ChatGPT

How to Measure Share of Voice in ChatGPT and Perplexity

Only 22% of marketers track AI visibility, and most measure just one platform. Here's how to calculate AI SOV across ChatGPT, Perplexity, AI Overviews and AI Mode.

73% of B2B buyers now use AI tools in their vendor research process, according to Averi’s March 2026 analysis of 680 million citations. AI-referred visitors convert at 4.4x the rate of traditional organic traffic, according to Semrush’s June 2025 study. Despite both figures, most marketers still measure a single platform. A brand performing well in ChatGPT can be simultaneously invisible in AI Overviews, Perplexity, and AI Mode.

Key takeaways

  • AI SOV measures how often your brand appears in AI-generated answers relative to competitors
  • A single-platform SOV score misses the full picture across ChatGPT, Perplexity, AI Overviews, and AI Mode
  • Only 22% of marketers track AI visibility, per Loganix’s 2026 analysis
  • AI-referred sessions convert at 4.4x the rate of organic traffic, making AI SOV a revenue metric

When we pull AI visibility data for a new client at FirstMotion, the first thing we look at is the gap between platforms. A brand with strong ChatGPT presence is often entirely absent from AI Overviews and Perplexity for the same prompts: different retrieval models, different citation patterns, different gaps. Our ContextualJourney™ platform is built specifically to surface those gaps across all four engines in a single measurement cycle.

What AI share of voice measures and why it matters

AI platforms now function as answer engines rather than link directories, synthesising recommendations instead of ranking results. Organic search metrics don’t capture presence in AI-generated responses: a brand can rank on page one and still be invisible when buyers ask AI assistants for vendor recommendations. 64% of marketing leaders say they’re unsure how to measure success in AI search today, according to Yext’s October 2025 study of 6.8 million citations.

AI SOV differs from AI visibility score, though the two are related. Visibility score measures how often a brand appears across all queries in a prompt set. SOV adds the competitive dimension: of all the brands mentioned in those responses, what share of total mentions belongs to this brand. The two metrics answer different questions: visibility tells you your reach, SOV tells you your competitive weight within it.

How AI models process brand signals

AI models don’t score content by links and authority metrics the way search engines do. They weigh the breadth and consistency of brand representation across sources they’ve learned to retrieve from. The result is that a brand’s AI share of category responses reflects its entire digital footprint. AI SOV = (Brand citations / Total category citations) × 100.

To calculate AI SOV accurately, the measurement needs to reflect the signals AI models actually use, rather than SEO proxies. Domain rating (DR) has limited predictive value for a brand’s AI share of category responses. Ahrefs’ analysis of 75,000 brands found branded web mentions correlate with AI Overview visibility at r=0.664, compared to r=0.218 for backlinks. Earned presence across the web outweighs technical site metrics by roughly 3 to 1.

Brands in the top 25% for web mentions earn up to 10x more mentions in AI Overviews than the next quartile, according to Ahrefs’ analysis of 75,000 brands. AI citations concentrate at the top of the mentioned distribution. Brands with the strongest independent third-party presence dominate category responses in AI-generated answers, regardless of domain authority.

How AI search engines measure brand presence

AI search engines don’t rank brands the way traditional search engines rank pages. They synthesise answers from sources they’ve learned to treat as authoritative: editorial coverage, brand mentions across trusted publications, review platform presence, and consistency of representation across independent sources. Earned media composition determines which brands dominate a given category in AI-generated answers.

The practical implication is that a brand’s AI share of voice is downstream of its earned media strategy. Brands that invest in original research, earned editorial coverage, and expert-attributed content build the external signal density AI systems retrieve from. Those that rely primarily on owned content and technical SEO optimization often find their organic search performance doesn’t translate into AI citations.

Why AI visibility differs by platform

The most common measurement mistake in AI SOV tracking is treating ChatGPT data as representative of AI search overall. Each major AI platform sources its answers differently, cites from different content types, and weights brand signals differently. ChatGPT, Perplexity, Google AI Overviews, and Google AI Mode each require separate measurement.

PlatformCitation behaviourKey SOV driver
ChatGPTSynthesises from training data and web retrieval; broad topic coverageBrand mention density across editorial and review sources
PerplexityReal-time web retrieval; favours recent, structured contentCurrent editorial coverage and recency of indexed content
Google AI OverviewsAppears in 20%+ of Google searches; draws from indexed top-10 resultsTraditional SEO combined with E-E-A-T signals
Google AI ModeDeep multi-step reasoning; pulls from broad web sourcesTopical authority across multiple related content clusters

Measuring SOV on only one of these platforms gives a partial picture that can produce misleading conclusions. A brand that invested in earned media coverage across trade publications may perform well in ChatGPT and Perplexity but underperform in AI Overviews if its traditional SEO lags. The reverse is equally common: strong Google rankings don’t automatically translate to strong AI SOV in ChatGPT or Perplexity.

