How to Measure AI Visibility

AI visibility is becoming a new marketing metric.

However, treating AI visibility the same way as traditional search rankings can be risky.

In SEO, marketers have become used to relatively familiar metrics:

rankings, impressions, clicks, CTR, organic traffic and conversions.

AI search is different.

A brand could show up in ChatGPT for one prompt, vanish if the wording changes a bit, get cited by Perplexity without being named, or have very different visibility in another country or AI engine.

That makes measurement harder.

This often leads people to want to sum everything up with a single number:

AI Visibility Score.

I’ve noticed more people treating that score as the final answer.

I don’t think it is.

A visibility score can be helpful as a headline metric, but it’s only as good as the method used to create it.

In particular:

AI Visibility Score is only as good as the prompt set behind it.

If the prompts don’t reflect how real customers research your category, the score might look accurate but actually tell you very little.

What is AI visibility?

AI visibility describes how often and in what context a brand, company, product or source appears inside AI-generated answers.

That can include several different forms of visibility:

• your brand being mentioned
• your website being cited
• your product being recommended
• your company appearing in a comparison
• your content being used as a source
• your competitors appearing instead of you
• the way an AI system describes your brand

This is already more complicated than a traditional ranking.

A brand can be highly visible without its website being cited.

A website can be frequently used as a source without the brand itself being prominently mentioned.

These are different signals.

And I would measure them separately.

AI visibility is not one ranking

A common mistake is trying to apply the SEO ranking model directly to AI search.

Traditional SEO trained us to ask:

What position do we rank for this keyword?

With AI search, the question becomes:

How often do we appear across a meaningful group of prompts, engines, markets and contexts?

There is no single GEO engine.

ChatGPT, Google AI experiences, Gemini, Perplexity, Claude and other systems may return different brands and use different sources for similar questions.

That means visibility needs to be treated as a distribution rather than a single position.

I explored this difference in more detail in SEO vs GEO: What Actually Changes in the AI Search Era.

Start with the questions you want to measure

Before choosing a platform or looking at dashboards, I would start with something more basic:

What questions matter to the business?

Suppose you are measuring AI visibility for a CRM company.

A weak prompt set might contain:

• best CRM
• CRM software
• CRM for businesses

Those prompts are relevant, but they tell only part of the story.

Real customers may ask:

• What CRM is best for a 20-person B2B sales team?
• Which CRM is easier to implement than Salesforce?
• What CRM integrates well with HubSpot Marketing Hub?
• What is a good CRM for a startup with a limited budget?
• Which CRM has simple reporting for a non-technical sales manager?

These prompts include more context.

They represent different stages of the buying process.

And they create different competitive environments.

That is why prompt design should come before measurement.

From seed topics to a real prompt set

I would normally start with a group of seed topics.

For example:

CRM software

could become several intent groups:

Discovery

What CRM tools should a startup consider?

Comparison

HubSpot vs Salesforce for a growing B2B company.

Recommendation

What CRM would you recommend for a 30-person SaaS company?

Problem solving

Which CRM is easiest to implement without a dedicated operations team?

Feature-driven

What CRM has good pipeline reporting and HubSpot integration?

Alternative

What are the best alternatives to Salesforce for a small sales team?

Now the measurement set begins to reflect the market rather than simply a keyword list.

This is closely related to the shift I discussed in How to Optimize Content for AI Search:

Stop thinking only about keywords. Start thinking about questions, context, intent and decision criteria.

The prompt set is part of the measurement methodology

This point is worth emphasizing.

Two companies can measure the same brand and get completely different visibility scores simply because they use different prompts.

Imagine one platform tracks 20 broad informational prompts.

Another tracks 100 commercial prompts.

A third tracks 500 automatically generated questions.

These aren’t always measuring the same thing.

So whenever I look at an AI visibility report, one of my first questions is:

What prompts are behind this score?

Then:

Are they relevant?

Are they commercial or informational?

Do they reflect real customer questions?

Are they too broad?

Are important topics missing?

Are they representative of different stages of the buying journey?

Without that context, the percentage itself has limited meaning.

Engine, country and language matter

Prompt selection is only one dimension.

The same prompt can produce different results depending on:

• AI engine
• country
• language
• date
• model version
• access to web search
• personalization or context

This matters particularly for international companies.

A company may have strong visibility in the US but limited visibility in Germany.

Or it may appear frequently when prompts are written in English but rarely for equivalent questions in another language.

Semrush, for example, has expanded its AI visibility data across dozens of country markets, illustrating why regional segmentation is becoming an important part of measurement.

So instead of asking:

What is our AI visibility?

I prefer:

What is our visibility for this topic, in this market, across these engines?

AI responses are not completely stable

Another difference from traditional SEO is variability.

Ask the same AI system the same question multiple times and you may not always receive exactly the same answer.

Brands can change.

Sources can change.

The wording can change.

