TL;DR Summary:
GA4 Is Only The Floor: Google Analytics now includes an AI Assistant channel for some chatbot referrals, but it still misses much of the traffic because many AI searches end without a click and referrers can disappear.Track Three Layers: The framework separates direct attribution, influenced attribution, and long-term visibility so marketers can connect AI search activity to leads, revenue, and brand presence.Watch The Proxy Signals: Rising branded search, more direct traffic, shorter sales cycles, better win rates, and stronger citation rates all suggest AI search is helping even when clicks are invisible.How do you know if AI search optimization is actually working? Kristina Frunze, founder of the SEO and AI search agency WebView SEO, built an AEO attribution framework to answer that question for her B2B SaaS clients, and it matters now because standard analytics tools miss most of the traffic coming from tools like ChatGPT. Her three-layer model gives marketers a way to show leadership what AEO spending produces, even when the data stays incomplete.
Why GA4 sets the floor for your AEO attribution framework
Google added an AI Assistant channel to GA4 reports earlier this year, giving marketers a direct way to spot AI-generated traffic for the first time. Frunze says this number still represents a floor, not the real total. Most search interactions on LLMs end without a click. Referrer headers often get stripped on mobile and desktop apps, so visits land in GA4 as direct traffic instead of AI traffic. People also switch devices mid-search, which breaks the tracking trail further. Any AEO attribution framework built only on GA4 will undercount your actual results.
Layer one: direct attribution inside your AEO attribution framework
Direct attribution covers the leads and traffic you can actually point to. Frunze recommends setting up an “AI search” channel inside Salesforce or HubSpot to tag leads with LLM referrers. She also advises adding “LLMs” as an option in the “How did you hear about us?” dropdown on demo and contact forms. Sales reps can ask directly on calls too, though Frunze calls this the more difficult route. These numbers stay small compared to your real LLM-driven traffic, but they connect to closed deals and revenue, which makes them worth tracking.
Layer two: influenced attribution and what it reveals about AEO impact
Influenced attribution covers the effects of AEO you cannot tie to a single click. Frunze points to four signals worth watching. Direct traffic and demo requests can rise as people copy your domain from an LLM answer instead of clicking a link. Branded organic search traffic can grow days after someone researches on an LLM, then searches your brand name on Google. Sales cycles often shorten because prospects arrive already informed about pricing and integrations. Win rates against competitors can improve when LLMs recommend your product based on accessible, accurate information. None of these carry a direct dollar value, but Frunze says they shape your cost per lead and sales cycle length over time.
Layer three: the future moat inside your AEO attribution framework
The third layer tracks longer-term positioning rather than immediate traffic. This includes AI share of voice, citation rate (how often LLMs reference your brand), and sentiment (whether LLMs describe your brand positively). Frunze recommends tracking these monthly for at least six months before drawing conclusions. The goal is making sure your brand shows up, and shows up accurately, in AI conversations happening 12 to 24 months from now. Since this layer depends on consistently checking how often and how favorably a brand gets cited across ChatGPT, Gemini, Perplexity, and Claude, tools like Nuwtonic SEO have emerged to handle that ongoing cross-platform monitoring, which is difficult to track manually at scale.
What this means for your AEO attribution framework going forward
Frunze presents this as one practitioner’s applied model, not an industry standard, and she calls it a work in progress given how fast AI search changes. Watch your GA4 AI Assistant numbers as a starting point, but build out CRM tagging and self-reported form data alongside it. Track branded organic traffic, sales cycle length, and win rates monthly, and give the pattern at least six months to show itself clearly. For teams building out that third layer of AI share of voice and citation tracking, a platform like Nuwtonic can automate the monthly monitoring of brand mentions across ChatGPT, Gemini, Perplexity, and Claude, giving marketers the citation-rate data Frunze’s framework depends on.
