Retailers are rushing to build AI apps. It’s unclear if shoppers will use them

AI Strategies Article Note
Retailers and brands are quickly launching shopping apps within ChatGPT and Claude. But whether or not shoppers will embrace them at a rapid pace remains to be seen.
In October, OpenAI launched ChatGPT apps, where users can connect apps like Expedia, Spotify or Zillow. Once adding the apps, customers can ask the chatbot questions about planning a trip or crafting a playlist, for example. They’ll then get information about such products and services from those brands. This month, Claude introduced its version of apps called “connectors” for consumer use, after previously supporting work-related apps such as Microsoft Teams, Outlook and OneDrive. It now can tap into Uber, TripAdvisor and Instacart, among others.
There are almost 900 apps on ChatGPT and 353 Claude connectors, according to AppDiscoverability.com, which tracks AI app data. About 10% of the ChatGPT apps listed are for shopping. Many of the largest retailers have already launched ChatGPT apps for shopping, including Target and Walmart. On ChatGPT specifically, more than 500 of these apps have been added just this month. However, there’s limited information about how these apps are performing and who is using them — and, most crucially for retailers, whether they can lead to purchases.
“For the moment, to be honest, adoption and conversion are pretty low,” said Dimitri Ewald, chief of staff at Alpic — a French company that builds and distributes apps for ChatGPT and Claude — based on conversations with his clients. “People don’t even know that there are apps in the ChatGPT store.”
OpenAI and Anthropic representatives did not immediately respond to requests for comment.
This flurry of new apps follows the discontinuation of ChatGPT’s Instant Checkout, which allowed users to ask ChatGPT what they’re looking for, receive recommendations of relevant products and check out with retailers like Walmart without having to connect a separate app. Several people involved with Instant Checkout told Modern Retail and other outlets that the program didn’t drive sales and that some merchants didn’t want OpenAI overseeing the checkout process.
ChatGPT apps can vary widely in how they function, what kinds of questions they can answer, how rich of a feature set they have and how integrated the checkout functionality is. Some just provide links to external websites; others allow you to add products to a cart and check out on a merchant’s website. Instacart lets you check out directly within ChatGPT, others do not.
Just a few examples of how these apps work: Last month, Sephora announced its ChatGPT app for beauty advice and recommendations that allows customers to use their loyalty rewards and member benefits, such as free shipping promotions and samples. Starbucks launched a beta ChatGPT app on April 15, where people can ask it for specific types of drinks, like one high in protein, nutty or sweet; or for a drink that matches their outfit. People can then customize the drink, choose the pickup location and finish checking out in the Starbucks app.
“I don’t think there’s anything wrong with a retailer or brand doing an app, and I think there are benefits to be had; there are learnings to be had, and data to be collected, and muscle memory to be built by brands and retailers doing it,” said Jason Goldberg, chief commerce strategy officer for Publicis Groupe. “I wouldn’t discourage someone from doing it. I don’t think it’s economically very exciting or material. In the long run, unless it evolves dramatically, I don’t think it’s going to be an important part of the agentic commerce ecosystem.”
One of the biggest flaws with ChatGPT apps is that they don’t have natural discoverability. The Sephora app won’t launch simply when you ask ChatGPT for cosmetics advice. Rather, you would have to know the retailer’s app exists, find it on the ChatGPT app store, install it and initiate it before getting an answer from Sephora. Unlike ChatGPT, Claude doesn’t require you to hit a plus icon next to the text box to tell it you want to use an app. Claude decides based on your prompt which apps can service it. However, it still requires you to add the app first before it recommends it.
“I would not say, ‘Apps are never going to work, and they’re dead,'” Goldberg said. “What I would say is, if they do work and they’re important to shoppers, they’ll probably look pretty different than the ones we have today.”
If, in the current moment, there is little evidence of the apps being successful as a commerce tool, why are retailers and brands launching them? Some say it’s FOMO, or the fear of missing out. Ewald, the AI app developer, said retailers who see all their competitors in ChatGPT want to be there, too. Additionally, he said some brands are hopeful that the discovery experience for ChatGPT apps will eventually improve and that it’s at least a way to control first-party information through ChatGPT.
Still, some retailers, brands and other app developers are hesitant to hand off customer information and payments to OpenAI, according to Bloomberg, which also reported that developers have complained about a tedious app approval process, a buggy coding system and the lack of usage data.
“We’re still early in building this out, and we recognize there are areas where the developer experience needs to improve,” a spokesperson told Bloomberg, adding that the company wants to make the platform “more reliable, more predictable and easier to build on over time.”
Goldberg said he often talks to retail executives who want an agentic commerce strategy, but most of the brands he works with aren’t immediately prioritizing apps as the way to tap into this new ecosystem. Instead, he said they’re more interested in influencing the native ChatGPT or Claude chatbots through generative engine optimization, figuring out what the systems are citing. They may also be interested in investing in AI features on their website or mobile app.
“While there’s a ton of traffic going into the LLMs, none of that traffic is available to the app unless you explicitly build your own brand recognition for your app,” Goldberg said. “If you’re going to do that, why would you build it for an app on someone else’s platform that you don’t own and that could take it away from you whenever they want? You’d build it for a platform that you own.”
It’s possible ChatGPT and Claude could someday add features that dynamically suggest apps without installing them first, said Juozas Kaziukėnas, an independent e-commerce analyst. “There are ways that these apps could be more user-friendly, but then, the question still remains is if using those apps, especially for consumer use cases, is an improvement on the AI experience,” he said.
Still, the apps can have a benefit in being a low-cost focus group, Goldberg argues. You can put analytics tags into the app and see what paths customers are or aren’t taking, and see the conversions in each part of the customer journey.
“Those learnings aren’t useful just to the app; they’re useful to the whole agentic commerce ecosystem,” Goldberg said. “You could think of launching a retail app on ChatGPT as a low-expense focus group that you can then use to improve the agents you might build for your own vertical website.”
Kaziukėnas compared the AI app launches to the metaverse and NFT launches several years back, where some companies are launching these because the technology is popular rather than having a specific use case in mind.
“There’s no visible success story of any of these apps being a meaningful driver for any of these companies,” Kaziukėnas said. “We can theorize how useful these apps could be or can’t be, but I think, for now, it’s quite clear they’re just experiments meant to drive press releases and don’t really impact anything at all.”