AI Share of Voice (SoV)

TL;DR — AI Share of VoiceAI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric. is how much of the AI-answer conversation your brand owns. You ask a bunch of AI tools (ChatGPT, Gemini, Perplexity, and so on) the questions your customers ask, then count how often your brand gets named or linked versus your competitors. More mentions = higher share. It tells you whether AI knows you exist — not whether anyone clicked.

What AI Share of Voice is

AI Share of Voice is a practitioner-defined measurement, not a standardized metric published by model providers. Evidence for this claim AI share of voice is an editorial measurement framework for how often selected entities appear across a defined sample of AI answers. Scope: Practitioner metric defined by its prompt set, platforms, entities, scoring rules, and observation window; not a standardized platform metric. Confidence: medium · Verified: 2026-07-14 OpenAI: ChatGPT search Results depend on the prompt set, sampling time, model, locale, and whether mentions or citations are counted. Evidence for this claim An AI share-of-voice result is sample-dependent and can change with the prompt set, model, product behavior, date, and scoring method. Scope: Measurement limitation; the source documents the changing answer-and-search product being sampled, not a universal AI-SOV formula. Confidence: medium · Verified: 2026-07-14 OpenAI: ChatGPT search

When someone asks ChatGPT “what’s the best project management tool for a small team?”, the answer names a handful of brands. AI Share of Voice (SoV) is a way to measure how often your brand is one of those named — across a whole set of questions, compared to your competitors.

The math is simple:

AI SoV = (your brand’s mentions ÷ total relevant responses) × 100

If you run 1,500 questions across AI tools and your brand comes up in 300 of the answers, that’s a 20% AI SoV.

Mentions vs. citations

Two different things get counted, and the difference matters:

• Mention — the AI names your brand in the text (“tools like Asana, Trello, and Monday…”). No link.
• Citation — the AI links to a page on your site as a source.

You can be mentioned without being cited, and cited without being mentioned. Good tracking watches both.

Why it matters now

People are starting to ask AI for recommendations instead of scrolling a page of blue links. An AI answer usually names just three to five brands — and most users take the first one suggested. If you’re not in that short list, you’re invisible in a channel a lot of buyers now start with.

The one thing beginners get wrong

High AI Share of Voice does not automatically mean more traffic. AI can recommend you without sending anyone to your site. SoV tells you whether AI knows and trusts your brand — it’s a visibility score, not a click counter. Treat it as one signal next to your actual traffic and sales, never as the whole story.

Want the formula details, the benchmark tiers, how to build a prompt pool, and why brand mentions (not keywords) drive AI SoV? Switch to the Advanced tab.

TL;DR — AI SoV = brand mentions ÷ total relevant responses × 100, measured across a sampled prompt pool run through multiple AI platforms. It is not traditional SoV reskinned: traditional SoV is keyword-rank × CTR × volume, AI SoV is mention/citation rate, and the signals differ — branded web mentions (0.664) and branded anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. (0.527) predict AI visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals. far better than Domain Rating (0.326). It’s volatile (40–60% citation churn month to month), varies hugely by platform (35% in Gemini vs. near-zero in ChatGPT for one brand), and is only as valid as the prompt pool you build. Benchmark tiers (Birkett): <20% visibility problem, 20–50% established, 50–90% category leader, >90% dominant; average mention rate sits around 17.2%. And the caveat that matters: SoV alone is a vanity metric — connect it to pipeline or it’s slideware.

What AI Share of Voice actually measures

The metric estimates observed answer presence within a defined sample; it does not measure all user conversations or internal model preference. Evidence for this claim AI share of voice is an editorial measurement framework for how often selected entities appear across a defined sample of AI answers. Scope: Practitioner metric defined by its prompt set, platforms, entities, scoring rules, and observation window; not a standardized platform metric. Confidence: medium · Verified: 2026-07-14 OpenAI: ChatGPT search Methodology and uncertainty should be reported with the number. Evidence for this claim An AI share-of-voice result is sample-dependent and can change with the prompt set, model, product behavior, date, and scoring method. Scope: Measurement limitation; the source documents the changing answer-and-search product being sampled, not a universal AI-SOV formula. Confidence: medium · Verified: 2026-07-14 OpenAI: ChatGPT search

AI Share of VoiceAI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric. measures how often and how prominently your brand appears in AI-generated responses, relative to competitors, across a defined set of prompts. The formula:

AI SoV = (brand mentions across tracked prompts ÷ total relevant responses) × 100

300 mentions across 1,500 prompts = 20% AI SoV. Simple enough. Two sub-signals get tracked separately by most tools:

• Mentions — your brand name appears as text in the response, no link.
• Citations — a linked URL from your domain shows up as a source.

