agentMention
Blog/AI Search & GEO
Guide · AI Visibility

AI Visibility metrics: how to measure Mention Rate, Citation Rate, Share of Voice and Sentiment

Search no longer ends with a list of blue links. We break down the four metrics that show whether your brand appears in AI answers — and how to measure each one.

AEagentMention editorialMay 28, 2026

Search no longer ends with a list of blue links. In ChatGPT Search, Perplexity, Microsoft Copilot, Google AI Overviews and AI Mode, the user often sees a ready answer, a comparison, a recommendation or a short list of options before ever visiting a website. For a marketer this means one simple shift: traffic is no longer the only way to tell whether your brand is present in search.

AI visibility shows how often a brand appears in answers from AI systems, whether they cite your site, who they place next to you and in what tone they describe the company. It takes four base metrics: Mention Rate, Citation Rate, Share of Voice and Sentiment. They don’t replace SEO metrics, but add a layer that classic rankings, clicks and sessions don’t show.

Google states directly that AI Overviews and AI Mode can show supporting links, use query fan-out and are part of the wider Search ecosystem, while Microsoft already provides an AI Performance dashboard with citation and grounding query data.

01In brief

The 4 core AI visibility metrics

01 Mention Rate

Baseline presence

“Are we named at all?” — how often the brand is mentioned in answers.
Shows the brand’s baseline presence in AI answers.
02 Citation Rate

Source trust

“Does AI use us as a source?” — how often it links to your site.
Shows source trust and referral-traffic potential.
03 Share of Voice

Competitive position

“Who wins the category in AI?” — share of mentions vs competitors.
Gives a competitive read, not just absolute visibility.
04 Sentiment

Perception quality

“Are we recommended or criticised?” — the tone used to describe the brand.
Shows brand risk and perception quality.

Most important: don’t read these metrics in isolation. A high Mention Rate with negative Sentiment can be a problem. A high Citation Rate without a brand mention means your content is used but the brand isn’t remembered. A high Share of Voice without citations may mean the brand is “talked about” but isn’t the source of the answer.

02The difference

AI visibility vs SEO visibility: what’s different

In classic SEO a team usually looks at rankings, impressions, CTR, clicks, organic sessions, conversions and revenue.

In AI Search, part of the impact happens inside the answer. A user might ask: “Which CRMs suit a B2B SaaS team in Ukraine?”, get 5 options, see a short comparison and form a shortlist without ever visiting a site. If your brand didn’t appear in that answer, you may have lost the consideration stage before the click.

This doesn’t mean SEO is dead. Google states directly that SEO best practices remain relevant for AI features in Search, and that appearing in AI Overviews or AI Mode requires no special technical requirements beyond normal indexing, content accessibility, quality and snippet eligibility.

The right frame

SEO visibility answers the question: “Did the user find our page in search?”

AI visibility answers the question: “Did the AI system pick our brand or our content as part of the answer?”

03Mechanics

How AI Search may choose and cite sources

AI Search doesn’t work identically across products, but the general logic is this: the system takes a query, breaks it into intents or sub-queries, looks for relevant sources, synthesises an answer and sometimes adds links. For AI features Google describes techniques such as retrieval-augmented generation and query fan-out, where the model generates several related searches to better answer a complex question.

In Google AI Mode it’s explained even more simply: the system splits the query into subtopics and searches them in parallel to assemble a broader answer with links for further exploration.

  1. 1Query
  2. 2Fan-out: sub-queries
  3. 3Grounding: sources
  4. 4Synthesis: answer + citations
How one query is split into sub-queries, grounded against sources and synthesised into an answer with citations.

For brands this has three consequences:

  1. 1You compete not for one keyword but for a set of sub-questions.
  2. 2AI may use not your homepage but a specific snippet, table, FAQ, documentation, review or third-party mention.
  3. 3Visibility depends not only on your site, but also on how the brand is described in media, reviews, alternatives lists, forums and partner sources.

That’s exactly why AI visibility metrics should be computed not from “one nice prompt”, but from a systematic set of queries.

04Metric 1

Mention Rate: whether the brand appears in answers

Mention Rate is the share of relevant AI answers in which the brand is explicitly mentioned.

