CMOs Can't Prove AI Search Visibility Drives Revenue
Marketing leaders are piecing together proxy signals and custom models to bridge the gap between LLM citations and actual sales.

Chief marketing officers face mounting pressure to demonstrate that their investments in AI search visibility translate to revenue, but the measurement tools don't exist yet.
According to John Barham, managing partner at performance media agency Roast, every CMO his firm works with has faced "excruciating pressure" over the past 18 months from boards and investors to prove the commercial value of appearing in large language model responses. While 73% of marketers have invested in tools to monitor AI visibility, according to a recent survey, platforms like SEMrush, Profound, and Scrunch can only estimate how LLMs present brands to users—they can't connect those citations to sales.
The triangulation approach
Without purpose-built attribution tools, marketing teams are constructing makeshift measurement frameworks. Alicia Yoon, founder and CEO of skincare brand Peach & Lily, described the process as "triangulating" proxy signals and data points to estimate the impact of their brand's profile within Google AI Overviews and ChatGPT on inbound leads and digital sales.
B2B SaaS company Rippling has developed a more sophisticated approach. Head of growth Neel Murthy said his team combines visibility measures from Profound and AirOps, conversion data from paid ChatGPT ads, web traffic from branded and unbranded queries, and a custom media mix model built on Google's open-source Meridian platform. "All these systems allow us to triangulate value," Murthy explained.
The stakes are rising. A Demandbase survey found that ChatGPT-referred visits to B2B brands jumped 303% from approximately 645,000 in June 2024 to 2.6 million in June 2025 as business buyers adopted AI tools for research. Yet linking that traffic to commercial outcomes returns marketers to attribution challenges that have "plagued" them for years, according to Demandbase CMO Rachel Truair.
Building custom models
Some agencies are turning to statistical modeling. Roast has deployed Google's Causal Impact model, an open-source solution using Bayesian logic to connect search inputs with business outcomes. "There isn't the tech or tools to draw a clear line from visibility on a given LLM and sales going up," Barham said.
The Interactive Advertising Bureau released standardized measurement guidelines in early 2025, and Google added a Search Console feature providing generative AI performance data in June. Mulenga Agley, founder and CEO of marketing agency Growthcurve, said the only viable path forward is for marketers to build their own models using first-party data while excluding known AI traffic.
In March, Stagwell media agency Assembly partnered with startup Emberos to develop a "Search+" solution. Dan Roberts, global senior vice president of search at Assembly, said the agency is working to integrate conversion, revenue, and lead data into the platform, though he declined to provide a release timeline.
Why it matters
The measurement gap forces CMOs to make significant budget decisions without clear ROI data at a time when AI search is fundamentally changing how consumers discover and research products. Companies that can't demonstrate commercial impact from AI visibility may redirect resources to channels with clearer attribution, potentially ceding ground in an emerging search paradigm. The lack of standardized measurement also creates competitive advantages for organizations with the data science resources to build custom attribution models.
A probabilistic future
Some marketers are accepting inherent limitations. Peach & Lily's Yoon noted that conversion tracking becomes impossible when users read AI search results on one platform but purchase through Amazon or other marketplaces. "So much [conversion data] is not directly captured," she said. "There's going to be a gap."
Barham suggested CMOs should reframe their thinking about search in "probabilistic" terms rather than expecting the clear performance metrics that once defined the channel. "Our teams have had to get more comfortable explaining to clients the value of their work and their media activity when there isn't a CSV they can download from a platform," he said.
These details were first reported by Digiday.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call