AI Share of Voice: The 2026 Definition and Formula

What AI share of voice means

AI share of voice (AI SOV) is the percentage of category mentions inside AI assistant answers that belong to your brand. If your brand appears in 18 of 50 prompts run across ChatGPT, Gemini, Claude, Grok, and Perplexity, your AI SOV for that prompt set is 36%. The number captures how often your brand is named when buyers ask the questions that matter for your category.

The term is contested in 2026. Ahrefs uses "AI share of voice" in Brand Radar to describe a percentage of brand impressions across tracked responses. HubSpot's AEO Grader uses it more loosely. About twenty smaller sites have definitional pages. Nobody has won the term yet, which means there's room for the cleanest definition to stick. This is ours.

Where AI share of voice comes from

AI SOV isn't a new invention. It's the next link in a chain of metrics that go back decades.

Classic share of voice was an advertising measurement: your media spend divided by total category media spend, times 100. Nielsen, Brandwatch, Sprout Social, and Talkwalker all carry roughly this definition. It captured how loud your brand was in the broadcast era.

Share of voice in social listening broadened the concept to share of conversation: your brand mentions across social and earned media, divided by category mentions. Same idea, different surface, no longer just paid.

Share of search (Les Binet, IPA, EffWorks 2020) proposed branded query volume divided by total category query volume as a digital era proxy. Binet found that share of search tracks share of market reasonably tightly over steady periods, which made it the most useful brand health metric of the 2020s.

AI share of voice is the next link. It captures presence inside the synthesized answers that increasingly replace search engine result pages. Where share of search measures who customers are searching for, AI SOV measures who AI assistants are mentioning when customers ask the category's questions. The chain is paid spend (classic SOV) -> earned conversation (social SOV) -> branded query volume (share of search) -> category answer presence (AI SOV).

How to actually calculate it

Four approaches show up in the wild. Each gives a slightly different read.

1. Mention based. Your brand mentions divided by total category mentions across the prompt set, times 100. Simple, but treats first mention the same as fifth.
2. Query based. Number of prompts where your brand appears divided by total prompts, times 100. The Ahrefs framing. "Appear in 18 of 50 prompts equals 36% SOV." Doesn't capture depth of mention.
3. Word count based. Share of the answer's words that refer to your brand. Used by a minority of tools. Valuable because it captures how much narrative the model spent on you.
4. Position weighted. Each mention scored on whether it was first, in the body, or buried, then summed. Best signal, more work to compute.

The reason we combine query based with position weighting: it's the version that survives scrutiny in both directions. Mention rate alone undervalues being the model's first pick. Pure word count overrewards verbose answers. Position weighting captures the thing buyers actually care about (was I named first, or did the model spend three paragraphs on my competitor before getting to me), without making the metric so complicated that nobody can reproduce the math.

Why you report per engine, not just aggregate

Cross engine citation overlap of cited domains is only about 11% in 2026 data. The brand that dominates Perplexity (which leans heavily on Reddit) is often invisible in ChatGPT (which leans on Wikipedia and trained knowledge), and the reverse. A single aggregate AI SOV number hides this picture entirely.

The right reporting pattern: track AI SOV per engine, present a weighted aggregate to executives (weighted by category traffic share if you have it, by user share if you don't), and use the per engine breakdown to decide where to invest. A brand at 8% AI SOV in ChatGPT but 41% in Perplexity has a fundamentally different problem than a brand at 24% across all engines, even if the aggregate average looks similar.

What counts as a good AI SOV

Honest answer: nobody knows yet. The category is young, no one has run rigorous benchmark studies the way Binet did for share of search, and the per category variance is wide. The heuristics in use today:

• Above 30% in your category: dominant. Your brand is the model's default answer.
• 10% to 30%: competitive. You're in the conversation but not winning it.
• Below 10%: at risk. The model has decided other brands are the answer in your category, and you're going to lose buyers you don't know you had.

The more useful framing is relative rather than absolute. Track your AI SOV against your top three competitors over time. A rising number against a falling competitor is the signal that matters, regardless of which side of 10% or 30% it sits on.

Real category examples

A few categories where the AI SOV pattern is recognizable enough to use as a reference point.

Medical alert systems

Bay Alarm Medical, Medical Guardian, MobileHelp, Lifeline by Philips, and Life Alert split most of the AI SOV in this category. Bay Alarm and Medical Guardian usually lead in ChatGPT and Perplexity (both have strong editorial coverage and Reddit presence). Life Alert is often invisible inside AI answers despite massive offline brand recognition, because its content footprint is thin and the model doesn't have much to draw on. High offline brand awareness does not translate automatically into AI SOV.

CRM

HubSpot dominates SMB queries inside ChatGPT and Gemini. Salesforce dominates enterprise queries. Pipedrive, Zoho, and Monday each cluster below 10% AI SOV in most prompt sets, despite all three being well known products. The split between SMB and enterprise AI SOV is the most important read in this category.

Project management tools

Asana, Monday, ClickUp, Notion, and Trello split most of the visibility. Linear punches above its size: lower category market share than the leaders, higher AI SOV than its market share predicts. The combination of strong editorial coverage and an active developer audience on the sources LLMs cite (Reddit, YouTube, Hacker News, X) produces an outsized AI SOV.

Adjacent metrics and how they relate

Several closely related metrics show up in the literature. Worth knowing where AI SOV sits among them.

• Citation share: the percentage of citation links (URLs) in AI answers that come from your domain. Narrower than AI SOV; useful for content teams targeting the citation graph.
• Brand impression share: AI SOV weighted by visibility context. A 30% mention rate in highly visible positions is worth more than a 30% mention rate buried at the end of answers.
• Answer share: used loosely as a synonym for AI SOV, sometimes restricted to whether your brand is named in "best of" listicle style answers.
• Share of model: per engine slice of AI SOV. Your ChatGPT share, your Gemini share, your Perplexity share. The breakdown that AI SOV rolls up.

Position AI share of voice as the umbrella metric. Treat citation share and answer share as components. Share of model is the per engine view inside it.

Frequently asked questions

The questions we hear most when teams start measuring AI share of voice.