These diverge constantly — AI will recommend a brand by name without linking to it, and link to a page without naming the brand in prose. Ahrefs Brand Radar splits these out (mentions, citations, impressions) and shows SoV history per platform, because blending them hides where you’re actually weak.

How AI SoV differs from traditional Share of Voice

This is the part people skip, and it’s the whole point. Traditional SoV is ranking-centric: keyword position × CTR × search volume, rolled up across a tracked keyword set, living inside GSCA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results., Ahrefs, or Semrush. AI SoV is mention-centric: brand mention/citation rate across a sampled prompt pool.

The inputs are different, so the optimization levers are different. A brand can dominate traditional organic SoV and still post near-zero AI SoV if its brand entity is weak in AI training data and third-party sources. They are measuring largely different intent pools too — a big share of AI prompts have no matching traditional search keyword at all, so your GSC keyword tracking and your AI SoV are looking at different questions.

What actually drives AI Share of Voice

In December 2025 I ran correlation analysis on factors associated with AI Overview visibility (presented at Ahrefs Evolve 2025). The results invert the traditional SEO playbook:

Read that order again. Off-page brand presence beats on-page keyword optimization, and beats raw domain authority. AI SoV is driven primarily by how present your brand is across the web as an entity — not by keyword density on your pages. That’s the core insight that makes AI SoV a distinct discipline rather than a new column in your rank tracker.

It lines up with how the systems behave. As Carolyn Shelby puts it, “AI doesn’t discover new brands — it selects from known entities.” And roughly 82–89% of AI citations come from earned media, not brand-owned pages. Digital PR, entity presence, and third-party coverage precede AI selection. Content strategy alone can’t close the gap.

Beyond raw mention rate, four signals shape competitive position:

• Mention order — about 74% of users select the top recommendation, making the order you’re listed in the most commercially loaded signal in the whole metric.
• Depth of explanation — thorough pages get cited far more than thin ones.
• Authority language — being described as “the industry standard” vs. “gaining traction” changes how the answer positions you.
• Comparative framing — “best for startups” vs. “best for enterprises” decides which queries you win.

The prompt pool is the measurement

Alex Birkett says it cleanly: “The prompt set IS the measurement. Its construction determines validity.” A poorly built prompt pool produces SoV numbers that look precise and mean nothing.

Quantity floors:

• 100–200 prompts — minimum viable set for tracking.
• 250–500 prompts — recommended for serious competitive analysis.

But quality matters more than quantity. Build prompts from real buyer language, not internal jargon: sales call transcripts, support tickets, Reddit threads, G2 reviews, People Also Ask, and autocomplete. Validate them against actual search data so you’re measuring questions people genuinely ask. Then refresh the pool quarterly — the questions move.

TIP Keep mentions and known-fact coverage separate

One sampled answer can mention a brand while omitting a fact you expected it to contain. Record those as different observations, and never treat unevaluated providers as zero share of voice.

Run a controlled brand-and-fact sample with my free AI Brand Visibility Checker Free

1. Use one prompt from the fixed pool and define any exact brand fact you expect the answer to include.
2. Record brand mention, fact coverage, provider, model, retrieval mode, and run date as separate fields.
3. Repeat across the planned provider sample before calculating share; an unevaluated provider is missing data, not a zero.

Tap to enlarge

A simulated Llama response mentions the fictional Acme Analytics brand by exact match. The known-fact check marks Founded in 2018 as missing. Mistral, ChatGPT Search, Gemini, Claude, and Perplexity are each marked not evaluated with a warning not to infer zero mention or citation rates.

