For years, SEO professionals have relied on well-established data points to measure organic performance. These principally included rankings, organic traffic, conversions and revenue. However, with the advent of AI search, there is now a completely new ecosystem of data points to measure. These include mentions, citations and cited pages, as well as share of voice, prompt visibility and sentiment.

As you may already be able to tell, measuring AI search is much more complicated than measuring traditional search. In addition, it’s also much less accurate. For example, whilst you can measure conversions from AI search referrals, that’s not really a reliable, holistic barometer of how well you are faring in AI search because it only measures users who clicked through to your site. It does not include those who saw your brand in an AI-generated answer and later visited and converted via another channel. Thus, there is a dual challenge of multiple metrics alongside an apparent lack of attribution.

Which AI Search Metrics Should I Measure?

There are dozens of different metrics you can measure, but at a minimum you should track:

  • Mentions – Instances where a brand is named in an AI-generated response, with or without a link.
  • Citations – Instances where an AI response includes a source link to the brand’s domain.
  • Prompt Visibility – The percentage of tracked prompts in which a brand appears.
  • AI Overviews – Google’s AI-generated summary block.
  • Referrals from AI Search – Website sessions originating from AI assistants or measurable AI-generated search results.
  • Sentiment – Whether a brand mention is framed positively, negatively or neutrally.

Some of this data can be obtained from GA4 and Google Search Console, whilst other metrics require specialist software.

Which Software Can I Use to Measure AI Search Performance?

  • Semrush

Semrush’s AI Visibility reports provide benchmarking and historical trends for AI mentions and citations. Prompt Tracking allows you to monitor a defined set of commercially important prompts over time. The standalone AI Visibility Toolkit starts from $99 per month, while Semrush One combines the AI and traditional SEO toolkits from $199 per month.

Metrics available across Semrush’s different AI Visibility reports include:

  • Share of Voice
  • AI Visibility Score
  • Mentions
  • Average Position
  • AI Overview Citations
  • Overall Sentiment

Above: The AI-search data available in dashboard view on SEMrush

It is worth noting that these metrics are spread across different reports. Prompt Tracking itself includes AI Visibility, Mentions, Owned Sources, Topic Volume and Average Position. Share of Voice is available within Semrush’s wider Brand Performance reports rather than Prompt Tracking.

  • DataForSEO

DataForSEO provides a flexible API for retrieving AI mentions, citations, historical trends and source data. It is not a direct replacement for a dedicated prompt-tracking platform but is an excellent option for building your own reporting.

At the time of writing, its database-backed LLM Mentions data covers ChatGPT and Google AI Overviews, so its platform coverage is narrower than some dedicated prompt trackers.

Pricing is pay-as-you-go following a minimum account top-up. Current pricing is approximately:

  • $0.10 per request
  • $0.001 per row returned

Above: DataForSEO dashboard

  • GA4

GA4 introduced an AI Assistant channel in May 2026 that categorises traffic from recognised AI-assistant referrers. AI Overviews continue to be reported under Organic Search. Some AI traffic arrives without a referrer and therefore appears as Direct traffic. Custom channel groups can recategorise identifiable AI referrals but cannot recover genuinely unattributed Direct visits.

Above: AI assistant data in GA4

  • Google Search Console

Search Console’s generative AI reporting currently provides impression data for AI Overviews and AI Mode within Search, while Discover’s generative AI reporting is shown separately. The Search report provides impressions by page, country, date and device, but clicks, CTR and query data are not yet available.

The reports are currently being tested with a subset of UK website owners ahead of a broader rollout.

Above: Google Search Console’s Generative AI features report shows impressions generated through AI Overviews and AI Mode over time.

  • Mentions.so

Mentions.so is primarily a prompt-tracking platform that allows you to measure visibility, average position, citations and sentiment across prompts that matter to your business. You can also inspect the URLs used as sources in individual answers, helping you identify publications and pages that influence AI responses in which competitors appear but your brand does not.

Typically, these might include branded prompts, category or product-related prompts, “best of” style prompts, as well as “versus competitor” prompts.

This is especially useful when trying to boost performance where your competitors are appearing but you are not. There are many other AI search tracking tools on the market—these are just the ones we use day to day.

Above: The prompt view in mentions.so which measured average position, sentimment score, visibility, competitors and GEO location

How Do I Attribute Revenue Increases to AI Search Performance?

If the conversion came directly from AI referral traffic, attribution is relatively straightforward within GA4. However, AI search is primarily about visibility and recommendations rather than immediate clicks.

A Similarweb study found that users who received a ChatGPT brand recommendation were around 2.5 times more likely to visit that brand’s website within seven days. Around 56% of those AI-influenced visits occurred via branded search rather than a directly attributable AI referral.

Although the study only covered the finance, travel and beauty sectors and should not automatically be generalised to every industry, it illustrates why referral traffic alone may understate AI’s wider commercial impact.

As mentioned at the start, AI search tracking is considerably more difficult than measuring traditional Google search because of the generative nature of responses. Modern AI search systems typically combine information retrieval and generative AI.

Google’s generative search features, for example, build on its existing Search ranking systems alongside retrieval-augmented generation, or RAG, and query fan-out techniques. Other AI platforms use different retrieval and grounding methods, making results harder to measure consistently.

Rather than simply measuring mention or citation counts, it is often more valuable to understand what your brand is being mentioned for and whether those mentions are positive, neutral or negative. Prompt tracking aligns particularly well with wider commercial objectives.

Some examples are below:

  • “Where is the best place to buy dog kennels in the UK?”
  • “Which provider is best for over-50s life insurance?”

The first prompt relates to a specific product category, while the second targets a defined audience segment: people aged over 50.

Aligning tracked prompts with specific areas of the business makes it easier to compare changes in AI visibility with revenue and conversion trends in the same categories.

How Can I Tell Whether an Uplift in Conversions Is Linked to AI Search Activity?

Although commercial data points such as sales and revenue are what you ultimately want to measure, it’s important to measure metrics such as mentions, citations and specific prompts for the categories or areas of the business you are working on from an AI search perspective.

In terms of ensuring the uplift is not something that would have happened anyway, you need to create a control group. For ecommerce sites, you should pick two categories, matched as closely as possible in terms of size, range, pricing and other relevant factors, and avoid doing this around seasonal changes in demand.

Focus all your activity on one category and ensure you do not make other changes to the control category in the meantime, for example, adding more content or building more links.

After three months, measure changes in revenue and AI search visibility across both categories to see if there is any uplift. A useful way to compare performance is to subtract the percentage change in the control category from the percentage change in the category you worked on.

This provides stronger evidence that your activity contributed to any uplift, for example, increased revenue or conversions. However, it should not be treated as absolute proof that AI search alone caused the increase, as the work may also influence traditional organic search and other marketing channels.

AI Search Is Here to Stay

If you are just beginning to measure AI search, the number of platforms and metrics involved can seem overwhelming. However, AI search is not going anywhere. OpenAI says ChatGPT now has more than 900 million weekly active users, while Google is introducing dedicated reporting for its generative Search features.

The answer is not to search for one perfect AI search KPI. Instead, businesses should use a balanced scorecard combining prompt visibility, mentions, citations, cited pages, sentiment, referral traffic, branded demand and commercial outcomes.

Together, these metrics provide a much more complete view of whether AI search is increasing a brand’s visibility and contributing to business growth.