The Hint Co. answer

How to measure AI visibility.

AI visibility is not one number. It is at least six: whether your page is retrieved, whether it is cited, whether your brand appears at all, whether your brand is picked, how often when picked, and whether any mentions are unfavorable. Most tools measure the first three and stop. The last three are the ones that map to revenue.

Why one metric is not enough.

A citation dashboard tells you the URL got surfaced. That is useful, and it is roughly a third of the story. Consider three things that all count as "the brand appeared" in raw measurement but produce four very different sales conversations:

Salesforce shows up in 28.5% of AI answers for B2B growth tools. 34 of those 77 appearances are neutral mentions. 9 are actively unfavorable ("Salesforce is bloated"). Salesforce is being cited, not chosen.
Clay shows up in 7.8% of the same category. Four times it is the sole vendor named. Four more it is the primary pick. Zero neutral. Zero unfavorable. When Clay is named, it is being chosen.
Snowflake shows up in 71.7% of data warehousing answers. Its pick share, meaning the percentage of total responses where it is the top pick, is 3.7%. Snowflake is cited constantly and almost never chosen.

All three would look identical on a citation counter. They are not identical. Salesforce needs a positioning fix. Clay needs a retrieval fix. Snowflake needs a recommendation fix. Different diagnoses, different work.

The six metrics that matter.

Retrieval rate. Percentage of prompts where any owned page is retrieved into the AI response context. This is where SEO ends and AEO begins.
Citation rate. Percentage of prompts where an owned page receives a visible citation.
Appearance share. Percentage of responses where your brand is named at all. This is the metric most tools stop at.
Pick share of total. Percentage of responses where your brand is the sole or primary recommendation. This is the metric that maps to revenue.
Recommendation-when-mentioned. When your brand appears, how often is it in the recommended set. A high number here with low appearance signals a retrieval problem. A low number with high appearance signals a positioning problem.
Unfavorable count. How often your brand is named negatively. Fifteen unfavorable mentions in a 270-response audit is an active sales conversation happening without you.

Corpus finding. Across seven audited categories, every category leader has a large gap between appearance share and pick share. Snowflake has the biggest, at 68 points. Cursor has the smallest, at 41 points. Nobody is closer than 35 points. The gap itself is a category signal.

The protocol.

The Hint Co. audit runs 30 buyer questions across 8 intent buckets, sampled three times against three engines (Claude web search, GPT-4o Search, Perplexity Sonar-Pro). Every brand mention runs through a two-stage matcher (word-boundary extraction plus LLM adjudication for ambiguous names) and a six-code recommendation classifier (SOLE, PRIMARY, ALTERNATIVE, CONDITIONAL, NEUTRAL, UNFAVORABLE).

The output is a structured report with appearance and pick share per brand, bucket-level breakdowns, verbatim excerpts, cited-domain classifications, and an adjudication log. Every audited brand can request a free counter-audit. Findings publish alongside.

The published methodology spells out every step. Full audits at the research library.

What most tools measure and where they stop.

Profound, Peec AI, Otterly.AI, Scrunch AI, and Semrush AI visibility all measure appearance and citation share. Some track sentiment. None publish a two-stage matcher or a six-code recommendation classifier as of July 2026, based on their public documentation.

That is not a criticism. Dashboards are a real product with real value. The distinction is that a dashboard tells you what is happening. A diagnostic tells you why and what to do about it. The Hint Co. sells the second kind.

Want this measured for your category?

Every 30-day pilot at The Hint Co. starts with a full AI visibility audit. Or buy the read on its own if the pilot is not the shape you need. Written by one operator, no agency team, no handoffs.

Apply