---
title: What Is AI Visibility? (And How to Measure It)
date: 2026-07-06T14:39:45Z
modified: 2026-09-15T17:32:07Z
permalink: "https://suparank.io/ai-visibility/"
type: post
status: publish
excerpt: AI visibility is how often AI answers mention or cite your brand. The four metrics and their formulas, what a good score is, and a step-by-step way to measure it.
wpid: 23
categories:
  - AI Visibility & GEO
tags:
  - AI Visibility & GEO
  - AI Citations
  - AI Visibility Tools
  - ChatGPT
  - Google AI Overviews
  - Perplexity
featured_image: "https://media.suparank.io/uploads/2026/07/thumb-ai-visibility.png"
featured_image_alt: What Is AI Visibility? (And How to Measure It)
author: Chase Allen
timestamp: 2026-09-15T17:32:07Z
---

AI visibility is the share of relevant AI-generated answers that mention or cite your brand. You measure it by running a fixed set of buyer prompts through engines such as ChatGPT, Gemini, Claude and Google AI Overviews, several times each, and counting how often your brand is named, linked and described correctly compared with competitors.

That definition is the whole idea in one paragraph. The rest of this guide turns it into numbers you can track: four metrics with formulas, an honest answer to “what is a good score”, a step-by-step measurement method, free ways to check today, and the levers that move the number.

## What is AI visibility?

![Three connected cards showing the layers of AI visibility: mentions, citations, and prominence](https://media.suparank.io/uploads/2026/07/sprk-ai-visibility-0-7.webp)

When someone asks an assistant “what’s the best payroll software for a 20-person company?”, the reply is a single synthesized answer, not a page of ten links. AI visibility describes your presence inside that answer, and it has three layers:

- **Mention:** the answer names your brand or product.
- **Citation:** the answer links to, or lists as a source, a page on your domain.
- **Representation:** what the answer says about you: where you appear relative to competitors, whether the tone is positive, and whether the facts (pricing, features, category) are right.

A brand can be mentioned without being cited (the model knows you but sourced a review site), and cited without being recommended (your page supplied a definition while a competitor got the recommendation). That is why one number is not enough.

## AI visibility vs SEO visibility

![Side-by-side comparison of traditional SEO versus AI visibility across optimization goal, key signals, and discovery model](https://media.suparank.io/uploads/2026/07/sprk-ai-visibility-1-7.webp)

| SEO visibility | AI visibility |  |
| --- | --- | --- |
| Unit of success | A position and a click | A mention or citation inside an answer |
| Surface | Ranked list of results | One generated answer (chat, AI Overviews, AI Mode) |
| Stability | Positions hold for days or weeks | Answers change from run to run |
| How you sample | One check per keyword | Many runs per prompt, per engine |
| Main data sources | Rank data, Search Console clicks and impressions | Prompt runs, Search Console’s generative AI report, AI referrals in GA4 |

The two overlap (crawlable, useful pages feed both) but they are measured differently.The stability row matters most. In a January 2026 study, SparkToro had 600 volunteers run 12 prompts through ChatGPT, Claude and Google’s AI a combined 2,961 times, and found a less than 1 in 100 chance that ChatGPT or Google’s AI returns the same list of brands in any two responses [1](#sprk-src-1). A single screenshot of one answer tells you almost nothing. Rates across many runs do. For the longer argument, see [why rank tracking falls short](/rank-tracking-is-dead/) in AI search.

## The 4 AI visibility metrics (with formulas)

Every metric below is calculated over the same inputs: a fixed prompt set, a fixed list of engines, and a fixed number of runs per prompt. Change any of those and the numbers stop being comparable.

### 1. Mention rate

**Mention rate = answers that name your brand ÷ total answers × 100.** This is the headline number: out of every answer your buyers could get, how often are you in it? SparkToro’s researchers reached the same conclusion from the other direction: rankings inside AI lists were close to random, but the percentage of runs in which a brand appears is “(probably) a reasonable way” to gauge how prominent it is in the AI’s consideration set [2](#sprk-src-1).

### 2. Citation rate

**Citation rate = answers that cite or link a URL on your domain ÷ total answers × 100.** Citations are the part of AI visibility that can send traffic, and they show which of your pages the engines trust as sources. Record the cited URL each time, because it tells you which content to protect and which to improve.

### 3. Share of voice

**Share of voice = your brand mentions ÷ mentions of all tracked brands (you plus named competitors) × 100.** Mention rate can rise simply because answers got longer. Share of voice controls for that by asking how much of the total conversation you own against a competitor set you choose in advance.

### 4. Sentiment and accuracy

**Accuracy rate = mentions that describe you correctly ÷ your mentions × 100**, with sentiment scored the same way (positive, neutral, negative). A mention that says you lack a feature you shipped last quarter can cost more than no mention at all. Log the wrong claim word for word so you can trace it to the outdated page or third-party listing that caused it.