Platform variance is real and large

Tracking one AI platform produces systematically biased data. The same brand recorded 35% SoV in Gemini and meaningfully lower in ChatGPT. Citation behavior varies structurally too: Perplexity and Copilot include external links in over 77% of responses, while ChatGPT does so in roughly 31% — so citation rate is partly a property of the platform, not just your content. I’ve seen one brand sit at 27% citation rate on Grok and 0.59% on ChatGPT. Same brand. Same content.

The minimum is three platforms. A serious setup tracks ChatGPT, Gemini, Perplexity, Copilot, and Google’s AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. (plus AI Mode and Grok where the budget allows).

Benchmarks

Alex Birkett’s tiers are the most-cited reference:

For context, the AthenaHQ State of AI SearchAI search uses large language models and retrieval-augmented generation (RAG) to synthesize an answer from multiple sources rather than returning a ranked list of links. Examples include Google AI Overviews, ChatGPT Search, and Perplexity. 2026 puts the average brand mention rate at 17.2% — so most brands are sitting in or below the “visibility problem” band. Calibrate by market: fragmented markets run lower (15%+ is already competitive), concentrated markets see leaders at 35–50%+.

It’s volatile by nature

AI SoV moves. 40–60% of cited sources change month-to-month (eMarketer), which is why a single-month snapshot is misleading. As Rand Fishkin frames it, “almost every GEOGenerative Engine Optimization (GEO) is the practice of optimizing content and brand presence so AI-powered search engines and assistants — Google AI Overviews, ChatGPT, Perplexity — cite, recommend, or mention you when generating answers. Google's position is that it's still SEO. response is different from every other GEOGenerative Engine Optimization — visibility inside AI answer engines. response.” You need a 12-week minimum window before drawing any trend conclusion. Brands that “earned” SoV one month can’t coast — ongoing monitoring isn’t optional.

The caveat that matters: SoV alone is a vanity metric

This is the load-bearing warning. Paul DeMott nailed it: “Citation share, presence rate, and AI Overview appearance counts are the new domain authority. They look defensible in a slide. For 95% of the agencies selling them, they aren’t connected to pipelines in any rigorous way.”

He’s right. SoV alone is a vanity metric. It’s a visibility signal, not a performance metric, and it is not a substitute for traffic analytics. AI can recommend you without linking to you; even citations don’t guarantee clicks. The discipline is to measure SoV alongside referral traffic, pipeline attribution, and brand-search-volume trends — never instead of them. If your SoV number can’t be traced toward revenue, you’re reporting slideware.

Where to go next

AI SoV is one metric in the broader measurement and reportingAI search measurement and reporting is the practice of tracking and communicating how AI search features — AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot — affect a brand's visibility, traffic, and revenue. It exists because most AI-referred traffic and exposure is invisible in standard analytics. discipline. Pair it with AI traffic attributionAI traffic attribution is the practice of correctly identifying and measuring website visits that come from AI tools — ChatGPT, Perplexity, Gemini, Claude, AI Overviews, and AI browsers. It's hard because many of those tools strip the referrer header, so the visits land in your analytics as Direct traffic with no source. (what AI actually sends you) and LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). visibility monitoring to get from a vanity number to a defensible one.

AI summary

A condensed take on the Advanced version:

• AI SoV = brand mentions ÷ total relevant responses × 100, measured across a sampled prompt pool run through multiple AI platforms. Mentions (named in text) and citations (linked) are tracked separately.
• It’s not traditional SoV reskinned. Traditional SoV = keyword-rank × CTR × volume; AI SoV = mention/citation rate. Different signals, different levers.
• Brand presence drives it, not keywords. Correlations with AI visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals.: branded web mentions 0.664, branded anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. 0.527, branded search 0.392, Domain Rating 0.326. About 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. come from earned media.
• The prompt pool is the measurement (Birkett). 100–200 minimum, 250–500 for serious tracking, built from real buyer language (sales calls, Reddit, G2, PAA), refreshed quarterly.
• Platform variance is large. 35% in Gemini vs. near-zero in ChatGPT for one brand; Perplexity/Copilot link in 77%+ of responses vs. ChatGPT’s ~31%. Track 3+ platforms.
• Benchmarks (Birkett): <20% visibility problem, 20–50% established, 50–90% category leader, >90% dominant. Average mention rate ~17.2%.
• It’s volatile: 40–60% citation churn monthly; use a 12-week-plus window.
• The caveat: SoV alone is a vanity metric (DeMott). Connect it to referral traffic and pipeline, or it’s slideware.