A 2026 OtterlyAI study tracked 520 prompts across seven AI answer engines and found that brand mentions were generally more stable than citations, but both could vary over time.

That has an important implication:

One query is not a measurement system.

You need repeated observations.

The trend matters more than one screenshot.

A practical AI visibility measurement framework

The basic process I would use is:

Seed Topics → Prompt Set → Engine + Country → Repeated Queries → Mentions & Citations → Competitor Comparison → Trend Over Time

Every stage is important.

1. Seed Topics

Start with the business areas where visibility actually matters.

Not every topic on the website needs measuring.

Focus on topics connected to:

• the product
• customer problems
• purchase decisions
• differentiation
• commercial intent

2. Prompt Set

Turn those topics into realistic questions.

Include a mix of:

• informational
• comparative
• commercial
• recommendation
• alternative
• problem-solving prompts

3. Engine and Country

Define where the prompts should run.

At minimum, separate the major engines rather than blending everything immediately into one score.

4. Repeated Queries

Run the prompts consistently over time.

Avoid making strategic conclusions from a single response.

5. Mentions and Citations

Track both.

They are not the same thing.

6. Competitor Comparison

A visibility number becomes more meaningful when compared with relevant competitors.

7. Trend Over Time

The important question is not simply where you are today.

It is:

Are we becoming more visible for the topics that matter?

What should you actually measure?

I suggest not relying on just one metric.

Instead, I would monitor a small group of complementary metrics.

Brand Visibility

What percentage of relevant responses mention your brand?

Share of Voice

How much of the brand presence in the tracked prompt set belongs to you compared with competitors?

Brand Mentions

How often is the company explicitly named?

Citations

How often is your website or content referenced as a source?

Source Visibility

Which domains and URLs are being used to support answers?

Prompt Coverage

For what percentage of the prompts do you appear at all?

Competitor Presence

Which competitors appear where you do not?

Engine Visibility

Are you stronger in ChatGPT, Gemini, Perplexity or another engine?

Topic Visibility

Which topics produce the strongest or weakest brand presence?

Trend

Is visibility improving or declining over time?

Some platforms also measure sentiment or the relative position of brands inside generated answers. Peec AI, for example, separates visibility, position, sentiment and share of voice, and also provides competitor and source analysis.

Mentions and citations are not the same thing

This distinction is especially important.

Imagine an AI response says:

For project management, companies commonly consider Asana, Monday.com and ClickUp.

Your brand may be mentioned even if your own website is not cited.

Now imagine the system uses research published on your domain as a source but never mentions your company by name.

Your content has source visibility.

Your brand does not necessarily have the same brand visibility.

Peec explicitly distinguishes between brand visibility and source visibility for this reason.

I would keep these metrics separate.

They diagnose different problems.

If the brand is frequently mentioned but the site is rarely cited, the market may know the brand while the site’s content is not becoming a strong reference source.

If the site is frequently cited but the brand is rarely mentioned, the content may have authority without that authority fully transferring to brand recognition.

Tools for measuring AI visibility

There are now several platforms designed specifically for this type of measurement.

I use tools such as Semrush and Profound, but I would not choose a platform based only on its headline visibility score.

The key questions are:

What prompts can I track?

Which engines are covered?

Can I segment by market?

Can I see individual responses?

Can I separate citations from mentions?

Can I compare competitors?

And can I understand the sources influencing those answers?

Semrush

Semrush is especially interesting for teams already using the platform for SEO.

Its AI Visibility tooling can monitor brand presence across AI platforms, regions and topics, benchmark competitors and track share of voice.

The advantage is the broader context.

AI visibility does not live in isolation from:

• SEO
• rankings
• keyword research
• content
• competitors
• backlinks

For teams already using Semrush, combining those data sets can be useful.

Profound

Profound is more purpose-built around answer-engine visibility.

Its platform runs structured prompts across AI systems and analyzes brand appearances, citations, sentiment, rankings and competitive presence. It also allows customers to customize their tracked prompts rather than relying only on a predefined list.

That prompt flexibility matters to me.

A generic prompt database may help with discovery, but the questions most relevant to a specific company often need to be defined manually.

Peec AI

Peec AI focuses strongly on prompt tracking, competitive benchmarking and AI share of voice.

It tracks visibility, position, sentiment and share of voice, and allows users to analyze the sources and citations associated with prompts.

One useful distinction in Peec is between the sources an AI system accesses and the citations that actually appear in the visible answer.

That can help identify where competitors are gaining authority outside their own websites.

OtterlyAI

OtterlyAI also focuses on AI search monitoring, particularly brand mentions, citations and source analysis.

Its citation reports can be filtered by dimensions such as AI engine and country and can help identify where competitors are being cited while your brand is absent.

It is also worth paying attention to its research into answer stability, because it illustrates why repeated measurement is important rather than relying on individual responses.

Don’t outsource the methodology to the tool

This is probably my top recommendation.

