Metric 1. Mentions — were you named or not
The most basic and the most telling. In Shopify’s framework (Greg Bernhardt) this is a binary signal: did your brand appear in the AI’s generated answer to a key query — yes or no.
How to collect it: a list of 20–50 key queries in your niche. Once a week you (or an assistant) put them to ChatGPT, Perplexity and Google AI Overviews and record “yes” or “no”. After four weeks you have a trend. After twelve, correlations with specific changes on the site.
Cost: 30–60 minutes a week. The alternative is specialised tooling such as OmniSEO or AgentFuel, which does it automatically.
Metric 2. Sentiment — in what tone you were mentioned
If the brand is cited, is it positive, neutral or negative? An AI can mention you as “a store with a good reputation” or as “a store with delivery problems according to Reddit”. Those are two different realities.
How to collect it: for every “yes” from metric 1, add a rating on three values: positive / neutral / negative. If you have the capacity, save the full sentence containing the mention. Within a month you will see patterns: what the AI “knows” about your brand and where it got that from.
Metric 3. Citation position — where you sit in the source list
An AI almost always lists the sources it cited after the answer. Position in that list matters: the first citation earns incomparably more real clicks than the fifth. This layer is a continuation of classic SEO and its ranking logic.
How to collect it: the same weekly run, but recording your position in the citation list. A rising trend means you are winning the RAG layer. A falling one means a competitor is doing something better.
Metric 4. AI referral traffic — actual sessions
Open Google Analytics → Acquisition → Source/Medium. Create a segment for the sources chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com. For stores with AEO working this is already 2–8% of total traffic, and the share grows every month.
What to look at inside it: session count, conversion rate, average order value, share of new visitors. AI traffic converts 1.5–2× better than cold organic — the user arrives having already done their research.
What you will not be able to measure
Traffic from Google AI Overviews — Google merges it with ordinary organic. To see anything at all, watch the trend for branded search in Google Search Console. A sharp rise in queries containing your brand with no obvious cause means you are being cited in AI Overviews.
That is a compromise you will have to accept. Until Google separates the sources in its reports, it is the only workaround.
A minimum dashboard in one day
One Google Sheet. Columns: Date, Query, Platform, Mentioned (Y/N), Sentiment, Citation position. 20 queries × 3 platforms = 60 rows a week. After eight weeks that is 480 data points — enough to test every change on the site as a hypothesis rather than as “it feels better now”.