AI discovery is growing, but clicks cannot show all its influence
Your brand can help a buyer choose a product inside ChatGPT and receive no credit when that buyer later converts through paid search. Adding an AI category to your reporting can recover identifiable visits currently assigned elsewhere. But measuring the earlier recommendation requires a broader approach because the buyer may never have clicked it.
That gap already matters for marketing planning: Salesforce’s 2026 State of Marketing Report found that 88% of marketers say they are optimizing for AI-driven discovery experiences. The figure measures adoption of those practices. To assess their business effect, teams need to connect evidence about discovery with evidence about observable conversion paths.
These kinds of evidence answer different questions. Brand visibility inside an AI answer shows how buyers may encounter a company, while a recorded visit or conversion shows an observable interaction. A useful measurement framework connects the two while keeping clear what each can establish.
AI can shape a decision before a measurable visit occurs
To understand that distinction, start with cross-channel measurement: combining information from marketing channels to assess their joint effect on customer behavior and business results. Familiar inputs include clicks, referral traffic, website sessions, and conversion paths. They describe a sequence in which someone searches for information, follows a link, visits a website, and converts.
Tracking connects those interactions through cookies, referral paths, and UTMs, tags added to URLs to identify campaign traffic. Multi-touch attribution models then distribute conversion credit across several recorded interactions. Much of this measurement answers a straightforward question: “Which channel drove the click?” The answer depends on buyers leaving a trail that analytics can observe.
AI discovery changes that condition because buyers can research and evaluate brands within an AI service. Three types of environment matter here. Each offers a different route to an answer, which affects both where a company can appear and what a measurement system can record.
In AI chat interfaces, buyers can ask questions, compare vendors, request summaries, and receive conversational recommendations. ChatGPT, Claude, Grok, and Microsoft Copilot belong to this category. These exchanges can help a buyer discover a company and narrow a shortlist before visiting any company website.
AI-powered search experiences bring generated summaries into an existing search environment. Google AI Overviews, Google AI Mode, and Bing Copilot can answer enough of a question to reduce the need for an external visit. Google AI Overviews also illustrates a measurement difficulty: generated information and conventional search behavior can occur together, making their contributions hard to separate.
AI-native answer engines organize discovery around AI retrieval and answers supported by citations. Perplexity AI is an example; its presentation differs from a conventional keyword-search results page. Across all three environments, buyers can compare products and select vendors while staying inside the AI service.
Staying inside the service creates two related gaps: discovery without a click and influence without immediate traffic. An answer may satisfy a research question, while a recommendation or citation may shape a later decision. For example, a buyer might ask ChatGPT for leading customer data platforms, evaluate its recommendations, and later convert through branded search or paid media. Attribution can credit that later interaction while missing the earlier AI influence.
Even when AI generates a visit, inconsistent referral information can cause analytics to classify it as direct or unassigned. Other measurable AI activity may appear within organic search or referral reporting, and AI platforms provide less standardized reporting than established advertising and analytics systems. AI can also shape what a buyer searches for before conventional search or paid media becomes visible. Along with blended search experiences, these conditions leave existing channel categories with an incomplete account of discovery.
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.
Discovery signals answer different measurement questions
To account for discovery, teams can organize measurement into five layers: visibility, citations, referral traffic, prompt-based discovery, and conversion influence. This proposed framework places AI alongside search, paid media, social, and email. It extends traffic attribution by examining how discovery contributes to decisions.
Each layer adds a specific observation. Visibility measures whether a brand appears in generated answers or recommendations, while citations show whether an answer references or links to its content. Referral traffic identifies measurable visits from AI platforms. Prompt-based discovery connects appearances to particular questions and topics, and conversion influence examines earlier AI interactions within journeys that end in a conversion.
Separating these observations helps teams decide what to investigate. Rare appearances in relevant answers raise questions about discovery and positioning; AI referrals that seldom convert raise questions about visitor behavior. AI interactions earlier in recorded conversion paths raise questions about their relationship with subsequent channels. Combining all three into one performance measure would hide the different decisions they support.
Connect brand appearances to the questions buyers ask
To investigate discovery and positioning, start with mentions, citations, and prompts. Mentions count brand appearances in answers, recommendations, and comparisons, while citations identify content used as a referenced or linked resource. Prompts are the questions or instructions people submit to an AI system; tracking them connects appearances to topics and use cases.