Official documentation

There’s no official “AI Share of VoiceAI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric.” doc — the search engines don’t publish a metric for their own AI surfaces. But what drives AI inclusion overlaps heavily with the brand and quality signals they do document. These are the primary sources worth groundingGrounding is anchoring an AI model's answer to source documents it retrieves at the moment you ask — not to the patterns frozen into its weights during training. Retrieval-Augmented Generation (RAG) is the most common way to do it. an AI SoV strategy in.

Google

• Creating helpful, reliable, people-first content — the E-E-A-T and quality guidance that underpins what gets surfaced, including in AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index..
• AI features and your website — Google’s own page on how AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. and AI Mode use content and how to be eligible.
• In-Depth Guide to How Google Search Works — the crawl → indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. → serve foundation AI Overviews are built on.
• Search Essentials — the baseline requirements for being eligible to appear at all.

Bing / Microsoft

• How Bing delivers search results — relevance, quality, and authority signals behind Bing and Copilot answers.
• Bing Webmaster Guidelines — Bing’s stated quality and credibility signals (note: Bing Webmaster ToolsMicrosoft's free portal for monitoring and improving how a site appears in Bing search — the peer to Google Search Console, plus IndexNow instant indexing, richer backlink data, and keyword volumes. Because Bing's index also feeds Microsoft Copilot, it doubles as a window into AI-search visibility. does not currently surface AI citationAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. or mention data, so Copilot SoV has to be measured with third-party tools).

Quotes worth keeping

The most useful statements on AI Share of VoiceAI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric. come from practitioners and researchers rather than search-engine reps — the engines don’t comment on their own AI SoV. These are the verbatim quotes I’d put in a deck, with sources.

On building the prompt pool

• “The prompt set IS the measurement. Its construction determines validity.” — Alex Birkett, Omniscient Digital. Source

On the vanity-metric trap

• “Citation share, presence rate, and AI OverviewAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. appearance counts are the new domain authority. They look defensible in a slide. For 95% of the agencies selling them, they aren’t connected to pipelines in any rigorous way.” — Paul DeMott, Search Engine Land, May 2026.
• “SOV alone is a vanity metric.” — Paul DeMott, Search Engine Land, May 2026.

On volatility

• “Almost every GEOGenerative Engine Optimization (GEO) is the practice of optimizing content and brand presence so AI-powered search engines and assistants — Google AI Overviews, ChatGPT, Perplexity — cite, recommend, or mention you when generating answers. Google's position is that it's still SEO. response is different from every other GEOGenerative Engine Optimization — visibility inside AI answer engines. response.” — Rand Fishkin (via Greg Jarboe, Search Engine Land).
• “40–60% of cited sources change month-to-month.” — eMarketer (via Greg Jarboe, Search Engine Land).

On how AI picks brands

• “AI doesn’t discover new brands — it selects from known entities.” — Carolyn Shelby.

On what moves the needle

• “74% of users select the top recommendation” — making mention order the most commercially loaded signal in AI SoV. — Wasim Kagzi, Search Engine Land, April 2026.

Build-your-prompt-pool checklist

The prompt pool is the measurement, so this is where the work is. Run it in order:

• Define category prompts — the unbranded questions buyers ask about your category (“best CRM for small business”, “how to do X”). These are where AI decides who to name.
• Add branded prompts — questions that name you and your competitors directly (“Asana vs. Monday”, “is [brand] good for enterprise”).
• Add competitive prompts — questions designed to surface the rival set (“alternatives to [competitor]”, “[competitor] vs. who”).
• Validate with search data — cross-check prompts against People Also Ask, autocomplete, and real query data so you’re measuring questions people actually ask, not internal jargon.
• Run across 3+ AI platforms — at minimum ChatGPT, Gemini, and Perplexity; add Copilot and AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index. for a real picture. Platform variance is large.
• Record results — capture mention rate, citation rate, citation position, and sentiment for your brand and each tracked competitor.
• Calculate SoV — brand mentions ÷ total relevant responses × 100, per platform and blended.
• Set a baseline — your first run is the line everything else moves against. Don’t react to it; record it.
• Track weekly/monthly — re-score on a regular cadence; refresh the prompt pool quarterly. Hold for a 12-week-plus window before trusting any trend.

