it". So we ask "what is paracetamol and where can I buy it" and anchor it to real demand. One keyword, one question, reproducible.

Real Google search volume as the measure of demand

Prompt-volume estimates are guesses. Verified Google search volume is something you can check yourself. We weight every score by it, so a 90,000-a-month intent counts more than a 90-a-month one. We only process keywords with at least 50 monthly searches, which keeps every number anchored in demand that actually exists.

The right-database advantage

Everyone else sells billions of keywords as the headline number. For AI search that is a liability, not an asset: you cannot run billions of prompts through LLMs at a sane cost, and most of those keywords are noise. We process our real, ranked-keyword databases per market, every keyword carrying real volume and intent. That is what makes coverage at this depth both possible and affordable. The big players' biggest boast is their biggest constraint here.

The metrics

• % mentioned (volume-weighted) is the headline. Across your tracked keywords, the share of AI answers where your brand is named, weighted by search volume. Keywords where no brand is named at all are excluded, so the score reflects only brand-competitive intents.
• % cited is secondary. The share of keywords where your own domain is used as a source, computed only where the engine shows citations: AI Overviews when present, ChatGPT in search mode. Per keyword we report your rank among the cited sources, not just yes or no.
• Per-engine state is reported separately for ChatGPT and AI Overviews: mentioned, not mentioned, or no brands named. AI Overviews adds "not present" when Google showed no overview at all, which is different from not being mentioned.
• Share of voice and sentiment round out the picture. Share of voice is paid: your mentions as a share of all brand mentions across the set, volume-weighted. Sentiment is coming soon: it reads the context around every brand mention and classifies it positive, neutral or negative, AI-confirmed.

Engines, cadence, and markets

v1 covers ChatGPT and Google AI Overviews. We always use the current default ChatGPT, the same one a logged-out person gets, because results track retrieved web content, not model version. The index is a single Q3 2026 snapshot today, refreshed quarterly for everyone, with monthly as the target as traction grows. Trends begin with the next update. UK and Romania are live; every other market is listed and you can join its waitlist, which moves it up the launch queue.

Branded and no-brand queries

The universe is non-branded, because that is where AI actually decides who to recommend. An exact brand name opens that brand's explorer instead. A pure navigational query ("Amazon contact") is not processed; we nudge you to the brand view or the de-branded intent. Keywords where AI named no brand at all stay in the table and can still carry a citation, but they are excluded from your mention score.