Measurement

Share of answers: the metric that replaces share of voice.

In search, you could count your rankings. In AI answers there are no rankings, only whether you were named. Here is how to measure that honestly.

Published October 2, 2026 ยท 5 min read

Why the old metrics break

Search gave marketers tidy numbers: position, impressions, click-through rate. An AI assistant gives one answer, written fresh each time, that may name two businesses, five, or none. There is no position two. The question that matters becomes simpler and harder at the same time: when a customer asks, are you in the answer?

What share of answers means

Share of answers is the percentage of sampled answers that name or recommend you, across a defined set of prompts and a defined set of assistants. If we run 40 prompts across 7 assistants once a week, that's 280 answers a week. If you're named in 115 of them, your share of answers is 41%.

A share of answers number is only meaningful with three things attached: which prompts, which assistants, and how often they were sampled.

How we calculate it

  1. Prompt map first. We agree the prompts with you before measuring anything, grouped into families such as discovery, comparison, local and purchase-ready. Changing the prompts changes the number, so they stay fixed month to month unless we both agree to change them.
  2. Fresh sessions. Assistants personalize. We run prompts in sessions with no history, with the location set to your market, so we measure what a new customer would see rather than what our own account has learned.
  3. Repeat runs. Answers vary between runs. Running each prompt more than once and averaging smooths the randomness, and we report a range when the variance is high.
  4. Score what counts. Named, recommended, or linked are scored differently, and sponsored appearances are tracked separately from organic ones. A paid placement never inflates your organic share.

What a good number looks like

For a local business starting from scratch, single digits is normal. A well-executed first quarter typically moves the discovery and local families fastest, because those prompts reward clear entity data and reviews. Comparison and purchase-ready prompts take longer; they reward depth and third-party citations. We'd rather show you a 12% that's climbing than a 60% built on a flattering prompt set.

What to watch out for

  • Prompt cherry-picking. Any tool can show a high share by choosing prompts you already win. Insist on seeing the full prompt list.
  • Single-run reporting. One run is an anecdote, not a measurement.
  • Platform updates. When an assistant changes its model or its sources, shares move for everyone. We flag those months so a drop isn't mistaken for your performance.

See how this looks in practice in our sample monthly report.

Want this done for your business?

Request a free AI Visibility Snapshot and we'll show you how you appear in AI answers today, or talk to a strategist about a plan.

Request a free snapshotTalk to a strategist

Frequently asked questions

What is share of answers?

The percentage of sampled AI answers that name or recommend your business, across a fixed set of prompts and assistants. If you appear in 41 of every 100 sampled answers, your share of answers is 41%.

How often should share of answers be measured?

Weekly sampling with a monthly report works well: often enough to see trends, with enough runs per prompt to smooth out the randomness in AI answers.

What is a good share of answers?

It depends on your category and starting point. A local business starting from scratch often begins in single digits; steady month-over-month gains in your most valuable prompt families matter more than a single headline number.