How we measure AI visibility results

Most brands can tell you their Google rank. Almost none can tell you whether ChatGPT recommends them. AI visibility feels invisible, but it is measurable: you just have to look in the right places and triangulate. Here is exactly how we do it, and what we report back to you every month.

Why the obvious metrics lie

The first instinct is to count mentions or raw traffic. Both mislead. A brand can be listed on every directory and still never get recommended, presence is not selection. A page can rack up impressions on off-strategy topics that never convert. So before any number means anything, we segment by intent: branded queries, high-intent decision queries, and everything else. We report those tiers separately, because a rising total can hide a flat result where it actually matters.

1. Daily per-prompt visibility tracking

We start from the questions your customers actually ask AI, your target prompts. Every day we put those prompts to the major models (ChatGPT, Gemini and others) and record how often, and how prominently, your brand appears versus your competitors. That gives each prompt a visibility score, and a scoreboard you can watch move. A prompt sitting at zero is not a failure, it is a topic we have not published for yet, and each one maps to the content pipeline. The pattern to expect: visibility moves in clusters, not a smooth line, flat while content is built, then a jump once it ships and gets indexed.

2. AI-attributed website analytics

Getting recommended only matters if it drives real visits and actions. We set up analytics that separate the traffic arriving from AI assistants from organic search and every other channel, so AI stops hiding inside “direct” or “unassigned”. Standard analytics systematically undercount AI, so this custom attribution, plus a simple “how did you hear about us?” field, is where the real contribution shows up. Inside Google we also watch which of your pages reach position one, the best proxy we have for being pulled into an AI Overview. Then we track two things: the volume of AI-referred visits, and, more importantly, what they do. In our own client data, AI-referred visitors convert around six times higher than Google organic, they arrive further down the decision, so even a small volume of AI traffic is worth tracking.

3. A triangulated visibility framework

There is no single source of truth for AI visibility, so we never rely on one. We layer proxies from Google Search Console with real visibility data from Bing Webmaster Tools, the only engine that currently exposes its own AI-citation data, which shows how often your pages are cited in Bing and Copilot's AI answers (Google's equivalent is coming). Read it over a rolling three-month window, not day to day. Cross-referencing channels also catches false signals: a spike in one source with no echo in the others is usually a ranking artefact, not real demand. Triangulating is what turns noisy data into a trustworthy read.

4. A monthly report that ties it together

Every month you get one report, in plain language. It opens with an executive summary, leads with your AI visibility score and its trend, breaks down traffic quality (are AI visitors engaging and converting?), and shows roadmap execution, what shipped, what moved, and what is next. The point is not a wall of numbers, it is a clear answer to one question: is AI starting to recommend you, and is it turning into business?

What good looks like over time

AI visibility is not an overnight switch, and it is not a slow bleed either. Expect the first results within a couple of months, and a clear trend within six months to a year, depending on the pace we work at. Because the work compounds, published content keeps earning citations, the gap between you and slower competitors tends to widen the longer it runs.