How AI Visibility Tools Really Measure Prompts: Semrush vs. Ahrefs vs. Google (2026)

TL;DR

• No tool measures real AI usage. Prompt data from ChatGPT, Gemini and other assistants never leaves the AI companies’ servers, so every “prompt volume” is an estimate.
• Semrush estimates demand: 317M+ prompts from clickstream data, grouped into topics, with volume estimated by machine learning.
• Ahrefs audits answers: it sends its own synthetic prompts, built from keyword and People Also Ask data, and records what AI platforms reply.
• Google gives two free reports: generative AI impressions in Search Console and an AI Assistant channel in GA4. Neither shows prompts.
• AI answers are volatile: around 70% of AI Overview content changes between two captures, according to an Ahrefs study.

Contents

1. Measured vs. estimated
2. How Semrush estimates prompt demand
3. How Ahrefs audits AI answers
4. Semrush vs. Ahrefs at a glance
5. What Google measures for free
6. Why AI visibility data is volatile
7. Limits in smaller-language markets
8. How to test AI visibility yourself
9. Vyara’s take

AI visibility tools promise to show what people ask ChatGPT and whether your brand appears in the answer. The numbers look precise. But the first question to ask about any of them is simple: was it measured, or estimated?

Measured vs. estimated

In classic SEO, keyword volumes are also estimates, but they are anchored in Google’s own data. In AI search, there is no equivalent. OpenAI, Google and other AI companies don’t share what users type into their assistants. As Ahrefs puts it, that data never leaves their servers.

So every AI visibility tool relies on a proxy. The two biggest SEO platforms use very different ones.

How Semrush estimates prompt demand

According to Semrush’s documentation, its AI Visibility Toolkit is built on:

• more than 317 million prompts and responses from ChatGPT, Gemini, Google AI Overviews and AI Mode;
• 117 regional markets, with daily rolling updates;
• prompts sourced from AI search clickstream data and Google’s keyword set for AI Overviews.

Clickstream data comes from browser extensions and consumer panels that collect browsing behavior with users’ consent. It captures real prompts from real people, but only from the people in the sample. Coverage depends on who installs these extensions, in which country and in which language.

Semrush does not report volume for individual prompts. It clusters similar prompts into topics and estimates topic volume with machine learning. The company states that “no platform can provide exact numbers” for AI visibility and describes its metrics as “reliable directional signals”.

How Ahrefs audits AI answers

Ahrefs Brand Radar takes a different approach. Instead of estimating what people ask, it asks the questions itself.

According to Ahrefs, prompts are seeded from its keyword database, which Ahrefs says contains 110 billion keywords, and from Google’s People Also Ask questions, paired with search volume. Ahrefs describes these prompts as synthetic. The answers are stored, and users search that archive for brand mentions and cited pages.

Brand Radar’s impressions metric is also modeled. According to Ahrefs’ help center, it is calculated by summing the Google search volumes of the prompts in which a brand appears. It assumes that Google search demand reflects AI prompt demand.

Ahrefs also offers free, limited tools, including an AI Overviews tracker, an AI Mode tracker and an AI visibility checker.

Semrush vs. Ahrefs at a glance

Neither tool measures audience. Both are visibility audits: they show what appears in AI answers, not who saw it.

What Google measures for free

Search Console. In June 2026, Google introduced generative AI performance reports, rolled out to all sites worldwide by August 31, 2026. The Search report shows impressions in AI Overviews and AI Mode by page, country, device and date. Discover has a separate generative AI report.

Limitations:

• impressions only, with no clicks and no positions;
• no queries, so you don’t see what people asked;
• AI Overviews and AI Mode are combined in one number.

Google Analytics 4. In May 2026, Google added an AI Assistant channel to GA4’s default channel group. Sessions referred by recognized assistants, such as ChatGPT, Gemini and Claude, are labeled automatically.

Limitations:

• it depends on the referrer, so visits from apps or copied links that pass no referrer land in Direct;
• traffic from AI Overviews and AI Mode stays in Organic Search, because it comes from Google.

The AI Assistant number is a minimum, not the full picture.

Why AI visibility data is volatile

An Ahrefs study of more than 43,000 keywords, reported by Stan Ventures, found that:

• around 70% of AI Overview content changed between consecutive captures;
• content changed roughly every 2.15 days;
• 45.5% of cited sources were replaced at each refresh.

A single check is therefore a snapshot, not a measurement.

Limits in smaller-language markets

Both approaches depend on samples that are strongest in English and in large markets. For smaller languages and local niches, clickstream samples are thin, and synthetic prompts built from keyword data may miss how local buyers actually phrase questions.

At the same time, competition in small niches is lower. A brand in a market with few players and few GEO-optimized sites may stay an AI recommendation far longer than the global volatility figures suggest.

How to test AI visibility yourself

• Start from the buyer, not the tool. Map the stages of the purchase decision and write the questions a buyer asks at each stage.
• Test in a clean environment. Use incognito sessions or accounts with no history or memory. A personalized account overstates your visibility.
• Repeat the test. Ask the same prompt on different days and record how often you appear, as a percentage.
• Check each surface separately. AI Overviews, AI Mode and chat assistants cite different sources.
• Validate tools manually. Before paying for a tool, check a sample of prompts by hand and compare the results.

Vyara’s take

“Every AI visibility number is an estimate, and the tools themselves say so. For smaller markets, I don’t start with a tool. I start with the business and the buyer: who buys, what worries them, what they would ask ChatGPT at each stage. If you have the map of the decision, you have the map of the prompts. My order is always the business, then the buyer, then the prompts, and only at the end, the website.”

Further reading: Why AI visibility tools can’t tell you what people ask ChatGPT, and how to research prompts instead – Vyara Stefcheva’s full method on stefcheva.com.

Sources: Semrush – AI Visibility Toolkit data · Ahrefs – You can’t track AI like traditional search · Ahrefs – Brand Radar · Ahrefs – AI visibility metrics · Google Search Central – Generative AI performance reports · Search Console Help – Generative AI performance report · Search Engine Journal – GA4 AI Assistant channel · Stan Ventures – AI Overviews volatility

About the Author

Vyara Stefcheva

Vyara Stefcheva is the founder and editor of Egorithms and an independent marketing expert. She follows how technology, AI and algorithms change the way people find, choose and buy, and writes about what that means for businesses and for the people who use them.

She has spent her career building marketing and business strategies for more than 300 brands across three continents. She built and exited a marketing agency and now works independently. She has taught more than 20,000 people through lectures, courses and corporate training, and speaks at conferences about AI and the future of the internet.

On her YouTube channel she tests new AI and marketing tools as they launch and gives each an independent verdict: adopt, watch or ignore.

See author's posts