Key Takeaways
- Use AI attribution models like Markov chains or Shapley values to finally put a real number on the value of non-click interactions in your conversion paths.
- Pull data from all your touchpoints, voice search logs, visual search queries, in-app discovery, to build a customer journey map that reflects how people actually find you.
- Focus on the user experience inside your AI discovery paths. Make sure the content AI surfaces is relevant and easy to use, because that’s what creates deeper engagement.
- Set clear KPIs for AI-driven discovery, like time on content or brand affinity scores, that go beyond the obsolete last-click metrics.
- Reallocate your marketing budget based on the full-funnel value AI attribution reveals, moving money to the early-stage discovery channels that build long-term customers.
Amelia had a problem that was costing her sleep. She was the Head of Growth for “Veridian Botanicals,” a DTC brand selling sustainable home goods, and while their marketing budget kept going up, the ROI felt… off. Their attribution models, almost entirely based on the last click, were telling a story that just didn’t match the brand’s real-world organic growth. People were talking about Veridian in online groups, sharing their latest eco-friendly finds, and even using visual search tools like Google Lens to ID their product designs. But the reports? They just credited a generic retargeting ad or a direct site visit. “It’s like we’re only seeing the final handshake, not the entire conversation,” she’d complain to her team. The problem was obvious: how do you measure the value of all this new AI-driven discovery when the journey is happening far beyond the click? Veridian Botanicals was all about sustainable living, and their customers tended to be early tech adopters. They weren’t just clicking banner ads. They were asking smart speakers, “Hey Google, where can I find ethically sourced beeswax wraps?” or using Pinterest’s visual search to find things with a similar vibe to a Veridian product they saw at a friend’s house. These interactions, all powered by algorithms, were clearly building brand awareness. Amelia just couldn’t prove it with a dollar figure. Her first-touch and last-touch models completely ignored these critical, early discovery moments. Most marketing dashboards, still stuck on last-click because it’s simple, completely miss how modern customers behave. A 2025 IAB report found that less than 30% of marketers believe their current attribution actually captures the full customer journey, especially with all the new AI discovery tools. You have to understand the *weight* of these touchpoints, particularly when an AI is doing the work of surfacing your brand to a potential customer. Amelia decided it was time to blow up their old strategy. She kicked off a project to rebuild Veridian’s attribution from scratch, starting with a deep dive into their customer data platform (CDP). The objective was to track every possible interaction, not just clicks with a UTM tag. This meant a massive data-integration project, pulling in everything from voice assistant analytics and visual search logs to social listening tools and their own internal site search data. It was a marathon. “We need to see the ghost in the machine,” she told her data science team, “the invisible threads connecting discovery to conversion.” One of the first ‘aha’ moments came from their voice search query analysis. People were asking things like, “What are the best non-toxic cleaning products for my kitchen?” and Veridian’s “Earth-Friendly Cleaning Kit” was being surfaced as an answer. These weren’t branded searches. They were problem-solving questions. No traditional analytics setup would ever give credit to “being recommended by a smart speaker,” but Amelia knew this was a massively influential first touch. Quantifying its influence was the real challenge. The team started digging into more sophisticated attribution models, moving past the simple linear or time-decay models that can’t handle non-click events. They began running experiments with data-driven attribution in Google Ads and building their own custom machine learning models. Specifically, they started working with Markov chain models and Shapley value attribution. A Markov chain model, for instance, calculates the probability of a user moving from one stage to the next, which lets you figure out the actual contribution of each touchpoint, click or no click. Shapley values, which come from cooperative game theory, provide another method by fairly assigning credit across all touchpoints, analyzing every possible combination of interactions that could have led to the sale. “This is where the magic happens,” Amelia explained to her CMO, pointing to a new dashboard. “Look here: a customer used visual search on Pinterest to find our ‘Recycled Glass Tumblers,’ then later asked Alexa about ‘sustainable glassware brands,’ and only then finally clicked a retargeting ad on Instagram. Our old model gave 100% of the credit to that Instagram ad. Our new model, using Shapley values, shows visual search was worth 30%, the voice search was 25%, and the Instagram ad was 45%.” This kind of granular data finally started to paint a true picture of the customer journey, revealing the huge role AI discovery was playing. The new data showed something huge: customers who found Veridian through AI discovery channels (voice, visual, or AI-powered content recommendations) had a much higher lifetime value (LTV) and lower churn. These users were on a mission, looking for solutions that fit their values, which meant they were already bought-in before the first click. A 2026 eMarketer forecast projects voice commerce will make up nearly 18% of all e-commerce by 2027. Any brand not tracking these interactions is essentially ignoring a huge, growing segment of their customer base.