## What is a good AI visibility score?

There is no universal good score, and be careful with anyone who quotes one. The number depends entirely on the prompts you chose, the engines you sampled and how crowded your category is. A 15% mention rate might lead a category with 40 credible vendors and trail badly in a category with three.

A score is useful in two ways:

- **Relative to competitors on the same prompt set.** Share of voice against three to five named rivals is the fairest comparison you can make.
- **Relative to your own past runs.** A move from 12% to 20% mention rate on an unchanged prompt set is a real signal, as long as each run samples enough answers to smooth out run-to-run noise.

We have not found a 2026 study that publishes credible benchmark ranges by industry, so we don’t give one here. Build your own baseline in the first month and compare against it.

## How to measure AI visibility, step by step

### Step 1: Build a prompt set

Write the questions buyers ask before they choose a vendor, not your keywords. Mix three types: category prompts (“best AI visibility tools for agencies”), problem prompts (“how do I see if ChatGPT recommends my brand”), and comparison prompts (“Profound vs Otterly”). Start with 10 to 30 prompts and freeze the wording, because users rarely phrase the same intent the same way and every rewording changes the answers [3](#sprk-src-1).



| Prompt type | Example | What it tells you |
| --- | --- | --- |
| Category | What are the best AI visibility tools for agencies? | Whether you make the shortlist |
| Problem | How can I see if ChatGPT recommends my brand? | Whether you’re associated with the job to be done |
| Comparison | Profound vs Otterly: which is better for a small team? | How you’re positioned against named rivals |
| Brand fact | How much does \[your brand\] cost? | Whether the facts engines repeat are correct |

A balanced prompt set covers four prompt types, not just category keywords.### Step 2: Run it across engines

Pick the engines your audience uses. For most brands that means ChatGPT, Google AI Overviews and AI Mode, Gemini and Claude; if GA4 shows referrals from other assistants, add them to your manual checks. Use logged-out or clean sessions where possible so personal history does not skew the answers.

### Step 3: Repeat runs, since answers vary

Run each prompt several times per engine and on a regular schedule (weekly is a sensible default). Record four things per answer: brand mentioned (yes/no), your URL cited (which one), competitors mentioned, and any incorrect claim.

### Step 4: Tie it to GA4 and Search Console

Visibility numbers need a reality check. In GA4, look at referral sessions from chatgpt.com, gemini.google.com and similar sources. In Search Console, Google’s generative AI performance report shows organic impressions from AI Overviews and AI Mode; as of August 31, 2026 Google says it has rolled this report out to all websites worldwide [4](#sprk-src-2). Our guide to [connecting GA4 and Search Console](/guides/connect-ga4-search-console/) covers the setup.

### A worked example (illustrative)

The numbers below are illustrative, not from a real brand. They show how the formulas fit together.



| Input or metric | Value | How it’s calculated |
| --- | --- | --- |
| Prompts × engines × runs | 10 × 3 × 3 = 90 answers | Engines: ChatGPT, Gemini, Claude |
| Answers naming the brand | 27 | Counted per answer |
| Mention rate | 30% | 27 ÷ 90 |
| Answers citing a brand URL | 9 | Counted per answer |
| Citation rate | 10% | 9 ÷ 90 |
| Mentions of all 4 tracked brands | 120 | Brand + 3 competitors |
| Share of voice | 22.5% | 27 ÷ 120 |
| Mentions with a wrong fact | 4 | e.g. outdated pricing |
| Accuracy rate | 85% | (27 − 4) ÷ 27 |

Illustrative example: one brand, three competitors, 90 sampled answers.Read it this way: the brand appears in almost a third of answers but is cited in only one in ten, so the gap is sources, not awareness. The next move is the page-level work in the last section, not more brand campaigns.

### How many runs are enough?

More than one, and more for list-style prompts. SparkToro’s volunteers ran each prompt dozens of times per tool to see which brands kept reappearing; one of their Google AI examples surfaced the same agency in 85 of 95 responses, which is the kind of stable signal a single run can’t show [5](#sprk-src-1). A practical rule: keep the prompt set and run count fixed, and only treat a change as real if it holds across two or more consecutive measurement cycles.