A platform can automate the measurement.

It should not define the strategy for you.

I would not automatically accept every suggested prompt.

Review them.

Remove irrelevant ones.

Add questions that come from:

• sales calls
• customer interviews
• Search Console
• product research
• competitor research
• support tickets
• sales objections
• category knowledge

The best prompt sets usually mix platform-generated discovery with real business knowledge.

Real prompts vs synthetic prompts

Another question worth asking is where the prompts come from.

Some may come from real search or query data.

Others may be generated synthetically based on category terms.

Both can be useful.

But they should not necessarily be treated as equivalent.

Synthetic prompts are excellent for:

• building coverage
• discovering scenarios
• identifying content gaps

Real customer language is more useful for understanding:

• actual intent
• terminology
• objections
• buying criteria
• context

I would use both, but label them clearly.

Share of Voice needs context

Share of Voice is likely to become one of the most popular GEO metrics.

It is intuitive.

If you appear in 25% of relevant brand mentions and your competitor appears in 40%, the comparison is easy to understand.

But again, the prompt set determines the result.

If you add 50 prompts highly relevant to your competitor, their Share of Voice may increase.

If you add 50 prompts where your product is stronger, yours may increase.

Neither change necessarily means AI systems changed their perception of the market.

The measurement set changed.

That’s why managing your prompts becomes part of analytics.

When you change the prompt universe significantly, document it.

Otherwise, trend lines can become misleading.

Look at individual prompts, not just dashboards

Aggregate metrics help executives understand direction.

But some of the most useful insights are usually found at prompt level.

Ask:

Why are we absent here?

Which competitor appears?

Which source is cited?

Does our website answer this question?

Is the AI using a comparison website?

A Reddit discussion?

A competitor’s documentation?

A publication?

An industry review?

This turns measurement into action.

Instead of:

Our AI visibility is 18%.

you can say:

We are consistently absent from prompts about implementation cost because third-party comparison sites dominate the sources used for those answers.

That is much more useful.

Connect AI visibility to business outcomes

Eventually, GEO measurement needs to go beyond mentions.

Visibility is an intermediate metric.

The business still cares about:

• awareness
• qualified traffic
• leads
• signups
• pipeline
• revenue

AI-driven journeys make attribution more difficult because a user may see the brand in an AI answer and visit the website later through direct or branded search.

So I would look for supporting signals:

AI referrals.

Branded search growth.

Direct traffic.

Conversion from AI-referred visitors.

Brand lift.

Sales feedback.

Customer survey responses such as:

How did you first hear about us?

AI visibility should eventually connect to commercial impact, even if attribution remains imperfect.

What should an AI visibility dashboard contain?

I would keep the executive dashboard relatively simple.

At the top:

AI Visibility / Share of Voice

Then underneath:

• visibility trend
• competitor comparison
• visibility by engine
• visibility by topic
• mentions
• citations
• prompt coverage
• top gains
• biggest gaps

Then add the ability to drill down into individual prompts and sources.

Executives need the trend.

Practitioners need the details and explanations.

The goal is not the highest possible visibility score

You can always build a prompt set that makes your numbers look better.

But that’s not the goal.

The goal is to measure whether the company is becoming more visible for the questions that matter to potential customers.

A smaller set of highly relevant commercial prompts may be more valuable than thousands of broad informational prompts.

Your measurement system should match your business strategy.

Not the other way around.

Frequently Asked Questions

What is AI visibility?

AI visibility measures how frequently and in what context a brand, product or source appears in answers generated by AI search systems.

How do you measure AI visibility?

Start by defining relevant topics and prompts, then track brand mentions, citations, competitor presence, share of voice and visibility across different AI engines over time.

What is AI Share of Voice?

AI Share of Voice measures your brand’s presence in AI-generated responses relative to competing brands within a defined prompt set.

What is the difference between an AI mention and a citation?

A mention means the AI response explicitly names the brand. A citation means the system references a particular source or page. A brand can be mentioned without its website being cited, and a website can be cited without the brand being prominently mentioned.

Which tools can measure AI visibility?

Platforms such as Semrush, Profound, Peec AI and OtterlyAI can monitor different combinations of prompts, brand mentions, citations, competitors, engines and geographic markets. Their capabilities differ, so the right choice depends on the measurement framework rather than the visibility score alone.

How many prompts should I track?

There is no universal number. The prompt set should be large enough to represent the important topics, intents and buying scenarios in the category. Relevance and coverage matter more than simply maximizing prompt volume.

Should AI visibility be measured separately for each engine?

Yes. Different engines can return different brands, sources and recommendations. An aggregate score can be useful, but it should be possible to break the result down by engine.

Can AI visibility be connected to revenue?

Partially. Direct AI referral traffic can be measured, but many AI-influenced journeys may later appear as branded search, direct traffic or another channel. AI visibility should therefore be analyzed alongside broader acquisition and brand metrics.