Formula
Mention Rate = Answers mentioning the brand / Valid answers × 100%
Example: 100 queries (40 informational · 30 comparison · 20 commercial · 10 reputation). The brand was mentioned in 28 → Mention Rate = 28 / 100 × 100% = 28%

This isn’t “good” or “bad” on its own. Context matters: which queries, which platform, which language, which competitors, where the brand appears in the list, whether the mention is positive and whether there’s a citation.

When Mention Rate is useful

It’s a baseline presence signal. It’s especially important for B2B/SaaS/tech, where a buyer often asks AI: “tools for…”, “best alternatives to…”, “which platforms suit…”, “compare X, Y and Z”, “what to pick for a 50-person team?”. If the brand isn’t mentioned in such answers, it doesn’t make the AI-generated shortlist.

A common mistake

It’s wrong to count only “was the brand mentioned in ChatGPT once”. AI answers depend on phrasing, language, country, time, model and context. 2025 research on AI Search shows that different services differ in domain diversity, freshness, cross-language stability and sensitivity to phrasing.

05Metric 2

Citation Rate: whether AI uses your site as a source

Citation Rate shows how often an AI answer includes a link to your site or to a source where your brand is mentioned. Here it’s important to distinguish two types.

Formula 1 · Source Citation Rate
Source Citation Rate = Answers citing your domain / All valid answers × 100%
Example: 100 prompts · 12 answers cite your site → 12%
Formula 2 · Mention-to-Citation Rate
Mention-to-Citation = Brand mentions with citation / All brand mentions × 100%
Example: the brand was mentioned 30 times · 9 mentions have a link → 9 / 30 × 100% = 30%

A mention means the brand is present in the narrative. A citation means AI uses a particular source as support for the answer. In its description of ChatGPT Search, OpenAI explicitly notes that answers can include links to sources and a sidebar with references. In its AI Performance dashboard Microsoft shows total citations, cited pages, grounding queries, page-level citation activity and visibility trends.

Owned vs third-party citations

Citation typeExampleWhat it means
OwnedAI links to your blog, docs, pricing pageYour site is the source
Third-partyAI links to a review, media outlet, directory, review siteThe brand is present via external authority
CompetitorAI mentions you but cites a competitor or an alternatives listYou’re in the narrative, but source authority is elsewhere
CategoryAI cites analytical content without mentioning the brandThere’s a chance to enter category content

What to do if the brand is mentioned without a link

This is a frequent scenario. Possible causes: the brand is known but the site has no easily citable pages; key facts are hidden in JS, PDF, gated content; AI relies on third-party sources; the product offer is described vaguely; there are no strong “who it’s for”, “comparison”, “alternatives”, “pricing”, “security”, “use cases” pages.

Practical action: create citation-friendly assets — pages with clear definitions, tables, comparisons, facts, update dates, authorship, sources and short self-contained blocks.

06Metric 3

Share of Voice: the share of AI answers the brand holds against competitors

AI Share of Voice shows what share of mentions or visibility a brand has compared with competitors across a given set of prompts.

Base formula
AI SoV = Your brand’s mentions / Mentions of all brands in the set × 100%
Category “CRM for B2B SaaS”, 100 answers: your brand — 25 · competitor A — 40 · B — 20 · C — 15. SoV = 25 / 100 × 100% = 25%

Why SoV is stronger than Mention Rate. Mention Rate answers: “Are we mentioned?”. Share of Voice answers: “Are we winning the category or losing to competitors?”. If your Mention Rate grew from 20% to 30%, but a competitor’s grew from 40% to 70%, your absolute presence improved while your competitive position got worse.

Weighted Share of Voice

In a real dashboard it’s worth counting not just the fact of a mention, but its weight.

SituationWeight
Brand mentioned first in a recommendation1.0
Brand mentioned in the top 30.8
Brand mentioned lower down0.5
There’s a citation to an owned domain+0.3
There’s positive sentiment+0.2
There’s negative sentiment−0.4

This isn’t a universal standard but an internal scoring model. The key is to fix the rules before you start measuring and not change them after the fact.

07Metric 4

Sentiment: how AI describes the brand

Sentiment shows the tone in which an AI system describes the brand: positive, neutral or negative. But for a CMO what matters more than just “positive / neutral / negative” is why exactly.

