Retail’s newest abandonment reason wasn’t meant to exist.
Every year at Humii, we analyse thousands of real online shopping journeys to answer one simple question: Why do shoppers leave?
Some reasons remain relatively consistent. Shipping and delivery information continues to top the list (18.3 per cent), followed by customer support (17.9 per cent) and site navigation (13.9 per cent).
This year, however, something entirely new emerged. For the first time, we introduced a brand-new abandonment category into our 2026 Online CX Report: Chatbot obstruction.
At the beginning of the year, chatbot obstruction already accounted for 3.4 per cent of all abandonment reasons before checkout, making it the eighth-largest reason shoppers abandoned a purchase.
That alone surprised us. But what surprised us even more is that six months later, it sits closer to 7 per cent.
For some retailers, chatbot obstruction is as high as 35 per cent. In those cases, the retailers may genuinely have been better off without a chatbot at all. In other words, one of the fastest-growing reasons shoppers are abandoning websites is a tool that was introduced to stop them from leaving. That’s quite the achievement.
Chatbots were meant to remove friction. Instead, many are creating it.
The original promise was compelling: Instant answers, no waiting for customer service, quick resolutions, more completed purchases. Yet increasingly, we’re seeing shoppers become obstructed by the very tool that was designed to help them.
That is not because chatbot technology is not improving; quite the opposite. The problem is that many retailers appear to launch them… and then stop looking after them.
Frankly, the number of chatbots proudly claiming “Ask me anything” while struggling to answer simple, everyday customer questions is astonishing. Many chatbots provide outdated information, many do not understand basic spelling mistakes, many do not know about current promotions.
And many simply send shoppers around in circles.
The examples are becoming hard to ignore
Recently, we asked Temple & Webster’s chatbot about an Aurora Lamp. It confidently replied that the company did not sell one. The irony? – The lamp had been promoted in one of its own EDMs earlier that day, and the chatbot even linked us… to the very lamp it had just claimed didn’t exist.
At BWS, we asked for a prosecco with the alcohol removed. The response? “Have a nice day.”
Polite. But not particularly useful.
The Iconic’s chatbot could not recognise the word “retrun” when asking about a return label. One simple typo and the conversation was over. Humans make spelling mistakes every single day. Your chatbot should expect them.
Myer’s Concierge was unaware of its current 40 per cent-off promotion. If your chatbot does not know today’s campaign, shoppers quickly begin wondering what else it does not know.
We asked Big W’s chatbot: “Can you send me a landing page link to kids pyjamas?” Its response? “Cya later!” Indeed, see ya later. Perhaps not anytime soon.
Customers don’t hate chatbots. They hate unreliable chatbots
One of the biggest behavioural shifts we’re now observing is shoppers actively avoiding chatbots altogether.
After enough disappointing experiences across different retailers, many shoppers simply won’t engage unless they know there’s a human behind the conversation. Ironically, this completely defeats the original purpose of introducing chatbots: They were meant to reduce pressure on customer service.
Instead, they’re often driving customers back towards it.
But the problem isn’t AI. It is ownership. Which raises much bigger questions: Who actually owns the chatbot? Who checks whether it knows about today’s promotions? Who updates it when new products launch? Who tests it after an EDM goes out? Who deliberately asks difficult questions? Who tries spelling mistakes? Who asks something completely unexpected?
Because from what we’re seeing, many chatbots feel like they’ve been launched… and then quietly forgotten.
Retailers continuously optimise their websites, they A/B test landing pages, they monitor checkouts, they review search functions. Yet one of the most customer-facing touchpoints on the website often appears to receive the least ongoing attention.
A few simple rules every retailer should follow
- Be clear who’s behind the conversation. If it is AI, say it is AI. If it is a human, say so. Transparency builds trust.
- Always provide a human escalation path. Too many chatbots send shoppers around in circles. If the bot cannot solve the problem, it should know when to hand over.
- Set expectations. Don’t say “Ask me anything” if your chatbot clearly can’t. The best bots explain exactly what they can and cannot help with.
- Stop asking for an email address before helping. This remains one of the most frustrating experiences our mystery shoppers report. Customers came looking for an answer, not another form to complete.
- Expect human behaviour. People make typos, they ask vague questions and they do not always use perfect terminology. Your chatbot needs to cope with that.
- Be transparent about when your chatbot or support team is available. A simple message such as, “We’re here to help from 8am–5pm. If you contact us outside these hours, we’ll get back to you the next business day,” sets clear expectations and avoids unnecessary frustration.
- Stress-test your chatbot. Ask awkward questions, search for products from your latest EDM. Ask about today’s promotion. Try spelling mistakes. Challenge it.
Trust us. We’ve stress-tested hundreds of them. It gets wild remarkably quickly.
Chatbots aren’t the problem. Neglect is
This is not an argument against chatbots. The best ones genuinely remove friction, improve conversion and make life easier for customers. But they require ownership, maintenance, continuous testing, and ongoing validation.
That is because the moment your chatbot becomes an obstruction rather than an assistant, it is no longer helping shoppers. It’s helping them leave.
About the author: Mareile Osthus is the founder and CEO of Humii, and a former CCO of The Iconic.