Another interesting insight came from their internal site search data. Customers using long, natural language questions on Veridian’s own website, phrases that looked a lot like voice search queries, had a 2.5x higher conversion rate than people using simple keywords. This told them that users accustomed to AI discovery expect that same intelligent experience everywhere. In response, Veridian rolled out an AI-powered site search tool that could understand natural language and give personalized recommendations. This fix, directly informed by the attribution data, improved the user experience and pushed conversion rates up. This new attribution strategy had immediate business implications. Amelia’s team started moving money around. They cut spend on some underperforming last-click channels and put a lot more investment into creating content optimized for voice search, visual search, and long-tail SEO. They also started working with influencers who were good at product placement in visually dense content, knowing those images could trigger a visual search. “It felt weird at first,” Amelia admitted, “putting budget into channels that don’t give you an immediate click-through. But the data from our new model was impossible to argue with.” The numbers spoke for themselves. Within six months, Veridian Botanicals boosted their overall marketing ROI by 15%, not by spending more, but by spending smarter. Their customer acquisition cost (CAC) for users from AI-driven discovery channels fell by 20%. This happened because these customers were better qualified and needed less convincing to make a purchase. The brand also saw its organic search rankings improve for broad, informational queries, a direct result of investing in content that answered those early-stage questions. Measuring AI discovery means adding up the influence of all the intelligent touchpoints that guide a customer to your brand. It means you have to finally ditch the outdated last-click models that can’t handle the complexity of how people actually find and buy things now. For Veridian Botanicals, this translated into a much deeper understanding of their customers and a far more efficient way to spend their marketing budget, proving that real value is measured far beyond the click. Getting this right isn’t just about better reports. It’s about finding the money to reinvest in the early-stage discovery that builds real brand loyalty, something last-click attribution will never show you.
What is AI-driven discovery in marketing?
It’s how people find things without typing in a traditional search box. This covers everything from asking a smart assistant a question (voice search) to using a phone’s camera to identify a product (Google Lens or Pinterest’s visual search). It also includes the personalized product and content recommendations that AI serves up on e-commerce sites and in social feeds based on your behavior.
Why is traditional attribution insufficient for AI discovery?
Because models like last-click give 100% of the credit for a sale to the final touchpoint, which is usually a click. This completely ignores the early, non-click interactions that build intent, like a voice query or a visual search match. As a result, you get a distorted view of what’s actually working and you misallocate your budget.
What advanced attribution models can measure AI discovery?
Data-driven models like Markov chains and Shapley value attribution are built for this. Markov chains work by analyzing the probability of a user moving between different stages in their journey, letting you quantify each touchpoint’s role in that progression. Shapley values, which come from game theory, assign credit more fairly by calculating a touchpoint’s marginal contribution across every possible sequence of events.
What data sources are important for measuring AI discovery?
You need to pull data from everywhere you can. The most important sources include logs from voice assistants, visual search platforms, and your own internal site search. You also need to integrate social listening tools and, most importantly, have a customer data platform (CDP) that can stitch all of this together into a single, cohesive view of the customer’s journey.
How can marketers optimize content for AI-driven discovery?
Focus on answering questions, not just targeting keywords. Create detailed, long-form content that directly addresses natural language queries people might ask a smart speaker. Use high-quality, well-tagged images and video so visual search algorithms can identify your products. And behind the scenes, use schema markup on your site so AI platforms can better understand the context of your content and serve it up as a solution.
