← Back to blog
Communicatie & Telefonie

Chatbots answer only 12 percent of questions correctly: what that means for your customer contact

By Zarioh Digital Solutions4 min read
Share

The Nationale Voice Monitor 2026, a Dutch annual study, shows that consumers use chatbots widely but are rarely satisfied with the result. Only 12 percent of questions get answered correctly, and frustration with automated contact keeps rising. What this means for your customer contact setup.

You call your energy supplier or phone provider and are met with a chat window asking you to describe your question in your own words. You type it in, get an answer that misses the point, and then have to figure out yourself how to reach an actual person. That feeling is common enough that it now gets measured nationally, every year, in the Netherlands.

The Nationale Voice Monitor 2026, run by research agencies Y.digital and Markteffect on behalf of the Dutch Klantenservice Federatie (customer service federation), surveyed just over a thousand Dutch consumers about their experience with automated customer contact. The outcome is clear: chatbot and voice assistant use is no longer growing fast, but their quality still falls well short of what customers expect.

The numbers: heavy use, little satisfaction

Seventy seven percent of respondents have used a chatbot at some point, and just over seventy percent a voice assistant. That use is now widespread and stabilising. Slightly more than half, 52 percent, are open to automated customer contact, but with one clear condition: only if it actually solves their question.

That is exactly where it goes wrong. Consumers say only 12 percent of their questions to a chatbot or voice assistant get answered well. That is a small improvement on last year, but still far from the level needed for chatbots to count as a channel people can rely on. The researchers cite 80 to 90 percent as the point where automation genuinely helps rather than getting in the way.

Frustration keeps rising, despite years of investment

Long wait times remain the biggest irritation in customer contact, at 52 percent. Right behind it, at 50 percent, is a chatbot that does not work. Forty percent of consumers name not being able to transfer to a human agent as a major drawback of automated contact. Overall irritation with automated customer contact rose from 38 percent in 2025 to 42 percent this year, despite companies having invested in this technology for years already.

A social listening analysis of more than two thousand messages within the study confirms the picture. Customers mainly complain that the bot does not solve their actual problem, that getting transferred to a person is made difficult, and that answers stay shallow: often information they had already found on the website before reaching out.

Phone and email still resolve issues best

Despite years of steering customers toward chat, no chatbot makes it into the top three preferred channels. Phone and email still hold those spots, and email and WhatsApp are also growing in use while scoring well on actually resolving issues. For organisations actively pushing customers toward chat to relieve their phone lines, that is an awkward fact: the channels customers trust are not the channels the policy is steering them toward.

Why chatbots miss so often

The study does not point to the underlying technology as the main cause, but to how organisations deploy it. Many companies build automated contact around their own processes and cost savings, not around the customer's actual question. A language model often gets bolted on top of existing systems, even though it is best suited to language and intent recognition, and far less suited to being the source of the content itself. Without a well structured knowledge base of the organisation's own information, a chatbot ends up guessing from shallow patterns.

The recommended approach reverses that order. First properly capture and structure the organisation's own knowledge, then put a language model on top of it to make that knowledge accessible in natural language. The result is hybrid customer contact, where AI mainly supports and speeds up staff rather than replacing them.

What this means for your own customer contact

For a director or customer service manager, this matters more than a technical report on chatbot performance. It directly affects how customers judge your organisation, and whether automation is actually saving you money or quietly driving customers away. A few things are worth checking directly.

Start with what you actually measure. Many reports show how many conversations the chatbot handled without transferring, the so called deflection rate. That says nothing about whether the problem was actually solved. Instead, track how often a customer calls back with the same question after a chatbot contact. Then check whether transferring to a human is possible within one or two steps, not only after working through a long menu.

Also assess whether your chatbot gives answers that go beyond what is already on your website. If customers checked the website first and then get the same answer from the bot, you have not added anything. And keep your phone line properly staffed: this study shows it remains the channel customers trust most, especially right after a chat contact has failed them.

A customer contact setup where telephony, chat and your knowledge base all run separately makes these kinds of improvements hard to implement. Zarioh helps organisations set up telephony and customer contact solutions where automation is built from the organisation's own knowledge, with a direct and simple route to a human agent whenever that is needed. Feel free to reach out for a no-obligation conversation about your current setup.

Z

Zarioh Digital Solutions

IT specialists from Utrecht, the Netherlands. We help businesses with Microsoft 365, AI agents, hosting and telephony, and share what we learn in practice. Follow us on LinkedIn

Related articles

← Back to all articles
Share