The AI Visibility Audit: How to Measure Your Brand's Score
A Mumbai B2B SaaS brand ranked well on Google and pulled in 4,200 organic visits that month. Across twelve AI queries, it appeared twice, both times near the bottom of the list.
AI visibility is not one number. It is four separate signals, and a high score on one engine predicts almost nothing about the others.
A brand manager at a Mumbai-based B2B SaaS company ran a simple experiment last quarter. She typed her company's category into ChatGPT, Perplexity, and Google AI Mode, twelve queries in total, and recorded what came back. Her brand appeared twice. Both times it sat near the bottom of a multi-competitor list. Her SEO dashboard showed 4,200 organic visits that same month. The AI visibility audit told a different story. Magnent works with Indian brands facing this gap every week: strong traditional search performance masking near-zero presence in the AI-generated answers where B2B buyers build their vendor shortlists.
Measuring AI visibility requires tracking four distinct metrics: mention rate, citation rate, citation position, and share of voice, across a structured prompt set on multiple AI engines. A high score on one platform does not predict performance on another. Magnent's AI visibility audit process gives Indian B2B brands a per-engine breakdown rather than a single misleading composite.
What's the difference between being mentioned and being cited in an AI answer?
Most brands conflate the two, and the confusion drives the wrong fix. A mention means the AI named the brand somewhere in its response. A citation means the AI attributed a specific claim, fact, or recommendation to the brand and pointed to a source. Both matter, but they are different signals that require different interventions.
| Signal | Definition | What a low score indicates |
|---|---|---|
| Mention rate | % of prompts where your brand appears anywhere in the answer | Weak brand awareness in the AI's category knowledge |
| Citation rate | % of prompts where the AI links to your content as a source | Content lacks the authority signals AI engines trust |
| Citation position | Whether your brand appears first, mid-list, or last | Strength of recommendation vs mere acknowledgement |
| Share of voice | Your mentions as a fraction of total brand mentions | Competitive positioning within AI-generated answers |
A brand can have a high mention rate and a low citation rate, meaning the AI recognises the company but does not treat it as an authoritative source. That gap points toward third-party authority deficits, not content volume. Conversely, a brand with strong citation rates on three prompts but zero presence on unbranded category queries has a different problem: it surfaces when asked about directly, but stays absent for the queries buyers actually run before they know which brands to compare.
The AI visibility audit Magnent runs in 30 minutes separates these four signals in the opening pass, because treating a citation-rate problem as a mention-rate problem produces effort with no measurable outcome.
Which AI engines actually need to be included in the audit?
Not all engines behave the same way, and optimising for one does not transfer to another. Reddit accounts for up to 46.7% of Perplexity's citations and only around 0.1% of Gemini's, based on citation analysis data (Tinuiti, Q1 2026). A brand that wins Perplexity through community discussion can be invisible on Gemini simultaneously, and vice versa.
The minimum useful set for an Indian B2B brand in 2026:
| Engine | Why it matters | Primary citation behaviour |
|---|---|---|
| ChatGPT (web browsing on) | Highest query volume; strong on comparison and category queries | Trained knowledge plus live retrieval; weights long-form editorial |
| Perplexity | Most citation-heavy engine; strong real-time retrieval | Heavy community source weight: Reddit, Quora, forums |
| Gemini | Critical for Google Workspace users; common in enterprise B2B | Anchored in Google-indexed URLs and structured data |
| Google AI Mode | Conversational layer over Google Search; rising fast in India | Blends organic rankings with trained AI responses |
Running the audit on ChatGPT alone produces a single-platform view. Brands frequently score above 80% on one engine and below 30% on another, because the retrieval architectures differ structurally. Magnent's analysis of AI visibility trends across Indian brand engagements from January to April 2026 shows per-engine divergence as the most consistently actionable finding in any first visibility audit.
How many prompts do you actually need to get results you can trust?
The minimum defensible prompt set for an Indian B2B brand is 20 prompts, each run three times per engine, producing 60 observations per platform. That volume filters out single-answer variance without becoming an unmanageable exercise.
Structure the 20 prompts across three categories:
- Category queries (10 prompts): Questions a buyer types when researching without a brand in mind. For example: "which project management tools work best for Indian engineering teams."
- Comparison queries (6 prompts): Direct "brand A vs brand B" questions. AI engines cite brands most reliably on these, making them the clearest diagnostic for share of voice.
- Branded queries (4 prompts): Explicit questions about the brand. Use these for accuracy checking, not visibility scoring.
The three-repetition discipline matters because every LLM answer is probabilistic. A single cold query produces noise, not signal. One-shot checks frequently generate misdiagnoses: a poor result on a single branded query gets treated as a category presence problem when it is actually a fact-accuracy issue in the model's training data.
One non-obvious structural finding from citation research across Indian B2B brands: intent queries such as "best procurement software for Indian mid-market companies" return near-zero citation rates for most brands across all major AI engines. These are the queries buyers type most naturally, but they are also the hardest to win. A brand scoring 0% on intent queries but 60% on comparison queries faces a different strategic priority than one scoring uniformly low across all three prompt types.