Google AI Overviews and AI Mode

Google AI Overviews and AI Mode are the two most commonly absent from AI SOV measurement programmes, despite being the AI surfaces most buyers encounter first. AI Overviews now appear in more than 20% of all Google searches, according to SparkToro’s 2026 analysis. When they appear, the zero-click rate rises to approximately 83% according to SparkToro and Similarweb data, meaning buyers who see an AI Overview response are highly unlikely to continue to the organic results below.

AI Mode requires separate tracking from AI Overviews: the two surfaces use different retrieval mechanisms and produce different citation patterns. A brand with strong AI Overview presence may have weak AI Mode visibility, and vice versa. Treating Google’s AI surfaces as a single measurement point is as misleading as treating all paid media channels as one number.

Measuring AI share of voice across AI platforms

Building your prompt set

AI SOV measurement starts with a defined set of category-relevant prompts reflecting real user intent: the questions buyers actually ask AI assistants when researching vendors in the brand’s category. A prompt set of 20 to 50 category-relevant queries gives a statistically meaningful AI SOV baseline. The same prompt set must be run consistently across all four platforms to produce comparable figures and a competitor comparison that holds up across engines.

The brand accounts for a share of all mentions generated in those responses. Running fewer than 20 prompts produces results too volatile to act on. A competitor comparison run across different prompts per platform produces numbers that can’t be compared, a common error that makes multi-platform SOV look inconsistent when the inconsistency is methodological.

Calculating your AI visibility score

AI visibility score is the percentage of prompts in the set that produce at least one mention of the brand, per platform. AI Visibility Score = (Prompts mentioning brand / Total prompts in set) × 100. This is the simpler metric to start with before calculating the brand’s AI share across the full competitive set.

Tracking AI mentions per prompt per platform identifies which queries the brand is absent from and which it dominates. A brand appearing in 12 of 30 tracked prompts on ChatGPT has a visibility score of 40%. If direct competitors appear in 25 of the same 30 prompts, the gap in the brand’s AI share is clear without the full SOV calculation.

Measuring AI SOV per platform

For each prompt and platform combination, record which brands appear and at what position in the AI generated responses. AI SOV per platform = (Brand mentions on that platform / Total tracked brand mentions on that platform) × 100. Running this across 30 prompts and four platforms gives 120 data points per measurement cycle.

The per-platform breakdown is more useful than the composite average. It shows exactly where AI-generated content is giving competitors an advantage and where the brand’s AI share is strongest. A brand with 35% ChatGPT SOV and 2% AI Overviews SOV has a different optimization brief from one sitting at 15% on every platform.

Tracking AI brand mentions and sentiment

Reading AI generated answers for brand data

AI SOV measurement captures brand appearance. How the brand is described in those answers is a separate question entirely, and both matter commercially. A brand with 35% SOV consistently described as “expensive” or “complex to implement” has a different problem from a brand with 35% SOV described as “the leading platform for mid-market B2B SaaS.”

Tracking AI brand mentions requires reading the actual responses. Recording which brands appear is only half the picture, so for each prompt where the brand appears, note:

  • Position (first, second, or third brand mentioned)
  • Framing (recommended, mentioned neutrally, or flagged with caveats)
  • Attributes (the specific capabilities or characteristics the AI assigns to the brand)

Structured data, review platform ratings, and schema-marked product descriptions shape what AI talks about when it mentions a brand. Positive reviews on G2 and Capterra contribute directly to the attributes of AI systems surface.

What AI answers reveal about brand positioning

The responses generated by AI platforms for category queries contain the AI’s synthesised description of the brand, and that description is what reaches the target audience during active vendor research. Brands losing ground in AI-generated answers often discover it through how AI talks about them: vague descriptions, outdated positioning, or association with problems.

Brand sentiment in AI answers reflects what AI systems have retrieved and learned from the brand’s digital footprint. If a brand’s content and third-party coverage emphasise one product feature heavily, AI-generated answers will often reflect that emphasis across the entire prompt set. Our earned media guide covers how to build the citation footprint across the publications AI engines weigh most heavily.

AI referral traffic and conversion data

AI referred traffic grew 527% year-over-year in the first five months of 2025, according to Previsible’s AI Traffic Report. AI SOV is the leading indicator; AI referred traffic is the commercial output it drives. AI-referred visitors arrive pre-qualified: the AI assistant has already described the brand and the buyer has chosen to investigate, which is what produces the 4.4x conversion premium Semrush identified across 500+ B2B topics.

