GEO KPIs your CMO should ask for

1. Named rate: how often AI engines name you

The core KPI is the share of answers, across a fixed set of buying questions and runs, in which your brand is named.

Define the denominator first: the full protocol (every question, every engine, every run), never a subset chosen after the fact. Report a margin of error with every rate, and keep questions that already contain your brand name out of the total, because they inflate the result.

Example from our published case study, a US B2B SaaS vendor at baseline: out of 1,000 answers, the brand was named in 128, appeared only as a source in 110, and was absent from 762 (full case).

2. Cited rate: how often your pages are used as a source

A citation is a different outcome: your page supports the answer, whether or not your brand is named in the text.

Ahrefs, for example, counts a mention “when a brand appears at least once in an AI generated response” and a citation “when a page appears at least once as a cited source” ^[1]. The two move independently. Adding them together produces a bigger number that means nothing.

3. Share of voice: your slice of the answers

Share of voice compares your mentions with those of named competitors on the same questions.

It depends entirely on who is in the competitor list and how answers are weighted. Ahrefs’ version, for instance, weights impressions by the Google search volume of each prompt ^[1]. That is a legitimate choice, but it is a vendor choice. Ask for the competitor list (validated with you, not detected by an algorithm) and for the weighting in writing.

4. Sources: which domains the engines reuse

Knowing which sources the engines cite on your questions tells you where content has to live to change the answer.

Report the top cited domains for your question set, and where your own site ranks among them. In the same case study, the client’s own site moved from 5th to 2nd among all cited sources between the baseline and the re-measurement.

5. AI referral traffic and conversions

Traffic from AI engines is measurable, partially, and still small. Track it as a lagging indicator, not as the main KPI.

• GA4 channel. Since May 13, 2026, GA4 provides “a dedicated way to measure and analyze traffic originating from popular AI assistants” ^[2]. It “excludes Google’s AI Overviews and AI Mode” ^[3], which stay in organic search. Google reports traffic from its AI features inside the Search Console Performance report, under “Web” ^[4].
• Dark traffic. SearchPilot’s summary of AI referral tracking notes that app handoffs and copy-paste strip the referrer, so part of AI-driven visits arrive as Direct ^[5].
• Size. Across 101,574 sites, AI platforms accounted for 0.32% of all website traffic in 2026, up from 0.02% in 2024 ^[6].
• Quality. Adobe measured AI-referred visits to US retail sites converting 60% higher than non-AI traffic in July 2026 ^[7]. That is retail, not B2B. The widely quoted “23x” figure comes from Ahrefs’ own website over 30 days ^[8]: an anecdote, not a benchmark.

KPIs to drop from the report

• A rank from one answer. Lists change almost every run. See why one AI answer is not a measurement.
• A composite score with no formula. If nobody can say what goes into it, nobody can say why it moved.
• Changes smaller than the noise floor. Measure the variation with nothing changed first; anything below it is not a result.
• “Pipeline from AI” without a method. Attribution from AI answers is still being built across the industry. Ask how it was calculated before it goes in a board deck.

A one-page GEO report