Shoppers Want AI to Help Them Shop, Not to Fix Their Problems or Fake Their Photos: Baymard

Date:

Share post:

US online shoppers are happy to let AI help them decide what to buy, but they draw a firm line when an order goes wrong or a product photo turns out to be AI-generated, according to new research from Baymard, the independent ecommerce UX research organization.

Baymard’s AI Engagement study, a survey of 2,273 US online shoppers, found most shoppers prefer an AI chatbot for low-stakes research tasks:

  • Comparing product features (63%)
  • Checking stock availability (59%)
  • Getting personalized recommendations (56%)

That preference is inverted when an order goes wrong, said the report. Roughly two-thirds prefer a human representative for:

  • Damaged or incorrect items (69%)
  • Late or missing orders (67%)
  • Shipping or billing mistakes (66%)

“Shoppers have made the division of labor very clear,” said Christian Holst, co-founder of Baymard. “They’re happy for AI to do the legwork of comparing products and checking stock. But the moment an order goes wrong, they want a person who can actually fix it. Retailers that trap customers in a chatbot loop during the busiest weeks of the year are putting their most valuable relationships at risk.”

60% of shoppers say knowing a product image is AI-generated would reduce their confidence in buying. 25% say it would slightly reduce their confidence and 18% say it would majorly reduce it. Around one in six (17%) say they would not feel confident enough to buy the product at all. Only 18% say it would make no difference, said Baymard.

“Product images are how shoppers check what they’re actually getting,” said Holst. “Once they know an image is AI-generated, that check breaks down, and for one in six shoppers, so does the sale. AI can help with things like lifestyle backgrounds, but the core product photos need to be real.”

What retailers can do

Baymard recommends that retailers:

  • Place AI assistants where shoppers want them, such as in product discovery, comparison, and stock checks.
  • Keep a clear, visible route to a human agent for order problems. Where live support isn’t available, offer transparent self-service options.
  • Use authentic photography for core product views, and take active steps to rebuild trust wherever AI-generated visuals are used.

In an interview with Retail Insider, Holst discussed the findings of the survey.

Your research found that 63 per cent of U.S. online shoppers prefer AI chatbots for comparing product features, but 69 per cent prefer human assistance when dealing with damaged or incorrect orders. What does this divide tell us about where consumers are comfortable with AI in the shopping journey, and where retailers risk damaging customer relationships by relying too heavily on automation?

We found that shoppers are broadly okay with AI assistance for gathering information (research, product suggestions), but not for decision-making matters or issue resolution (purchases, order issues, returns).

This tells us that the need for human support isn’t going away, but it may be shifting in the customer journey. The data shows that the majority of users aren’t yet comfortable giving AI control over making purchases for them and that post-purchase support is where human assistance is still critical.

Users need clarity, flexibility, and reassurance throughout the return process. If the site makes users hunt for help, re-explain the problem across channels, or guess what happens next, support burden rises fast and trust drops. When something breaks, users want conversational reassurance from a human.

Canadian retailers are increasingly introducing AI-powered shopping assistants and customer service tools. Although your research surveyed U.S. consumers, do you believe these findings reflect broader North American consumer expectations? What should Canadian retailers consider when implementing AI across their ecommerce platforms?

While this survey only looked at US consumers, it’s showing patterns related to human behavior. It shouldn’t be a surprise that consumers prefer human assistance for issue resolution. There’s a greater reassurance that a human will be able to understand a problem and empathize, and this reassurance is necessary regardless of geography.

What we’ve seen with our UX research is that core usability problems remain remarkably consistent across geographies.

As for implementing AI across ecommerce platforms, it’s important to remember that currently only a subgroup of users actively seek out and engage with AI features. Retailers shouldn’t expect that a majority of users will interact with their AI features. It’s still critical to ensure their sites are performing well across all traditional areas of ecommerce UX (traditional search, category navigation, product page, etc.) and not assume most of their users will rely on any AI features to find products.

MART PRODUCTION photo
MART PRODUCTION photo

The research found that 60 per cent of shoppers would have less confidence in a product if they knew its image was AI-generated, while 17 per cent said they would not feel confident enough to purchase it. Why are consumers responding this way, and what does this suggest about the importance of authentic product photography in online retail?

Product images are one of the most important inputs online shoppers use to decide whether a product meets their needs. When people shop online, they can’t pick up the product, turn it around, or move closer to inspect it. So images become the primary substitute for physical evaluation. Baymard has extensive research on product images, ranging from how many images are necessary, to what to show in images, to image quality. While images are important for all products, we’ve found that they’re especially critical for visually-driven industries, like apparel, furniture, home decor, or jewelry where shoppers are often judging style, fit, texture, finish, and real-world appearance.

When authentic product photography is replaced with AI-generated images, there isn’t the same trust that the product is being accurately represented. Without trust in the product images, users don’t have the ability to properly evaluate the product.

With Black Friday and the holiday shopping season approaching, retailers are under pressure to improve efficiency, manage higher customer service volumes and adopt new technologies. How should retailers balance the cost and operational benefits of AI with consumers’ expectations for human assistance, particularly when dealing with complaints, returns, shipping problems and other sensitive transactions?

Our testing indicates AI implementation varies widely on ecommerce sites. While on some ecommerce sites users were able to successfully use AI as part of their product finding and exploration, on others the AI feature led to some feature or site abandonments due to its poor implementation.

As mentioned above, the majority of users won’t seek out and use AI features, and when it comes to resolving order issues human assistance is still preferred.

Our testing indicates that AI features can be a good “first point of contact” for users dealing with complaints, returns, shipping problems, and sensitive issues. However AI features are prone to hallucination and providing incorrect answers, and therefore users will expect to be able to reach a human if necessary to resolve complex ecommerce issues.

Looking ahead, how do you see consumer expectations around AI in ecommerce evolving over the next two to three years? What are the biggest mistakes retailers could make as they expand their use of AI, and what practical steps should Canadian retailers take now to build consumer trust while making the most of the technology?

Consumers are likely to expect on-site AI features to improve over the next 3 years. While currently many AI features do a reasonably good job assisting with product finding and exploration, our testing indicates users’ experiences are inconsistent. Thus, poor-performing AI features are likely to stand out in several years as indicative of a site that’s not “up to par”.

One of the biggest mistakes is thinking all users will use AI features as part of their online shopping. Our quantitative testing indicates that while there is a subgroup of users willing to use AI as part of their ecommerce shopping, the vast majority still rely on traditional site navigation and exploration. Therefore spending all your resources on having a top-performing AI feature may not represent the best investment in your site, as currently most users won’t rely on AI to help find and purchase products.

A practical step retailers can take is to always make it clear how to access the site’s AI features and be transparent when AI is assisting with product finding. Additionally, noting that AI features can make mistakes will help reduce users’ expectations if the feature fails to perform as they expect. This helps protect trust in the brand. Finally, focusing on traditional product finding and exploration, and checkout UX is a smart way to ensure you’re meeting the needs of the majority of users who will prefer manual exploration of the site.

More from Retail Insider:

Mario Toneguzzi
Mario Toneguzzi
Mario Toneguzzi, based in Calgary, has more than 40 years experience as a daily newspaper writer, columnist, and editor. He worked for 35 years at the Calgary Herald covering sports, crime, politics, health, faith, city and breaking news, and business. He is the Co-Editor-in-Chief with Retail Insider in addition to working as a freelance writer and consultant in communications and media relations/training. Mario was named as a RETHINK Retail Top Retail Expert in 2024.

MORE FROM AUTHOR

Subscribe to the Newsletter

Subscribe

* indicates required

Related articles