B2B buyers in India increasingly turn to AI assistants in the early stages of vendor research, which means the prompt set needs to map to how real buyers phrase those queries, not how a brand's marketing team would write them. The 2026 AI and AEO visibility trends report from Magnent covers how Indian brands are adapting their prompt sets as AI search behaviour evolves.
Why a brand shows up on Perplexity but not ChatGPT, and what it means
This is the most common finding after a first AI visibility audit for Indian B2B brands, and the explanation is structural rather than tactical.
Perplexity functions closer to a live search retrieval engine: it pulls from Reddit threads, Quora answers, community forums, and current web content in real time. A brand with genuine community presence on Indian B2B forums or active participation in relevant subreddits surfaces on Perplexity without necessarily having strong trained-knowledge representation in ChatGPT.
ChatGPT's category and comparison query behaviour draws more heavily on what it absorbed during training: long-form editorial content, independent review pieces, structured product comparisons, and consistently mentioned brand claims across multiple third-party sources. A brand absent from industry roundups, editorial comparisons, and independent review content scores low on ChatGPT regardless of its Perplexity performance.
The reverse pattern also appears: brands with polished editorial coverage and strong technical SEO score well on ChatGPT and Gemini while remaining invisible on Perplexity because they have no genuine third-party community footprint. An AI visibility audit covering all four major engines makes these platform-specific gaps visible and separable. Without per-engine data, brands typically invest in the wrong channel and wonder why aggregate visibility does not improve.
Research on AI adoption in enterprise buying contexts consistently shows how different AI tools serve different stages of the purchase process, with search-oriented AI tools like Perplexity used more in early discovery and model-oriented tools like ChatGPT used more in structured comparison (McKinsey Global Institute, 2025).
What does a good AI visibility score actually look like?
There is no universal benchmark. Visibility score norms are category-specific: in competitive B2B SaaS verticals, the top-three Indian brands frequently reach 70 to 80% mention rates on comparison prompts; in less-saturated categories, a 40 to 50% mention rate on transactional queries can represent a first-mover position before rivals invest.
The most important threshold is diagnostic rather than comparative. A brand with a mention rate below 30% on category queries and more than 60% of its citations coming from its own domain has a structural problem that more owned content cannot close. Third-party source presence is the underlying gap: coverage on review platforms, genuine community forum participation, independent editorial, and structured data signals.
A non-obvious finding from cross-brand AI visibility analysis deserves attention here. Technical AI readiness, clean crawlability, schema markup, structured content, and actual AI visibility in live answers are different problems with different solutions. A brand can achieve full technical readiness and still score near-zero on visibility, because citation data consistently shows that the large majority of AI citations point to third-party sources rather than the brand's own domain (industry citation analysis, May 2026). Reddit threads, Quora answers, review platforms, news articles, and independent editorial carry the bulk of the citation weight. The brand's website was never going to be the primary citation surface; third-party authority is the work.
The trend across quarterly re-audits is more informative than any single absolute score. A brand moving from 22% to 38% mention rate over two quarters on the same prompt set has demonstrable momentum. A brand holding flat at 55% despite content investment is losing competitive ground as rivals improve.
Frequently Asked Questions
Isn't running 20 prompts across four engines manually too time-consuming for an Indian brand team?
The core 20-prompt set, run three times per engine across four platforms, takes roughly three to four hours for a first audit. Subsequent audits run faster once the prompt set and tracking spreadsheet are built. Most Indian B2B brand teams run a structured audit quarterly and use a simpler five- to eight-prompt spot-check between audits to catch major model shifts without the full exercise.
Can the AI visibility audit be done with automated tools, or does it need manual prompt testing?
Several tools automate parts of the process, platform-specific citation tracking, share of voice scoring, and source attribution. The limitation is that most skip the three-repetition discipline needed to filter out variance, and few cover all four major engines Indian B2B buyers use. A hybrid approach works well: automated tracking for the monthly spot-check, manual structured runs for the quarterly deep audit where per-engine diagnosis is the primary goal.
What is the difference between an AI visibility audit and an SEO audit?
An SEO audit measures how web pages perform in traditional search rankings. An AI visibility audit measures how a brand appears in conversational AI-generated answers across ChatGPT, Perplexity, Gemini, and Google AI Mode, regardless of whether the brand's own pages are the source. A brand can rank on the first page of Google and score near-zero on ChatGPT category queries, because ChatGPT is summarising what third-party sources say about the brand, not what the brand's own pages claim.
Which prompt type produces the most actionable findings for an Indian B2B brand audit?
Comparison queries consistently surface the most actionable findings. AI engines cite brands most reliably on direct "A vs B" questions, making these the clearest diagnostic for share of voice and citation position. Category queries reveal reach beyond the brand's known audience but have lower baseline citation rates and require a larger prompt set to produce reliable results.
How long before an AI visibility audit shows measurable change from content and community work?
Most Indian brands working on citation-focused content and off-site community presence should expect 60 to 90 days for the first measurable movement on a 20-prompt audit set, and four to six months for compound improvement across multiple engines. Perplexity and Google AI Mode typically respond faster than ChatGPT to off-site changes, because they rely more heavily on real-time retrieval. Tracking per-engine rather than composite scores surfaces earlier signals of movement.