The AI SoV methodology

A repeatable framework for standing up AI Share of VoiceAI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric. tracking from scratch.

1. The formula

AI SoV = (brand mentions ÷ total relevant responses) × 100

Track mentions and citations as separate sub-signals — they diverge, and the gap between them tells you whether you have a presence problem or a linking problem.

2. Prompt pool size

• 100–200 prompts — minimum for viable tracking.
• 250–500 prompts — for serious competitive analysis.

Quality over quantity: a tight 150-prompt set built from real buyer language beats a sloppy 500.

3. Sources for prompts

Pull from where buyers actually talk: sales call transcripts, support tickets, G2 reviews, Reddit threads, People Also Ask, and autocomplete. Internal jargon is the single biggest cause of invalid prompt pools.

4. Platforms to track

Minimum three; ideally ChatGPT, Gemini, Perplexity, Copilot, and AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index.. Add AI Mode and Grok where budget allows. Never track just one — variance across platforms is too large to extrapolate.

5. Metrics to capture

For your brand and every tracked competitor:

• Mention rate — was the brand named?
• Citation rate — was a brand URL linked?
• Citation position — first mention? the primary recommendation?
• Sentiment — leader language vs. challenger language.

6. Refresh cadence

• Prompt pool: refresh quarterly (the questions move).
• Scoring: re-run weekly (or at least monthly), and hold a 12-week-plus window before drawing trend conclusions.

7. The honesty layer

Wrap the whole thing in a line to pipeline. SoV is the input metric; referral traffic, conversions, and brand-search trends are how you prove it’s worth tracking.

AI SoV cheat sheet

The formula

AI SoV = (brand mentions ÷ total relevant responses) × 100

Benchmark tiers (Birkett)

Reference point: average brand mention rate ≈ 17.2% (AthenaHQ, 2026). Most brands live in or below the “visibility problem” band.

Reading a competitive SoV breakdown

A category snapshot across a 1,500-prompt pool might look like this:

SoV is roughly zero-sum within a pool — when a competitor gains share, someone loses it. Watch the gaps, not just your own number.

Don’t-forget facts

• Track 3+ platforms — 35% in Gemini vs. near-zero in ChatGPT for one brand.
• Platform link rates differ: Perplexity/Copilot 77%+, ChatGPT ~31%.
• 40–60% of citations churn month-to-month → use a 12-week window.
• Branded web mentions (0.664) drive AI SoV; Domain Rating (0.326) barely.
• 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. come from earned media, not your own pages.
• SoV alone is a vanity metric — always report it next to traffic/pipeline.

Changing the prompt pool and calling the result growth

The prompt set is the measurement frame. Adding easier branded prompts or removing weak categories changes the denominator and can manufacture an improvement. Version the pool and restate the baseline when it changes.

Combining platforms into one unexplained score

Visibility can differ sharply by product, model, retrieval mode, geography, and date. Report platform-level results before an aggregate and document any weighting.

Treating mentions, citations, prominence, and sentiment as interchangeable

A brand named last in a list, recommended first, cited as a source, or criticized has different value. Store those fields separately rather than letting one count imply all of them.

Reporting share of voice without business context

A higher SoV can come from irrelevant informational prompts and produce no qualified demand. Connect cohorts to audience, referral visits, assisted conversions, pipeline, or another suitable outcome without claiming causation from visibility alone.

Prompt: audit an AI SoV prompt pool

Audit this AI share-of-voice prompt pool for measurement bias. Classify each prompt by
journey stage, topic, audience, branded/unbranded status, geography, and likely intent.
Flag duplicates, leading wording, prompts no real buyer would ask, overrepresented
topics, missing high-value categories, and prompts that name one competitor unfairly.
Do not rewrite the baseline silently: return proposed additions/removals as a new
version with the impact on comparability explained.