Levels of sentiment analysis

LevelWhat to assessExample
OverallThe overall tone of the mention“reliable tool”, “expensive option”
AttributeTone about a specific attributeprice, support, integrations, security, UX
ComparativeHow the brand is described against competitors“better for enterprise”, “worse for SMB”
RiskNegative associationscomplaints, lawsuits, data breach
AccuracyA factual errorold price, non-existent feature, wrong market

In 2026 BrightEdge specifically flagged to CMOs that AI engines can surface negative brand sentiment differently: Google AI Overviews more often picked up controversy-driven negativity, while ChatGPT more strongly surfaced product-evaluation negativity closer to the purchase stage. This is a secondary industry source, but it illustrates well why sentiment should be measured per platform, not as a single average number.

Sentiment score formula
Sentiment Score = (Positive − Negative) / All mentions × 100
Example: 50 mentions (20 positive · 25 neutral · 5 negative) → (20 − 5) / 50 × 100 = 30

But for practical work it’s better to look not only at the score, but at the themes of the negativity: expensive; complex onboarding; poor support; few integrations; not for small business; outdated information; confusion with another brand.

08System

How to build an AI visibility measurement system

1. Assemble a prompt universe

Don’t start with a single prompt like “recommend tools like X”. Build a set of queries by intent.

Prompt groupExamples
Category discovery“what tools are there for AI visibility tracking”
Problem-solution“how can a CMO measure brand visibility in ChatGPT”
Comparison“Brand A vs Brand B for B2B SaaS”
Alternatives“best alternatives to [competitor]”
Recommendation“recommend a platform for GEO monitoring”
Reputation“is [brand] reliable”
Pricing / purchase“how much does [brand] cost and is it worth buying”
Localized“AI Search visibility tools for the Ukrainian market”

For a starter audit 50–100 prompts is enough. For serious tracking — 200+ prompts, split by intent, funnel stage, platform, language and market.

2. Fix the platforms

A minimal set for B2B/tech: Google AI Overviews / AI Mode; ChatGPT Search; Microsoft Copilot / Bing; Perplexity. Google AI Overviews were expanded to over 200 countries and territories and over 40 languages in May 2025, and AI Mode supports Ukraine and the Ukrainian language among its supported languages.

3. Localise for Ukraine

For Ukraine it’s worth testing at least three language layers:

  • Ukrainian: “how to measure AI visibility”, “GEO metrics”, “brand visibility in ChatGPT”.
  • English: “AI visibility metrics”, “GEO metrics”, “ChatGPT visibility tracking”.
  • Mixed: “what is Mention Rate”, “Citation Rate formula”, “AI Share of Voice for SEO”.

Also add local modifiers: “in Ukraine”, “for Ukrainian SaaS”, “in Ukrainian”, “for a B2B CMO”, “for a tech company”.

4. Record the test conditions

For each measurement record: date; platform; model / mode if visible; language; country or local context; prompt; intent; whether there was a brand mention; which competitors were mentioned; citation URL; sentiment; errors or hallucinations; recommended action.

5. Repeat the measurements

AI visibility shouldn’t be assessed once. A practical cadence:

  • Weekly — for high-priority commercial prompts.
  • Monthly — for the full prompt set.
  • After major changes — product / pricing / positioning, PR campaigns, updates to search / AI platforms.
09Template

A dashboard template for AI visibility

FieldWhat to record
PromptThe exact query
IntentInformational, comparison, transactional, reputation
Funnel stageAwareness, consideration, purchase, retention
PlatformChatGPT, Perplexity, AI Overviews, AI Mode, Copilot
Language / marketuk-UA, en-UA, ru-UA, global
Brand mentioned?Yes / No
Mention position1st, top 3, list, paragraph, passing mention
CompetitorsList of competitors
Citation present?Yes / No
Citation URLURL or domain
Citation typeowned, third-party, competitor, media, forum
Sentimentpositive, neutral, negative, mixed
Sentiment reasonprice, support, features, trust, security
Accuracy issuewrong price, outdated info, confusion
Actionupdate page, create comparison, PR, reviews, technical fix
10Interpretation

How to interpret the results

SituationWhat it meansWhat to do
Low Mention + Low CitationAI barely sees the brandCategory content, entity clarity, PR, third-party visibility
High Mention + Low CitationThe brand is known, but the site isn’t a sourceCitation-friendly pages, docs, comparisons, FAQ, authorship
Low Mention + High CitationContent is used, the brand isn’t rememberedStrengthen brand/entity signals, authorship, About
High SoV + Negative SentimentVisibility exists, but it hurtsReputation, reviews, support, honest limitation pages
Low SoV + Positive SentimentThe brand is liked, but barely presentScale content, PR, comparison pages
High Citation + Low trafficAI uses the content but doesn’t send clicksAssisted discovery, branded search, demo conversions
11Myths

