David Chmelař: turning AI's threat into revenue at Heureka

Three years ago, many were predicting a swift death for the Czech price-comparison site Heureka under the onslaught of artificial intelligence. But David Chmelař, who took over the company shortly before the boom in language models, is convinced: AI isn't a killer but an ally, one capable of bringing Heureka a “golden era.”

According to Chmelař, around half of people who shop online already use AI in some way when choosing what to buy — not at the purchase stage, but earlier, when they're looking for ideas. Users ask ChatGPT, Gemini or Perplexity, for instance, which stroller would suit them, and get detailed, context-aware recommendations.

The problem, in the view of Heureka's CEO, starts right where the shopper needs to move on to actually buying something. Language models are built first and foremost for large markets — the US and potentially China — and for universal use cases that look the same everywhere, like programming. E-commerce, by contrast, he says, is intensely local: Czechia alone has around 50,000 online stores, a unique environment that's hard for outside models to navigate.

This, Chmelař believes, is exactly where Heureka's new role comes in. According to a study of more than a million AI-model queries, the service is the fourth most frequently cited source for ChatGPT, behind only Wikipedia and Reddit, and it ranks second in Perplexity. “The models aren't confident about the Czech market, it's opaque to them. That's why we're seen as an authoritative source with a huge trove of unique data,” he explains.

What's at stake specifically are data sets that ordinary AI web-crawling can't reach — precise information, for example, about stock levels and colour availability at specific stores. It's on monetising exactly this kind of data that Chmelař is building the company's future business model. Right now, the online stores themselves pay to be listed on Heureka, but in future, he expects, big AI companies will join them, using the service as a “pass-through” source of data on the Czech market.

According to Heureka's CEO, this benefits both sides: the company monetises data that's hard to gather, while AI models get precise, up-to-date recommendations for their users. As an analogy, he points to Sportradar, which tracks the results of 80 million sporting matches a year and already has contracts worth hundreds of millions with major AI players — deals over e-commerce data, Chmelař believes, could work the same way in the future.

That said, he admits, for services like Heureka worldwide this will mean the end of their familiar shape: full-fledged shopping websites will give way to dynamic interfaces and plug-ins, where even the company's own brand may go unnoticed by the end user.