Consider an illustrative company that appears often for “best CDP for ecommerce,” but rarely for “top customer data platforms.” CDP means customer data platform, so the difference suggests stronger visibility for an ecommerce use case than for the broader category question. A marketing team can use that observation to review content coverage, messaging, and positioning priorities. Establishing a contribution to conversions requires journey data.
Keeping prompts attached to appearances makes that review more useful because an overall mention count can hide differences between use cases and broad category research. Teams can examine the category relevance and perceived authority reflected in generated answers. They can then interpret frequent appearances in the context of the questions that produced them.
Examine how answers present the brand
Once teams know where a brand appears, position and sentiment help explain how it is presented. Semrush, which sells Semrush Enterprise AIO and benefits commercially from demand for AI-discovery measurement, offers measures of where a brand appears when an answer names several companies. The platform also measures sentiment: whether the answer presents the brand positively, negatively, or neutrally. Its AI share of voice, or AI SoV, compares the brand’s appearance frequency with competitors.
These measures matter because generated answers can contain more detail and evaluation than conventional search snippets. In an illustrative unfavorable pattern, answers might repeatedly favor unnamed competitors or associate a brand with negative reviews, pricing concerns, or outdated capabilities. A mention count would record its presence, while sentiment and competitive context would reveal a reputation-management concern. The team could then examine the repeated framing and the content associated with it.
Semrush Enterprise AIO is also recommended for the broader approach, covering mentions, citations, visibility, share of voice, sentiment, and connections to marketing performance. Semrush has a commercial interest in that recommendation. These measures describe how a brand appears in the answers examined; establishing whether a particular buyer saw those answers or converted because of them requires additional observations.
Use crawl activity to investigate content collection
Before content can appear in an answer, an AI system may access it through crawling, the automated collection of website pages. Website access logs record that activity. The proposed process begins by obtaining those logs and uploading them into a server log analysis tool.
Within that tool, the team identifies bot traffic through user agents, the identifiers software sends with requests. It then looks for AI crawlers such as OpenAI’s GPTBot and ClaudeBot, examines how often they visit, and identifies the pages they select. If GPTBot visits several pages, the logs record that collection activity even when no buyer referral accompanies it.
The proposed interpretation is that GPTBot gathers material that may later appear in ChatGPT responses, with crawl patterns indicating collection frequency and page priorities. Under that interpretation, frequent AI-bot visits can suggest active ingestion for training and response generation. Teams can combine crawl observations with mentions, citations, and referral data to investigate which accessed pages also appear elsewhere in the discovery process.
A frequently crawled page therefore gives teams a useful starting point for citation and visibility checks. The crawl establishes access to content; answer tracking establishes whether the content appears in the answers examined. Keeping those events separate prevents teams from treating collection activity as buyer exposure.
The same separation applies across the framework. Each signal supports a specific observation, while further evidence is needed to answer the next question.
| Signal | What it concerns | Additional observation needed |
|---|---|---|
| Crawl activity | Collection of website content | Answer inclusion or buyer exposure |
| Mentions, citations, position, sentiment | Observed answer visibility and framing | Buyer exposure, website visits, or conversion participation |
| Referral traffic | Measurable visits from AI platforms | Discovery that occurred without a click |
| Recorded assists | Observable interactions before conversion | Earlier unobserved influences or evidence of causal revenue contribution |
These distinctions give teams a practical sequence for investigation. Crawl observations can identify pages worth checking in citation data, while prompt tracking explains which topics accompany brand appearances. Referral and conversion data can then show what happens within observable website journeys. Each step adds evidence without changing what the earlier observations establish.
A dedicated AI channel organizes observable conversion paths
Once an AI platform sends a measurable visit, conventional attribution has a clear role. A dedicated AI channel helps distinguish visits from ChatGPT, Perplexity, or Copilot from other acquisition activity. Without that separation, relevant sessions can remain scattered across existing reporting categories, making visitor behavior harder to assess.
The proposed setup begins by defining AI as a separate reporting channel and creating a category such as “AI Discovery” or “AI Assistants.” In Google Analytics 4, or GA4, teams can use custom channel groups to recategorize identifiable AI traffic, then create a customized report for that group. The report separates those referrals from organic search, paid media, social media, and email.
With those visits grouped, teams can examine traffic volume and compare visitors arriving from different AI platforms. They can assess which platforms are associated with stronger signs of interest in buying and whether AI-assisted visitors convert differently from organic or paid visitors. The reporting group also gives them a place to investigate remaining classification problems.
Questions about content exposure still require the discovery data described earlier. Citation and visibility observations identify which content appears most often in AI responses, while the channel report describes measurable website sessions. Comparing those views lets a team investigate the relationship between exposure and traffic while allowing for citations that generate no referral.