Tracking AI referred traffic in Google Analytics requires identifying the AI platform referral sources:

  • chat.openai.com
  • perplexity.ai
  • claude.ai
  • Google AI properties (including AI Overviews and AI Mode referrals)

Free tools like GA4’s channel grouping configuration separate AI referred sessions from direct traffic, which standard setups lump together and understate. A brand with 30 ChatGPT-referred sessions per month converting at 15% is commercially more significant than one with 300 organic sessions converting at 1.5%.

AI SOV benchmarks and tracking cadence

There is no universal AI SOV benchmark because category competitiveness varies significantly. In highly competitive categories with many active players, a 30% AI SOV represents strong positioning. Track SOV relative to three to five direct competitors rather than every brand that ever appears in category responses: the competitive subset gives more actionable signal than a broad category average.

AI SOV rangeInterpretation
0–5%Near-invisible in the category
6–15%Emerging presence; competitors dominate
16–30%Competitive presence; part of the conversation
31–50%Strong AI visibility; consistent reference for category queries
50%+Category leadership; majority of AI answers include the brand

Track the full prompt set monthly. Run high-priority prompts (the highest-intent buying queries) weekly, since AI model updates can shift citation patterns quickly. Monthly audits catch shifts in AI SOV before they compound. Monthly tracking of AI SOV also reveals if content and earned media activity from the previous period is translating into improved visibility.

How to improve AI share of voice

Improving AI SOV requires generative engine optimisation: structuring content and building the external brand signals that improve how AI systems synthesise brand data. The content types that produce the strongest AI SOV gains are detailed how-to guides with direct answers, and original research with primary data: formats that give AI models information gain they can’t reconstruct from other sources. A brand’s content and earned coverage together determine what AI systems learn to say about it.

Our entity authority guide covers how AI systems learn to associate a brand with its category. The digital PR guide addresses the earned media programme that builds the distribution footprint AI engines retrieve from. Structured data implementation, review platform presence, and consistent expert-attributed content all contribute to AI brand mentions improving over time.

If you’re not measuring AI SOV, here’s where to start

Start with a prompt set of 20 to 30 prompts: the entry point of the 20 to 50 range that gives a meaningful AI SOV baseline. Run them across ChatGPT, Perplexity, Google AI Mode, and Google AI Overviews. Record which brands appear in each response. Calculate your visibility score per platform. That’s the baseline.

The gaps become visible immediately, and the gaps are where the work is. A brand that’s invisible in Perplexity but present in ChatGPT has a different programme to build from one that’s invisible everywhere. Measurement is what tells you which problem you have.

Book a free GEO audit to pull your brand’s AI SOV baseline across all four platforms and identify which prompts your direct competitors are winning that you’re not.

FAQ

Frequently asked questions

What is AI share of voice?

AI share of voice (AI SOV) measures how often a brand appears in AI-generated answers relative to all tracked competitors, for a defined set of category-relevant prompts. It's calculated as (Brand citations / Total category citations) × 100. Unlike traditional share of voice, which measures paid media or search ranking presence, AI SOV measures inclusion in synthesised outputs: the answers AI tools give buyers when they ask for vendor recommendations.

How is AI SOV different from AI visibility score?

AI visibility score measures the percentage of tracked prompts that produce at least one mention of the brand. AI SOV adds the competitive dimension: of all brand mentions produced by the prompt set, what share belongs to this brand. A brand can have high visibility and low SOV if direct competitors dominate. Appearing in a single AI response doesn't establish share, both metrics are necessary for a complete picture.

How many prompts do you need to measure AI SOV accurately?

20 to 50 category-relevant prompts gives a statistically meaningful AI SOV baseline for most B2B categories. Fewer than 20 produces results too volatile to act on. Running different prompts per platform produces numbers that can't be compared, which defeats the purpose of measuring across four engines.

Why does AI SOV differ across ChatGPT, Perplexity, and Google AI Overviews?

Each platform sources its answers differently: ChatGPT draws from training data and web retrieval, Perplexity from real-time results with a preference for recent content, Google AI Overviews from indexed top-10 results, and Google AI Mode from multi-step reasoning across broad source pools. A brand optimised for one citation pattern may underperform in others, which is why multi-platform measurement is essential.

How does FirstMotion track AI visibility for clients?

We run structured prompt sets across ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, and Meta AI through our ContextualJourney™ platform, tracking citation rate, share of voice, brand position, and sentiment for each client brand and up to five competitors simultaneously. We run weekly citation rate monitoring, monthly sentiment reviews, and quarterly competitive audits, connecting AI visibility data to pipeline metrics.

Work with us

Ready to be the first answer in your category?

Book a call