How to track AI visibility: a zero BS guide for beginners

Jump to section:

• Can you accurately track AI visibility?
• The challenges of tracking AI visibility: a closer look
• The most popular AI visibility (GEO/AEO) metrics
• Other AI search metrics: tracking commercial impact
• How to build your list of prompts to track in AI search
• Some tips for generating AI search prompts
• What are AI visibility monitoring tools?
• How do AI visibility monitoring tools work?
• Getting started with AI visibility tracking: your 3-step action plan

Can you accurately track AI visibility?

Despite what some SEO bros might say on LinkedIn, there is no definitive or simple way to track AI visibility.

But that doesn’t mean you shouldn’t try. You just need the right information and the right set of expectations. This guide explains what you can actually track, how to do it, and (just as importantly) what the limitations are.”

Firstly, what does AI visibility mean? 

It’s the extent to which your brand is discoverable, recognisable, and accurately represented within AI-powered search tools and large language models (LLMs) like ChatGPT, Perplexity, or Google’s Gemini (AI assistant, AI Overviews and AI Mode). You may also see it referred to as Answer Engine Optimisation (AEO), Generative Engine Optimisation (GEO), or something else entirely.

The challenge is that unlike traditional SEO, where you can track fixed rankings on stable search results pages, AI visibility is probabilistic and context-dependent  — meaning you get different answers to the same question every time. Probabilistic simply means the AI generates responses based on likelihood rather than certainty, so results vary even with identical queries.

Think of it this way: you’re not measuring a position on a nice, sturdy ladder – you’re estimating how often you appear in a hazy cloud of possible answers.

LLMs generate different responses to the same question over time and across users, so there’s no stable ranking or attribution layer to measure. Sure, you can estimate patterns and track directional trends, but you can’t measure AI visibility with the precision you’re used to in SEO.

This blog explains what you can track, how to do it, and – just as importantly – what the limitations are. Zero BS.

The challenges of tracking AI visibility: a closer look

Understanding why AI visibility is hard to measure helps you set realistic expectations and choose the right approach. These aren’t just technical quirks – they’re fundamental limitations that affect what you can track, how accurately, and what it tells you about your brand’s visibility.

These core challenges are:

1. AI systems weren’t designed to be measured
2. The same query can produce different answers
3. You can’t inspect what’s “in” the LLM
4. Information retrieval doesn’t guarantee citation
5. People search in endless different ways
6. The more people optimise, the weaker the signal becomes
7. It’s tricky to isolate what’s actually working

Let’s take a closer look at each.

1. AI systems weren’t designed to be measured

Traditional search engines have indexes, rankings, and logs. They give you clear URLs you can track and measure.

LLMs work differently. They generate language (i.e., answers) – not rankings or fixed “results”. There’s no built-in way to know “you ranked #3” or “this answer used 62% of your content.”

So, from the start, AI visibility lacks a proper measurement layer. You’re reverse-engineering what you see in outputs, not reading data from the system itself.

2. The same query can produce different answers

LLMs don’t retrieve fixed answers. They generate new responses each time based on how you phrase your question, your search history (if using a logged-in account), your location, and other contextual factors. Your brand might appear in one response and disappear in the next.

This is what’s meant by “probabilistic” – the model calculates the likelihood of different words and phrases appearing next, rather than pulling from a stored answer. It’s more like rolling dice than looking something up in a filing cabinet.

You’re not asking, “Did I rank?” – you’re asking, “How often do I appear across possible answers?”

And that probability is inherently unpredictable.

3. You can’t inspect what’s “in” the LLM

LLMs don’t store knowledge as documents or web pages the way traditional search engines do. They store compressed patterns in their neural networks, not searchable indexes you can audit.

This means you can’t inspect whether your content is “in the model,” trace exactly which source influenced which answer, or get the model to reliably explain where information came from.

Visibility becomes about what you observe in outputs, not how it happened behind the scenes. This makes perfect tracking impossible in principle, not just in practice.

4. Being retrieved doesn’t mean being cited

Even when AI systems retrieve your pages or documents from the web, clean measurement breaks down.

Having your information retrieved doesn’t guarantee being cited. Multiple sources may be blended together. Citations may be partial, rotated between responses, or left out entirely.

This is one reason why Google and other AI providers are facing backlash from publishers and content-heavy websites. Their concern is that AI systems are scraping and using their content to generate answers – often reducing traffic to their sites – without properly attributing or compensating them for that use.

5. People search in endless different ways

In SEO, you track known keywords and known search results pages. In AI, users ask infinite variations in conversational language that can shift intent mid-conversation.

You’re sampling from an unbounded space of possible questions. Any tracking system is necessarily incomplete and can never cover the full surface.

This is a mathematical limitation, not a software limitation.

6. The more people optimise, the weaker the signal becomes

The moment people start optimising for “AI visibility,” content becomes more similar, models adapt, and the signal decays.

This happens in SEO too, but AI absorbs optimisation faster and reflects it back into outputs more quickly. Measurement changes the thing being measured.