Once referral tracking is configured, teams can connect those visits with conversions. Start with immediate conversion activity, then examine whether a recorded AI interaction preceded a later conversion. An assist is a channel interaction that appears in the conversion path before the final interaction. Assisted-conversion analysis in GA4 and other analytics platforms can therefore show how identifiable AI visits participate in longer journeys.
That analysis expands reporting from the channel that completed a conversion to the recorded channels that participated. Teams can examine which AI platforms appear in conversion paths, which channels accompany them most often, and how conversion behavior differs from other acquisition sources. They can also investigate relationships between AI visibility, branded search, and later revenue activity, while treating causation as a separate question.
The customer data platform journey introduced earlier illustrates why that distinction matters. Suppose the buyer discovers a company through a ChatGPT recommendation and leaves without converting. A week later, the buyer searches for the brand, clicks a paid Google ad, and submits a demo request. Last-click attribution assigns the conversion entirely to paid search, even though the earlier recommendation helped shape consideration in this illustrative journey.
Assisted reporting can reveal that earlier AI interaction when the available data contains a visit or another recorded event that can be included in the path. A recommendation that produces no recorded interaction remains outside the path, even after the team creates an AI channel. The example explains a gap in conversion credit, but reconstructing the journey depends on the interactions captured.
Within captured paths, AI interactions may precede conversions through paid search, organic search, or direct traffic. They may also precede demo requests and later pipeline creation, when a potential deal enters the sales process. Examining these sequences helps teams understand how channels work together. The AI channel supplies the identifiable traffic and recorded interactions needed for that analysis.
Unified reporting connects observations across systems
Those interactions often sit in separate systems, creating another reporting problem. In an illustrative review, a customer relationship management system, or CRM, shows one result, GA4 shows another, and paid media platforms report different numbers again. AI discovery may have no clear representation in any of them. A budget decision based on one isolated report would therefore use only part of the available evidence.
This fragmentation reflects journeys spanning AI assistants, organic search, paid media, social platforms, email, and direct traffic. Because each system records a different portion of activity, bringing the information together starts with a centralized measurement layer. The proposed process is to ingest data from the separate systems, normalize it into consistent forms, and connect the available data points in a unified dataset.
Customer data platforms, GA4, enterprise analytics environments, and data warehouses can support that process. Centralization gives teams a common place to compare acquisition activity, customer records, and conversions. The aim is a “single source of truth”: a shared basis for analysis that preserves the meaning and limits of the contributing records.
Once the data is centralized, reporting tools such as Google Data Studio can blend datasets into unified dashboards. Those dashboards can show journey-level trends across AI visibility, paid media, search, social, CRM activity, and conversion behavior. Bringing the datasets together makes relationships easier to investigate than they would be across disconnected platform reports.
The dashboard should also place discovery and reputation measures alongside traditional search engine optimization, or SEO, reporting. A change in AI sentiment raises a different question from a change in organic website traffic, even when both concern the same brand or content. Keeping both visible helps teams examine how the company appears during research and what happens when buyers later interact with its channels.
To support that investigation, the dashboard must keep the type of observation clear wherever datasets meet. If a page receives crawler visits, appears in AI citations, and later receives referral traffic, the report should present those as separate events whose relationship can be examined. Combining them cannot create a record of an unseen buyer interaction or automatically assign conversion credit.
The bottom line
AI discovery creates a measurement gap that a new channel category alone cannot close. A dedicated AI channel makes identifiable visits and recorded assists easier to assess. Visibility, citations, and sentiment show how a brand appears during research. Neither establishes the full contribution of AI to revenue.
For business leaders, the priority is to connect these observations without treating them as interchangeable. Ask teams to distinguish what buyers could have encountered, what interactions were recorded, and what those interactions show about conversion paths. That discipline makes reporting more useful for investment decisions and reduces the risk of assigning credit the evidence cannot support.
Start by separating identifiable AI traffic, tracking brand appearances against relevant buyer questions, and bringing those findings into existing performance reviews. Expand the measurement effort where it can inform a specific decision about content, positioning, reputation, or channel spending.
The goal is a clearer basis for action, not a claim that every influence can be measured. Better reporting should help leaders see where AI participates in discovery, where observable journeys connect to business results, and where uncertainty remains.
A project in mind?
Schedule a 30-minute meeting with us.
Senior experts helping you move faster across product, engineering, cloud & AI.


