Lendi pilots AI discovery measurement as chatbot ads deliver middling returns

Australian home-loan company Lendi Group is putting more of its AI effort into understanding how borrowers discover it than buying chatbot ads. In an October 19 AdExchanger report, growth general manager Michael McCash described middling advertising returns alongside a different signal: prospective customers arriving better informed after researching loans with AI chatbots.

Lendi is a pilot customer of Mutinex’s AI Visibility tool, which examines how brands appear in chatbot responses. The company’s measurement effort sits within a paid-media operation spending more than $10 million annually, according to McCash. It combines an in-house system for assigning credit across customer interactions with Mutinex’s marketing mix modeling.

Paid placements and informed arrivals

Lendi has made only a limited investment in chatbot advertising so far. McCash told AdExchanger that its marketing mix modeling—a method used to assess the contribution of different marketing channels—showed middling performance. The cost per thousand ad views and the cost of acquiring customers were too high to justify substantial spending.

Yet McCash sees customers reaching Google Search or Lendi’s own website having already done considerable research through large language models, the systems behind AI chatbots. He described these prospective borrowers as more knowledgeable and further along in their buying process. Their first visible visit to Lendi can therefore follow research conducted inside another platform.

Probing answers instead of counting clicks

That early research is difficult for Lendi to observe. McCash said AI platforms make it hard to see how particular keywords or searches surface products. Marketers accustomed to clicks, impressions and search-term data have less visibility into the questions people ask chatbots and the brands those answers present.

Mutinex’s AI Visibility approaches that gap by sending large numbers of relevant prompts to language models and analyzing the answers. A search for home loans in Western Australia is one example cited by AdExchanger. The method examines generated responses rather than drawing on a dashboard of actual customer searches.

Lendi supplements that exercise with two other sources of information. They add competitive brand tracking and direct customer research to the picture, rather than asking the prompt-testing tool to answer every measurement question:

• Tracksuit measures brand metrics and tracks how AI treats Lendi relative to competing brands.
• Askable provides customer and mass-market surveys, adding respondents’ perspectives to the company’s measurement work.

A budget lesson from broadcast TV

Lendi has already seen measurement challenge its assumptions about another channel. It had been a major buyer of streaming television and data-driven video on demand. But its marketing mix modeling showed that free-to-air television had received too little investment and was an efficient way to spend, McCash said.

The company redirected money from less efficient channels into television, which continued to increase its return on investment. McCash’s explanation was partly about buyer habits: media buyers can favor channels they personally use, while marketers tend to trust the measurements they can readily see.

For AI discovery, McCash expects platforms to supply more useful information as their advertising businesses grow. He wants anonymized insights into the questions people ask and how brands appear. That would give marketers a clearer view of the research customers complete before arriving—a view Lendi is currently assembling through external tools and surveys.