Common myths and mistakes

  • Myth 1. “AI visibility can be measured with one prompt”. No. One prompt is a screenshot, not a measurement. You need a set of queries, segmentation and repeatability.
  • Myth 2. “Citation = AI used the source correctly”. Not always. In 2025 Nature Communications published a study of LLM medical answers where a significant share wasn’t fully supported by the cited sources. Citations need to be verified.
  • Myth 3. “Just add llms.txt or special AI schema”. For Google Search that’s not the case. Google states directly that there’s no need to create special machine-readable AI files or schema.org markup to appear in generative AI features.
  • Myth 4. “Mention Rate matters more than anything”. Without sentiment and competitive context it can mislead. If a brand is often mentioned as “expensive” or “not for SMB”, high visibility doesn’t necessarily help sales.
  • Myth 5. “SEO is no longer needed”. It is. Google links generative AI features to its core Search ranking & quality systems and advises continuing to work on structure, content accessibility and unique value.
12Actions

What to do after an AI visibility audit

If you’re not mentioned

Create or update: category pages; use case pages; comparison pages; alternatives pages; “best for” pages; glossary / definitional content; customer proof pages; expert-authored guides.

If you’re mentioned but not cited

Improve citation-readiness: add short answer-first blocks; use tables; show the update date; give clear product facts; don’t hide important content behind JS or gated forms; make pages accessible to crawlers. OpenAI separates OAI-SearchBot for search from GPTBot for training; Perplexity describes PerplexityBot as a crawler for surfacing/linking in results, not for training.

If third-party sources cite you but not your site

This is a signal for digital PR and owned content: earn mentions in niche media; work with reviews; update directory profiles; produce research-led content; create pages that are easier to link to than someone else’s summary. In its “What is AI Reading?” study Muck Rack analysed over 1M links cited by AI tools and showed the weight of earned media in the citation ecosystem.

If sentiment is negative

Don’t start with “rewrite the SEO copy”. First find the source of the negativity: a real product issue; outdated reviews; negative reviews; confusion with a competitor; poorly explained pricing; an old incident; the lack of a page answering objections. Then act: update product messaging; create an honest comparison; add a “limitations / best fit” page; strengthen customer proof; work through reviews and PR; fix wrong facts on owned properties.

13FAQ

Frequently asked questions

It’s how often and in what context a brand, site or product appears in answers from AI Search systems: ChatGPT Search, Perplexity, Google AI Overviews / AI Mode, Microsoft Copilot and others.
Mention Rate shows how often the brand is mentioned in an answer. Citation Rate shows how often AI adds a link to your site or another brand-related source. A brand can be mentioned without a citation, and a site can be cited without strong brand recall.
As the share of your brand’s mentions among the mentions of all competitors in a given prompt set. For deeper analysis you can compute weighted SoV factoring in mention position, citation and sentiment.
Yes, at the start you can run a manual audit in a spreadsheet: 50–100 prompts, several platforms, logging mentions, citations, competitors and sentiment. But for regular tracking, change history and a large prompt set, a dedicated tool or dashboard is better.
Google notes that sites appearing in AI features are included in the overall Search Console Performance report under the “Web” search type; no universal “AI Overviews only” filter is stated in that documentation.
For important commercial prompts — weekly. For a full audit — monthly. It’s also worth re-measuring after product updates, pricing changes, PR campaigns and major changes at Google, OpenAI, Microsoft or Perplexity.
First find the source of the negativity: reviews, media, forums, old pages, wrong data or real product gaps. Then update owned content, close objections, fix facts, strengthen customer proof and work with PR/reputation.
For Google AI Overviews / AI Mode — not as a mandatory condition. Google states directly that there’s no need to create new machine-readable files or special AI markup. Other AI systems may have their own crawler recommendations, so robots.txt, server logs and bot access are still worth checking.
Ready to start?

See where your brand is already present in AI Search

We’ll show Mention Rate, Citation Rate, AI Share of Voice and Sentiment for your brand across ChatGPT, Google AI Overviews / AI Mode, Copilot and Perplexity — and which pages, sources and queries to optimise first.