7. It’s tricky to isolate what’s working

AI visibility doesn’t exist in a vacuum. It overlaps heavily with traditional SEO, digital PR, and broader brand reputation signals.

This means you can’t really pinpoint one factor that’s driving AI visibility. It’s almost always a combination of content quality, digital PR, website authority, and how the AI system weighs and retrieves information.

This means you might be “visible” in AI results without deliberately optimising for them, simply because your wider digital presence is strong.

These challenges don’t mean you shouldn’t track AI visibility, but they do mean you need to approach measurement differently than you would with traditional SEO. The key is knowing what you can track reliably, and what those metrics actually tell you.

The most popular AI visibility (GEO/AEO) metrics

As you can imagine, there’s no industry-standard set of AI visibility metrics (not yet at least). Different tools use different terminology, and what one platform calls “visibility score” might be known by a different name elsewhere.

But a few core metrics have emerged as the most common and useful – and most AI visibility monitoring tools offer some version of these.

You can look at these metrics overall across all your tracked queries, or drill down into specific topics or subsets of prompts that matter most to your business.

Here are some of the metrics you’ll find in AI search tracking tools:

Other AI search metrics: tracking commercial impact

AI visibility monitoring tools give you a picture of how often you appear in answers, but there are a couple of other AI search metrics you can track to understand the commercial impact:

Referral traffic from AI platforms

You can track referral traffic from sites like ChatGPT and Perplexity in Google Analytics (GA). Look for traffic sources labelled “chat.openai.com” or “perplexity.ai” in your acquisition reports.

The reality check: for most businesses, this referral traffic still makes up a tiny fraction of overall web traffic – typically around 1% or less. But it’s worth monitoring as AI adoption grows.

Leads generated from AI traffic

If you want to understand whether AI visibility is driving business outcomes, track leads that come from AI sources. You can do this by:

• Using custom UTM parameters in any links you control that might be cited by AI systems
• Adding “Through AI search” or similar as an option in the “How did you hear about us?” section on your contact forms

This gives you a clearer picture of whether AI visibility is translating into tangible commercial value.

A note on Google metrics for AI visibility

It’s also important to mention that Google doesn’t currently offer dedicated metrics for AI search in the same way it does for traditional search (organic clicks, impressions, sessions, users).

The current state of play is that any clicks, impressions, or sessions generated via Google’s AI Overviews or AI Mode are blended into your overall figures in GA and Google Search Console, with no easy way of determining how much was contributed by AI visibility specifically.

This makes it harder to isolate the impact of Google’s AI features on your traffic and performance – another reason why tracking AI visibility requires a different, more directional approach than traditional SEO measurement.

How to build your list of prompts to track in AI search

Deciding which prompts to track is one of the hardest parts of AI visibility measurement.

The reality is that people can ask AI systems questions in endless different ways, and you’ll never be able to track them all. What you’re tracking will only ever be a snapshot of the possible.

This means you need to settle on a realistic, relevant set of prompts that reflect what your target customers or personas might be searching for.

This isn’t an exact science, but it doesn’t have to be. Start by looking at these three key areas, using a fictional project management software as an example.

1. Recommendations

What do you want your brand to be recommended for? Which middle- or bottom-of-funnel queries should surface your products or services when someone’s actively looking for solutions?

An example prompt to track might be: “What’s the best project management software for remote teams?”

2. Citations

Which topics do you want to influence or be quoted as an authority on? This is more about thought leadership and top-of-funnel visibility – establishing your brand as a credible source.

An example prompt to track might be: “How do agile project teams track sprint progress?”

3. Reputation

Which branded queries are you interested in tracking to see how AI represents your brand? What sentiment is being expressed when people ask about you directly?

An example prompt might be: “Is [insert fictional software company name] good for small businesses?”

Some tips for generating AI search prompts

Once you’ve answered those questions, you need to build a matrix of specific prompts to track.

Here are some practical ways to determine which prompts to include:

• Look at your existing SEO keywords: There’s natural overlap between what people search for in Google and what they ask AI systems. Start there.
• Organise prompts by product or service: Group queries around specific offerings so you can measure visibility where it matters most commercially.
• Ask your sales, business development and product teams: They hear the same questions repeatedly. Those real-world queries are gold for AI tracking.
• Use market and customer data: Surveys, interviews, call logs, and support tickets can reveal the language your audience actually uses.
• Prioritise commercial impact: Focus on areas where improving AI visibility is most likely to lead to tangible business results – qualified traffic, leads, and sales.
• Think conversationally: AI search is more natural and question-based than traditional keyword search. Consider full questions people might ask, using human language, not just keyword phrases.

Start small and expand over time. Begin with 20–30 core prompts that cover your priority areas, then add more as you learn what’s working and where the gaps are.

Going in with hundreds of prompts immediately will be overwhelming and make it harder to spot meaningful patterns or build a coherent strategy.

What are AI visibility monitoring tools?

Over the past year or so, we’ve seen an explosion in AI visibility tracking tools. Lots of new options have come to market – including Profound, Peec, Semrush (an extension of their existing SEO software), Otterly, and Writesonic.

