Broken Store Chatbots Are Costing Polish Retailers Sales, PAYBACK Study Finds

Online stores are rolling out AI chatbots en masse, betting on a sales boost, but a new PAYBACK study shows the effect can backfire. As many as 59 percent of Polish internet users abandoned a purchase when a store's built-in assistant failed to answer their question or resolve the issue they raised.

The study, commissioned by PAYBACK, reveals a gap between the trust people say they place in artificial intelligence and their actual shopping experience. Poles are happy to ask AI for advice when choosing products, but when they run into a poorly performing bot on a specific store's site, they would rather close the browser tab than keep talking to a machine that doesn't understand their problem.

Where Chatbots Fall Short

The data shows that nearly 40 percent of shoppers don't turn to any AI-based shopping assistant at all, even though these tools are increasingly offered by e-commerce platforms. Among those who do try them, 59 percent end up abandoning their cart once they hit a question that falls outside the bot's standard script.

That figure should worry e-commerce teams investing in customer service automation. A chatbot meant to shorten the buying journey and take work off human consultants often ends up dragging out a customer's frustration instead, and the sale is lost rather than completed.

Trust with Reservations

A separate strand of the study looks at how Poles feel about AI-generated recommendations themselves. 80 percent of respondents say they trust AI's shopping suggestions, at least in part. At the same time, as many as 94 percent still verify those recommendations on their own instead of accepting them blindly.

This pattern suggests Poles treat AI as a starting point for a purchase decision rather than the final word on it. Consumers want the algorithm to narrow down the options and point them in a direction, but the final call still belongs to them.

Consumers aren't so much afraid of manipulation as they are correctly assuming that the recommendations generated by algorithms inside store apps are based on less data - Grzegorz Kielczewski, senior product expert at PAYBACK

In-House Bots Lose Out to ChatGPT

The study also points to a clear preference for general-purpose tools over solutions built into individual stores. 59 percent of respondents say they reach for outside tools like ChatGPT or Gemini more often than for the assistant offered by a given store platform, which only 23 percent of those surveyed actually use.

According to Kielczewski, this comes down to a rational calculation on consumers' part. A store's bot inherently operates on a limited set of products and offers from a single seller, so it can hardly be expected to give an objective comparison of prices or quality against the competition. General AI models, which draw on a wider range of sources, are simply more credible in that role.

What It Means for Retailers

For Polish e-commerce companies, the findings signal that simply having an AI chatbot on the site isn't enough. If the tool can't actually help, it doesn't build customer engagement, it actively pushes shoppers away from finishing the purchase. Investing in service automation without ensuring its quality can therefore lower conversion instead of raising it.

PAYBACK's findings fit a broader trend seen in the Polish market in recent months, where consumers increasingly use AI at the research stage of a purchase but keep their distance from tools controlled directly by the seller. Further market research will likely show whether stores can improve the quality of their assistants before customers switch for good to external chatbots as their default starting point for shopping.