Glossary — primary metric
Share of paragraph.
The AI visibility primary metric, defined.
Share of paragraph is the percentage of an AI assistant's response devoted to a specific brand, computed across a locked panel of buyer-intent prompts. It's the AI-search analog to share-of-voice in traditional media — and the primary KPI tracked across the AI visibility tooling category.
Last reviewed: May 2026
Definition
What share of paragraph measures.
For a given AI assistant response to a buyer-intent prompt, share of paragraph is the proportion of the response's content (typically measured in words or sentences) that references a specific brand. Averaged across a panel of category-relevant prompts, it captures both how often the AI mentions your brand AND how much of the AI's reasoning your brand occupies when mentioned.
The metric became standard in 2024-2025 as the AI visibility tooling category emerged and is now the primary KPI tracked across measurement platforms including Profound, Peec AI, OtterlyAI, Evertune, AthenaHQ, Bluefish, and Lynceus.
Formula
How to calculate it.
Per-prompt share = (words referencing brand) ÷ (total response words) × 100
Aggregate brand share = average per-prompt share across prompt panel
Per-assistant share = aggregate brand share, segmented by ChatGPT, Claude, Gemini, Perplexity
Variations exist — some implementations weight by mention position (first-mentioned scores higher), some use sentence count rather than word count, some include implicit pronoun references. The core formula is consistent across the category. For the complete measurement methodology see how to measure AI visibility.
Benchmarks
What scores actually mean.
Share of paragraph benchmarks are highly category-dependent. Some directional anchors:
- →Concentrated categories (3-5 named competitors): leader holds 25-40% share; #2 holds 15-25%; rest split 5-15%.
- →Fragmented categories (10+ named competitors): leader holds 15-20%; individual competitors hold 3-8%.
- →Emerging categories (no clear leader): top 5 brands cluster at 8-15% each.
The more actionable benchmark is your specific category's leader and your 2-3 most-named competitors. Track quarter-over-quarter movement against them rather than against absolute thresholds.
Frequently asked questions
What is share of paragraph?
Share of paragraph is the percentage of an AI assistant's response devoted to a specific brand, computed across a locked panel of buyer-intent prompts. It is the primary metric for AI visibility — the AI-search analog to share-of-voice in traditional media. A brand at 25% share of paragraph in its category occupies meaningful real estate in AI buyer evaluations; a brand at 3% is structurally absent from the conversation. The metric was popularized in 2024-2025 as AI visibility tooling emerged and is now the standard primary KPI across measurement platforms including Profound, Peec AI, OtterlyAI, AthenaHQ, Evertune, Bluefish, and Lynceus.
How is share of paragraph calculated?
For each AI response in your prompt panel, count the words, sentences, or characters that reference your brand specifically (named mentions, pronoun references to your brand, product names attributed to your brand). Divide by the total response length to get the per-prompt share. Average across the prompt panel to get the aggregate brand share of paragraph. Variations exist — some implementations weight by mention position (first-mentioned scores higher), some use sentence count rather than word count — but the core formula is consistent across the category.
Why is share of paragraph the right primary metric?
Three reasons. First: it captures both whether you're mentioned AND how much of the AI's reasoning your brand occupies — a mention buried in a 500-word response is different from a mention that drives the recommendation. Second: it's comparable across competitors, across prompts, and across time in a way that binary 'mentioned vs not mentioned' metrics aren't. Third: it correlates with downstream buyer behavior more strongly than simpler metrics — buyers who read an AI response where your brand has high share of paragraph convert at higher rates than buyers who read responses where you're a passing mention.
What's a good share of paragraph score?
Category-dependent. In categories with 3-5 named competitors and a clear leader, the leader typically holds 25-40% share of paragraph; #2 holds 15-25%; the rest split 5-15% each. In fragmented categories with 10+ named competitors, the leader may hold only 15-20% and individual competitors 3-8%. Benchmarking against your specific category's leader and your 2-3 most-named competitors is more useful than absolute targets. Track quarter-over-quarter movement rather than absolute thresholds.
Does share of paragraph differ across AI assistants?
Yes — meaningfully. ChatGPT, Claude, Gemini, and Perplexity weight different signals and produce different paragraphs for the same prompts. A brand can hold 30% share of paragraph in ChatGPT and 8% in Claude because Claude weights primary-source content more heavily. Brands should track share of paragraph separately for each assistant and aggregate weighted by audience usage. See /how-chatgpt-cites-brands, /how-claude-cites-sources, /how-perplexity-works, /how-gemini-cites-sources for the per-assistant mechanics.
How is share of paragraph different from share of voice?
Share of voice measures brand presence in traditional media (press mentions, ad impressions, social mentions). Share of paragraph measures brand presence in AI assistant responses — a discrete, structured surface with much more constrained word counts. Share of voice operates across an unbounded media landscape; share of paragraph operates within finite AI responses where every word competes for attention. The metrics serve similar strategic purposes but are not interchangeable measurements.
Can share of paragraph be gamed?
Not easily, by design. Share of paragraph reflects what AI assistants choose to cite when constructing responses. Brands cannot directly modify AI training data or live retrieval results. The only legitimate way to move share of paragraph is to publish more authoritative, entity-rich, structurally-clean content on your topic so AI assistants have stronger reasons to cite your brand. Attempts to game the metric (keyword stuffing, fake reviews, AI-detected manipulated content) typically backfire — modern LLMs deprioritize obviously manipulated sources.
Measure yours
Your current share of paragraph
in three minutes.
The free Lynceus AI Visibility report computes your share of paragraph across ChatGPT, Claude, Gemini, and Perplexity for three category prompts. Real baseline, no signup gate.