How we calculate AI visibility

Every metric on the Cited Index in plain English. Hover the column headers or the brand detail panel labels on the dashboard for the one-line definition; deeper explanations live below.

The Cited AI Visibility Score measures how frequently a brand appears in AI-generated answers to non-branded consumer-intent prompts across 5 major AI platforms: ChatGPT, Perplexity, Gemini, Google AI Overviews, and Google AI Mode. Scores range from 0 to 100, where 100 means the brand is mentioned in every AI response across every platform. The score is calculated as a simple average of per-platform mention rates, with each AI platform weighted equally.

Brands are discovered from AI responses rather than pre-selected, so the index reflects which brands AI actually recommends. The August 2026 edition carries 339 ranked brand entries across 12 categories in India. A brand that competes in more than one category is ranked separately in each, so the count of distinct companies is lower than the count of entries.

Every brand is measured across 5 AI platforms, weighted equally:

Every prompt is written the way a real shopper with no brand in mind would type it into ChatGPT. No brand names, no marketing jargon, no leading language.

We use 6 consumer-language prompt archetypes, each reflecting a different way real shoppers ask AI for help:

Quality control: Every prompt is tested against a banned vocabulary list (no marketing jargon like “D2C”, “premium segment”, “value proposition”) and scored for consumer authenticity: “would a real person actually type this?”

Each category runs between 19 and 29 prompts, and every prompt is run once on each of the 5 platforms. The smallest category in this edition is Higher Education at 19 prompts and 95 responses; the largest run 29 prompts and 145 responses.

Four steps:

1. Run all category prompts across 5 AI platforms
2. Parse every response and extract every brand entity mentioned (LLM-driven discovery)
3. For each brand, calculate mention rate per platform
4. AI Visibility Score = simple average across all platforms

Display threshold: a brand must reach a 5% mention rate on at least 2 of the 5 platforms to appear in the index. Every one of the 339 entries in this edition clears it.

Sentiment is classified from how AI describes the brand. Positive means AI uses favourable language, mixed means a blend of pros and cons, and negative means AI flags concerns. Sentiment is aggregated across all mentions across all platforms.

Position measures where in the AI response a brand appears. #1 means it's the first brand mentioned. Lower is better. Position measures recommendation priority. Being mentioned first carries significantly more weight than being listed fifth.

Framing tags are the specific words AI uses to describe a brand (for example “stylish”, “affordable”, “durable”). The index publishes up to five per brand, for every ranked brand. Which prompts produced them, and how the framing moves over time, live in the paid platform.

Every two months. Each edition runs all prompts fresh across all 5 AI platforms, discovers brands from the new responses, and recalculates every score. New categories are added as we build the prompt sets for them. Edition-over-edition comparisons resume from the August 2026 edition. June was a basis-change edition, moving from 4 platforms to 5, so deltas are deliberately not shown for it.

Brands are discovered automatically from AI responses. None are manually added, and none are sponsored. Any brand that meets the display threshold (a 5% mention rate on at least 2 platforms) appears in the index.

If your brand is missing, it means the AI platforms are not naming you in answers to consumer prompts in your category, or your category is not one of the 12 we cover yet. A free AI Visibility Report shows you which platforms are silent on you and which pages they cite instead.

The Cited AI Visibility Score does not measure:

• Purchase conversion or click-through rates
• Paid ad placements within AI responses
• Regional variations within India (national average only)
• Real-time changes (each edition is a two-month snapshot)

AI responses are non-deterministic. The same prompt can produce a different answer on a different day. Each prompt is run once per platform, so confidence comes from the number of prompts in a category and from aggregating across all 5 platforms, not from repeat runs of the same prompt. A single brand-prompt result is noisy; a category-level score built on 95 to 145 responses is not.

The index reflects consumer-intent prompts. Enterprise B2B prompts or highly technical searches may show different brand distributions.