But they all broadly try to do the same thing: automate the process of testing prompts across multiple AI systems and tracking which brands appear in the answers.

Before we get into how these tools work, a few important caveats:

• They can get expensive. Especially if you choose to track hundreds of prompts, which is another reason to stick with a smaller subset of priority queries.
• Features vary widely. Some offer pure visibility tracking, while others include built-in capabilities like content recommendations, AI-powered content generation, off-site PR suggestions, and various analytics dashboards.
• They’re not a substitute for strategy. You still need a proper SEO or AI visibility strategy, and ideally a dedicated team or agency to interpret the data and act on it.
• They’re data aggregation platforms. Ultimately, these tools solve the problem of collecting data from multiple prompt executions across multiple LLMs on a regular basis without doing it manually. You could technically do it manually if you wanted to – but it would be incredibly time-consuming.

How AI visibility monitoring tools work

Here’s a simplified, five-step summary of how AI visibility monitoring tools work in practice.

1. Create your list of prompts

Think of these as “test” searches. There are no right or wrong prompts – they should reflect the questions your target audience might ask.

2. The tool executes the prompts

The tool runs your prompts in real-time across different AI models (ChatGPT, Perplexity, Google AI Mode, etc.). It captures data on which brands, products, or services appear in the answers.

It also shows you which online sources are most influencing the answers – whether those are corporate websites, editorial sites, social media platforms (like LinkedIn), user-generated content sites (YouTube, Reddit), or institutional sites.

3. Track for mentions and citations

The tool works out how often your brand is mentioned across all the prompts you’re tracking. This becomes your “visibility score” – usually measured out of 100%.

You can also see how frequently your brand’s content is being used as a cited source.

4. Repeat over time

Prompts are executed and new data is collected every day (or whatever frequency comes with your package).

This helps you see clearer patterns and trends over time, despite the imperfections in the data.

5. Track against competitors

In most tools, you can compare your visibility scores against a defined list of competitors. This gives you a sense of your share of voice and where you’re leading or falling behind.

Getting started with AI visibility tracking: your three-step action plan

Tracking AI visibility isn’t straightforward, but there’s increasing pressure on brands to do it. Here are three key steps to get started.

1. Build a starter list of 20-30 priority prompts.
2. Choose a tracking tool and establish your baseline.
3. Review monthly and adjust your content/PR strategy accordingly.

More broadly, our advice for brand and marketing leaders would be to:

• Accept that measurement will be imperfect and directional rather than precise.
• Focus on a realistic set of prompts that reflect real customer questions.
• Track the metrics that matter most to your business goals.
• Monitor trends over time, rather than obsessing over individual data points.
• Combine AI visibility with your broader SEO, PR, and brand strategy – because they all influence each other.

Want to improve your brand’s visibility in AI search?

Our team helps brands increase their AI visibility scores. Speak to our B2B SEO, PR and social media experts about building a no-BS strategy for AI search.

Written by Matthew Robinson, Senior PR and SEO Strategist at Definition on 26/01/2026.

Additional FAQs about tracking AI visibility

1. How often should I check my AI visibility metrics?

There’s no need to check daily. AI visibility changes more slowly than traditional search rankings, and daily fluctuations are mostly noise. Weekly or fortnightly reviews are usually sufficient to spot meaningful trends.

Monthly reporting works well for most businesses, especially if you’re just starting out. The key is consistency: track the same prompts over time so you can identify genuine patterns rather than random variations.

2. Can I track AI visibility without paying for specialist tools?

Yes, but it’s labour-intensive. You can manually test prompts across different AI platforms (ChatGPT, Perplexity, Google AI Mode) and log which brands appear, where they’re positioned, and whether they’re cited.

The downside is that manual tracking doesn’t scale well, makes it harder to spot trends, and won’t give you competitive benchmarking. Paid tools automate the heavy lifting and provide more comprehensive data.

3. How long does it take to see improvements in AI visibility?

It varies, but expect months rather than weeks. AI visibility is influenced by many of the same factors as traditional SEO – content quality, domain authority, backlinks, brand mentions – and those take time to build.

If you’re creating new content or running digital PR campaigns, it can take 3–6 months before you see meaningful changes in AI visibility metrics.

4. Should I optimise differently for different AI platforms?

Not radically differently. The fundamentals are the same: high-quality, authoritative content with clear structure, credible backlinks, and strong brand signals.

However, there are nuances. Google’s AI features draw heavily from its existing search index, so traditional SEO still matters enormously. ChatGPT and Perplexity rely more on real-time web retrieval and citation-worthy sources.

5. What’s a realistic AI visibility score to aim for?

There’s no universal benchmark. It depends entirely on your industry, competitors, and the prompts you’re tracking.

For example, a score of 30–40% might be strong in a highly competitive sector, while 60–70% could be achievable in a niche market. Don’t fixate on hitting a specific number. Instead, focus on improving your score relative to competitors and tracking upward trends over time. The goal isn’t perfection – it’s consistent, measurable progress.