AI search is breaking attribution, and the early AI ad landscape currently skews B2C. Gintarė Rimolaityte, CEO and co-founder at Trendos, explains where misattribution comes from and how to build measurement around organic AI visibility, paid ads exposure and business results.

No one questions that LLMs have upended search. Performance marketers know this better than anyone, and they’ve discovered firsthand how this shift has turned analytics on its head. A brand can be highly visible inside an AI platform but functionally invisible on an analytics dashboard. A marketer has no way of knowing if their products are being suggested by ChatGPT, Gemini or another model to solve a particular problem, let alone if that search leads to a site visit later on or a conversion. Evidence of original AI interaction is guesswork on a good day. In many ways this is a familiar performance-marketing problem set against a new backdrop, where the channel influencing demand can disappear from the reporting layer that is supposed to prove its value.

The visibility problem starts with attribution. AI search draws a distinction between organic AI visibility and paid AI placement, so marketers shouldn’t treat them as interchangeable. Organic visibility is shaped by the information and sources that models draw on when deciding which brands to mention. Paid placement sits alongside that experience, operating more like a conventional performance channel with targeting, bidding, conversion measurement and optimisation capabilities.

Neither form of visibility maps neatly onto the conventional search-click attribution model. An AI answer can influence a purchase without producing an immediately trackable session, while a paid placement can generate a measurable click without necessarily receiving the same level of attention as the answer itself. OpenAI labels ChatGPT ads clearly and keeps them separate from the answer itself. Advertisers can already optimise for conversions and track results with the OpenAI Pixel, but geographic targeting is still out of reach.

This becomes a basic diagnostic issue for performance teams. If an AI appears to be underperforming, is it really? Are you actually measuring the interaction that is influencing the customer?

AI advertising is still finding its performance model

Further complicating the situation is how varied the current advertising environment is between different platforms and audiences. While the majority of brands currently running ChatGPT ads are B2C, there is already a significant B2B segment. Professional services and developer tools dominate the B2B category, with business software alone representing almost 20% of observed impressions between March and August 2026.

The divide between B2C and B2B ads is in flux. As consumers turn to AI systems for product discovery and purchase research, retail and consumer categories will only grow as an AI ad category. There will be no single AI advertising market whose audience profile can simply be extrapolated across every model. Gemini has a different user and ecosystem relationship from ChatGPT. Claude skews professional and enterprise, while models such as DeepSeek and Grok introduce additional geographic and audience considerations.Scale still sets a baseline:  according to Similarweb, ChatGPT accounts for roughly 60% of AI chatbot traffic and Gemini around 25%, so both deserve broad monitoring, as does Google AI Overviews, especially for B2C brands. Beyond that, model-specific behaviour should determine where incremental measurement and media investment goes.

When brands operate internationally, these considerations multiply. A brand’s performance can vary widely between markets or models, meaning any overarching, ‘global’ AI visibility score can end up concealing the commercially important variation underneath it. There is little substitute for monitoring individual models in specific geographies.

The next measurement layer is outcome visibility

AI advertising is part of a long history of digital advertising, and there are relevant lessons the past can teach us. Early search and social advertising formats struggled when the commercial message felt too disconnected from the user experience. Performance usually improved as platforms developed advertising formats that felt increasingly native to the environment. Similar adaptations may not arrive quickly within AI. ChatGPT currently keeps advertisements clearly labelled and separate from the answer, which OpenAI claims is intended to protect user trust and keep advertising from influencing model responses.

For performance marketers, the challenge is demonstrating commercial results within an advertising environment designed around maintaining user trust. Then again, the closer the paid experience eventually becomes to the user's actual decision-making flow, the more important it will be to understand whether the placement contributes to the eventual commercial outcome.

In practice, measurement should start with the questions customers are actually asking. Identify the prompts that reflect those customer questions and track them consistently across relevant models and markets: how often the brand is mentioned, which sources are cited, how it is described and which competitors appear alongside it. On the paid side, UTMs and tools such as the OpenAI Pixel can connect clicks with downstream behaviour, but marketers still need visibility into the advertising environment around those results, including which prompts trigger ads and what competitors are running. Together, this creates three connected measurement layers: organic visibility, paid AI exposure and downstream business results.



Paid measurement, however, is only part of the picture. Attribution on the organic side is inevitably more difficult. If visibility improves after a new piece of coverage, stronger first-party content or an increase in reviews, marketers should be cautious about assuming one action caused the change. AI answers are probabilistic and competitors are changing at the same time. A better approach is to isolate specific changes, monitor related prompts and citation sources over a longer period, and treat movement as evidence of correlation rather than automatic proof of causation.

There’s no getting around the real measurement challenges. Nevertheless, there is an opportunity here to treat AI visibility as another measurable part of the acquisition system, with its own incrementality questions, market-level benchmarks and revenue outcomes. Like everything else in performance marketing, this is a discipline. Established early enough, that discipline can give brands a clearer view of where AI is actually influencing demand, even when the dashboard makes the whole channel look invisible.