Prompt pool:
[PASTE PROMPT ID | PROMPT | CURRENT CATEGORY | SOURCE]

Prompt: code captured answers consistently

Using only these captured answers, code each tracked brand for mention present,
citation present, recommendation/prominence position, sentiment with quoted evidence,
and cited canonical URL. Do not infer a mention from a citation tray or a citation from
unlinked answer text. Return an uncertainty field for ambiguous aliases and preserve
the prompt ID, platform/model, date, mode, and run ID.

Brand alias dictionary:
[PASTE BRAND | APPROVED ALIASES]

Captured answers:
[PASTE RUN METADATA AND COMPLETE ANSWERS]

Test yourself: AI Share of Voice

Resources worth your time

My speaking

• GEO? AEO? LLMO? What’s With All This AI SEO Stuff? — my Ahrefs Evolve 2025 talk on measuring brand visibility in AILLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals., including my discussion of Ahrefs’ correlation data.
• Slides on Speaker Deck — the deck version, if you’d rather skim.

Ahrefs Brand Radar

• Do Brand Mentions Correlate With AI Overview Visibility? — the original 75,000-brand study by Louise Linehan and Xibeijia Guan behind the 0.664 branded-mention correlation.
• Brand Radar — Ahrefs’ AI visibility tracker. It tracks SoV, mentions, citations, and impressions across ChatGPT, Perplexity, Gemini, Copilot, AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index., AI Mode, and Grok, backed by 350M+ search-derived prompts.
• Brand Radar methodology — how the prompt set is built (People Also Ask plus semantic fanout) and per-platform query volumes and cadence. Worth reading before you trust any AI SoV number — your own or a tool’s.

From others

• How to Measure AI Share of Voice (+ 3 Tools) — Alex Birkett’s methodology piece and the source of the benchmark tiers.
• AI Share of Voice: The New SEO Battleground — BirdEye’s overview of AI SoV as a competitive metric, with practical framing for brand managers.
• AI Answer Visibility and Share of Voice Benchmark — NeuralAdX/Spotlight benchmark data from 2.4M+ AI responses, including the 77% vs. 31% citation inclusion rate by platform.
• Share of Voice: Definition, Measurement and Benchmarks — LLMA large language model (LLM) is a deep-learning model trained on massive text corpora to predict the next token and generate human-like text. LLMs use the transformer architecture and power AI search features like Google's AI Overviews (Gemini) and Bing Copilot (GPT-4). Pulse’s glossary entry covering definition, measurement methodology, and benchmark ranges.

Stats worth citing

• Branded web mentions are the strongest driver of AI visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals.. Ahrefs’ 75,000-brand study by Louise Linehan and Xibeijia Guan foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore.: branded web mentions 0.664, branded anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. 0.527, branded search volume 0.392, Domain Rating 0.326. Off-page brand presence beats on-page keyword work. Source
• Average brand mention rate ≈ 17.2%. Across relevant prompts — meaning most brands sit in or below the “visibility problem” band. (AthenaHQ, State of AI Search 2026.)
• 40–60% of cited sources change month-to-month. Citation churn is why a single-month snapshot misleads and a 12-week window is the floor. (eMarketer, via Greg Jarboe, Search Engine Land.)
• Platform variance is enormous. One brand: 35% SoV in Gemini vs. far lower in ChatGPT; 27% citation rate on Grok vs. 0.59% on ChatGPT. Same brand, same content.
• Citation inclusion is structural. Perplexity and Copilot include external links in 77%+ of responses; ChatGPT in roughly 31% (Spotlight analysis, 2.4M+ responses).
• 82–89% of AI citationsAn AI citation is the visible source link an AI answer engine shows next to its generated text — the clickable reference that credits the web page it used. A citation's presence is a separate thing from whether the cited page actually supports the statement, and from being retrieved (read behind the scenes) or merely mentioned (named without a link); citation is driven more by brand mentions and being retrievable than by traditional ranking. come from earned media, not brand-owned pages — digital PR and entity presence precede AI selection.
• AI referral converts well but is small in volume. AI referral traffic converts at roughly 9× the rate of organic Google traffic per session, but total AI click volume is still small for most brands — which is exactly why SoV is a visibility signal, not a traffic guarantee.
• Birkett benchmark tiers: <20% visibility problem, 20–50% established, 50–90% category leader, >90% dominant.
• B2B SaaS benchmark (DerivateX, 2026 study, 50 firms, 1,400 prompts): average AI Presence Score of 56.9/100; 44% of firms scored under 50 — a majority of established B2B brands are below the midpoint.
• Top-quartile vs. bottom-quartile citation gap (Data-Mania, 2026): top-quartile SaaS sites are cited ~31× per month vs. 3.7× for the bottom quartile — an 8.4× gap between winners and the rest.