## Reading the numbers: what each pattern means



| Pattern | What it usually means | Where to look first |
| --- | --- | --- |
| Low mention rate, low citation rate | Engines don’t associate you with the topic yet | Crawl access, topical coverage, third-party mentions |
| High mention rate, low citation rate | Engines know the brand but source other sites | Your answer pages: are they direct, current and quotable? |
| High citation rate, low mention rate | Your pages supply facts, competitors get recommended | Comparison and use-case content, reviews, positioning |
| Mentions rising, share of voice flat | Answers got longer, competitors rose too | Competitor pages that gained citations |
| Mentions steady, accuracy falling | Outdated facts are being repeated | Old pricing or feature pages, directory listings |
| Good in one engine, absent in another | Engine-specific retrieval or crawl issue | That engine’s crawler access and the sources it cites |

Diagnostic patterns for AI visibility metrics. Use them as starting hypotheses, not verdicts.The cited URLs you logged are the most useful evidence here. If a competitor keeps getting cited from one comparison page, open it and compare: is the answer stated in the first sentence, is there a table, is there a date, does it cite sources? Those differences are usually more actionable than any aggregate score.

## How to check your AI visibility for free

1. **Manual prompts.** Run five of your most important buyer questions in ChatGPT, Gemini and Claude three times each and fill in the table above. It takes under an hour and gives you a baseline.
2. **Search Console.** Open the generative AI performance report to see impressions from AI Overviews and AI Mode for your own site [6](#sprk-src-2).
3. **Ahrefs’ free AI Visibility Checker.** As of September 2026, Ahrefs says the free tool checks ChatGPT, Google Gemini, Perplexity, Microsoft Copilot, Google AI Overviews and Google AI Mode [7](#sprk-src-3).
4. **A crawl-access check.** If AI crawlers and agents can’t fetch your pages, nothing else matters. Our free [agent-readiness checker](/tools/agent-readiness/) tests whether a URL is readable by AI agents.

Free checks give you a snapshot. When you need the same prompt set run on a schedule with history, competitors and share of voice, compare dedicated [AI visibility tools](/ai-visibility-tools/). Suparank, which publishes this guide, runs scheduled checks across ChatGPT, Google AI Overviews and AI Mode (plus Gemini and Claude on its Growth and Enterprise plans) and puts them next to your GA4 and Search Console data; see how we [track AI visibility](/product/ai-visibility/).

## How to improve AI visibility: 5 levers

Measurement tells you where you’re missing. These five levers are where to act, each with a deeper guide.

1. **Let the crawlers in.** Check robots.txt, your firewall or CDN bot rules, and whether key content renders without JavaScript. Generate a clean policy with the [robots.txt for AI crawlers generator](/tools/robots-for-ai-generator/).
2. **Write answer-first pages.** Put a direct, quotable answer near the top, use descriptive headings and tables, and keep facts current. Google’s own guidance is that foundational SEO and unique, people-first content are the basis for visibility in its generative AI features [8](#sprk-src-4). Engine-specific tactics are in [how to get cited by ChatGPT](/get-cited-by-chatgpt/) and [how to rank in Google AI Overviews](/google-ai-overviews-optimization/).
3. **Add evidence.** Original data, statistics and cited sources give an engine something worth quoting. The research that coined generative engine optimization found that optimizing content for generative engines could boost its visibility in their responses by up to 40% [9](#sprk-src-5).
4. **Earn third-party mentions.** Engines corroborate. Reviews, comparison articles, community threads and partner pages that name you correctly feed both mention rate and accuracy.
5. **Work through a checklist.** The [GEO checklist](/geo-checklist/) lists the technical, content and entity checks in priority order, and [generative engine optimization](/generative-engine-optimization/) explains the discipline behind them.

Then re-run the same prompt set. If mention rate rises but citation rate doesn’t, focus on source pages; if citation rate rises but accuracy falls, fix the facts on the pages being cited.

## Sources

1. [SparkToro, 2026](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/?utm_source=suparank.io) [1](#sprk-fnref-1) [2](#sprk-fnref-2) [3](#sprk-fnref-3) [5](#sprk-fnref-5)
2. [Google Search Console Help, 2026](https://support.google.com/webmasters/answer/16984139?utm_source=suparank.io) [4](#sprk-fnref-4) [6](#sprk-fnref-6)
3. [Ahrefs, 2026](https://ahrefs.com/ai-visibility-checker?utm_source=suparank.io) [7](#sprk-fnref-7)
4. [Google Search Central](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?utm_source=suparank.io) [8](#sprk-fnref-8)
5. [Aggarwal et al., KDD 2024](https://arxiv.org/abs/2311.09735?utm_source=suparank.io) [9](#sprk-fnref-9)

## Topics

**Categories:** [AI Visibility & GEO](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/category/ai-visibility-geo.md)

**Tags:** [AI Citations](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/post_tag/ai-citations.md), [AI Visibility Tools](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/post_tag/ai-visibility-tools.md), [ChatGPT](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/post_tag/chatgpt.md), [Google AI Overviews](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/post_tag/google-ai-overviews.md), [Perplexity](https://media.suparank.io/uploads/wp-mfa-exports/taxonomy/post_tag/perplexity.md)