The standing KPI for AI visibility

Share of voice is the metric this whole topic is built on — but measuring it honestly means pinning down what you count and accepting that the underlying answers are non-deterministic. The number below is a trend line, not a precise gauge.

Citation / mention share against a competitor set

• Metric — Your brand’s citations (or mentions) across a fixed set of prompts, divided by the total citations for you plus a defined competitor set — the AI-era analog of ranking share.
• What it tells you — How often assistants surface you versus rivals for the questions that matter to your category, rather than whether you’re merely present at all.
• How to pull it — Define a stable prompt set, run it across the assistants on a fixed cadence, and log which brands/domains get cited; the Citation Gap Checker surfaces where competitors are cited and you are not, so you can turn the gap into a content target. Honesty caveat: outputs are non-deterministic and personalized, so hold the prompt set and cadence constant and read the result as a trend, not a to-the-decimal figure.
• Benchmark / realistic range — There’s no defensible universal benchmark — the third-party figures cited in the Stats tab are other people’s samples, not your target. Establish your own baseline against a named competitor set and track movement from there.
• Cadence — Monthly, or after a major model release (which can reshuffle citations). Expect real run-to-run variance, so trust the direction over several reads rather than any single snapshot.

AI Share of Voice

AI Share of Voice (SoV) measures how often and how prominently a brand appears in AI-generated responses relative to competitors, across a defined pool of relevant prompts. It's a visibility signal, not a traffic or revenue metric.

Related: LLM Visibility, Measurement and Reporting, AI Traffic Attribution

AI Share of Voice

AI Share of Voice (SoV) is your brand’s slice of the AI-answer conversation. Where traditional SoV is built from keyword rankings weighted by CTR and search volume, AI SoV is built from how often your brand gets mentioned (named in text) or cited (linked) across a sampled pool of prompts run through AI systems like ChatGPT, Gemini, Perplexity, Copilot, and Google’s AI OverviewsAI Overviews are the AI-generated summary box Google shows above or within its regular search results, written by Gemini models from pages retrieved out of Google's normal Search index. It's a Search feature, not a separate platform or index..

The formula is simple: brand mentions ÷ total relevant responses × 100. The hard part is everything around it — the prompt pool you measure against is the measurement, so a poorly built set produces meaningless numbers. Most tools track mentions and citations separately, since AI can recommend a brand without linking to it.

Three things make AI SoV its own discipline rather than a reskin of traditional SoV:

• Different signals drive it. In my correlation work, branded web mentions (0.664) and branded anchor textAnchor text is the visible, clickable text of a hyperlink. It tells readers what they'll find on the other end and gives search engines context about the linked page. (0.527) predict AI visibilityLLM visibility (or AI visibility) is the aggregate measure of how often and how prominently a brand or page shows up in AI-generated answers — across AI Overviews, ChatGPT, Perplexity, Copilot, and Gemini. It's the AI-search analog of organic visibility, but it's driven by different signals. far more than Domain Rating (0.326). Off-page brand presence beats on-page keyword optimization.
• It’s volatile. 40–60% of cited sources change month-to-month, so single-month snapshots mislead — you need a 12-week-plus window for real trends.
• It varies wildly by platform. The same brand can hit 35% SoV in Gemini and a fraction of that in ChatGPT. Tracking one platform gives you systematically biased data; track at least three.

And the caveat that matters most: AI SoV alone is a vanity metric. It tells you about visibility, not pipeline. Measure it alongside referral traffic and revenue, not instead of them.

Related: LLM Visibility, Measurement and Reporting, AI Traffic Attribution

Changelog