AI doesn’t care about your beautiful website.

Visit any fashion brand’s homepage and you’ll see all sorts of dynamic or interactive elements from image carousels to dropdown menus that are designed to catch shoppers’ eyes and ease navigation.

To the large language models that underlie ChatGPT and other generative AI, many of these features might as well not exist. They’re often written in the programming language JavaScript, which for the moment at least most AI struggles to read.

This giant blindspot didn’t matter when generative AI was mostly used to write emails and cheat on homework. But a growing number of startups and tech giants are deploying this technology to help users shop — or even make the purchase themselves.

“A lot of your site might actually be invisible to an LLM from the jump,” said A.J. Ghergich, global vice president of Botify, an AI optimisation company that helps brands from Christian Louboutin to Levi’s make sure their products are visible to and shoppable by AI.

The vast majority of visitors to brands’ websites are still human, but that’s changing fast. US retailers saw a 1,200 percent jump in visits from generative AI sources between July 2024 and February 2025, according to Adobe Analytics. Salesforce predicts AI platforms and AI agents will drive $260 billion in global online sales this holiday season.

Those agents, launched by AI players such as OpenAI and Perplexity, are capable of performing tasks on their own, including navigating to a retailer’s site, adding an item to cart and completing the checkout process on behalf of a shopper. Google’s recently introduced agent will automatically buy a product when it drops to a price the user sets.

This form of shopping is very much in its infancy; the AI shopping agents available still tend to be clumsy. Long term, however, many technologists envision a future where much of the activity online is driven by AI, whether that’s consumers discovering products or agents completing transactions.

To prepare, businesses from retail behemoth Walmart to luxury fashion labels are reconsidering everything from how they design their websites to how they handle payments and advertise online as they try to catch the eye of AI and not just humans.

“It’s in every single conversation I’m having right now,” said Caila Schwartz, director of consumer insights and strategy at Salesforce, which powers the e-commerce of a number of retailers, during a roundtable for press in June. “It is what everyone wants to talk about, and everyone’s trying to figure out and ask [about] and understand and build for.”

From SEO to GEO and AEO

As AI joins humans in shopping online, businesses are pivoting from SEO — search engine optimisation, or ensuring products show up at the top of a Google query — to generative engine optimisation (GEO) or answer engine optimisation (AEO), where catching the attention of an AI responding to a user’s request is the goal.

That’s easier said than done, particularly since it’s not always clear even to the AI companies themselves how their tools rank products, as Perplexity’s chief executive, Aravind Srinivas, admitted to Fortune last year. AI platforms ingest vast amounts of data from across the internet to produce their results.

Though there are indications of what attracts their notice. Products with rich, well-structured content attached tend to have an advantage, as do those that are the frequent subject of conversation and reviews online.

“Brands might want to invest more in developing robust customer-review programmes and using influencer marketing — even at the micro-influencer level — to generate more content and discussion that will then be picked up by the LLMs,” said Sky Canaves, a principal analyst at Emarketer focusing on fashion, beauty and luxury.

Ghergich pointed out that brands should be diligent with their product feeds into programmes such as Google’s Merchant Center, where retailers upload product data to ensure their items appear in Google’s search and shopping results. These types of feeds are full of structured data including product names and descriptions meant to be picked up by machines so they can direct shoppers to the right items. One example from Google reads: Stride & Conquer: Original Google Men’s Blue & Orange Power Shoes (Size 8).

Ghergich said AI will often read this data before other sources such as the HTML on a brand’s website. These feeds can also be vital for making sure the AI is pulling pricing data that’s up to date, or as close as possible.

As more consumers turn to AI and agents, however, it could change the very nature of online marketing, a scenario that would shake even Google’s advertising empire. Tactics that work on humans, like promoted posts with flashy visuals, could be ineffective for catching AI’s notice. It would force a redistribution of how retailers spend their ad budgets.

Emarketer forecasts that spending on traditional search ads in the US will see slower growth in the years ahead, while a larger share of ad budgets will go towards AI search. OpenAI, whose CEO, Sam Altman, has voiced his distaste for ads in the past, has also acknowledged exploring ads on its platform as it looks for new revenue streams.

“The big challenge for brands with advertising is then how to show up in front of consumers when traditional ad formats are being circumvented by AI agents, when consumers are not looking at advertisements because agents are playing a bigger role,” said Canaves.

Bots Are Good Now

Retailers face another set of issues if consumers start turning to agents to handle purchases. On the one hand, agents could be great for reducing the friction that often causes consumers to abandon their carts. Rather than going through the checkout process themselves and stumbling over any annoyances, they just tell the agent to do it and off it goes.

But most websites aren’t designed for bots to make purchases — exactly the opposite, in fact. Bad actors have historically used bots to snatch up products from sneakers to concert tickets before other shoppers can buy them, frequently to flip them for a profit. For many retailers, they’re a nuisance.

“A lot of time and effort has been spent to keep machines out,” said Rubail Birwadker, senior vice president and global head of growth at Visa.

If a site has reason to believe a bot is behind a transaction — say it completes forms too fast — it could block it. The retailer doesn’t make the sale, and the customer is left with a frustrating experience.

Payment players are working to create methods that will allow verified agents to check out on behalf of a consumer without compromising security. In April, Visa launched a programme focused on enabling AI-driven shopping called Intelligent Commerce. It uses a mix of credential verification (similar to setting up Apple Pay) and biometrics to ensure shoppers are able to checkout while preventing opportunities for fraud.

“We are going out and working with these providers to say, ‘Hey, we would like to … make it easy for you to know what’s a good, white-list bot versus a non-whitelist bot,’” Birwadker said.

Of course the bot has to make it to checkout. AI agents can stumble over other common elements in webpages, like login fields. It may be some time before all those issues are resolved and they can seamlessly complete any purchase.

Consumers have to get on board as well. So far, few appear to be rushing to use agents for their shopping, though that could change. In March, Salesforce published the results of a global survey that polled different age groups on their interest in various use cases for AI agents. Interest in using agents to buy products rose with each subsequent generation, with 63 percent of Gen-Z respondents saying they were interested.

Canaves of Emarketer pointed out that younger generations are already using AI regularly for school and work. Shopping with AI may not be their first impulse, but because the behaviour is already ingrained in their daily lives in other ways, it’s spilling over into how they find and buy products.

More consumers are starting their shopping journeys on AI platforms, too, and Schwartz of Salesforce noted that over time this could shape their expectations of the internet more broadly, the way Google and Amazon did.

“It just feels inevitable that we are going to see a much more consistent amount of commerce transactions originate and, ultimately, natively happen on these AI agentic platforms,” said Birwadker.

IBL News | New York

Sixty-eight organizations have signed to date the White House’s Pledge to America’s Youth: Investing in AI Education over the next four years, which follows President Trump’s April 23 executive order in this regard.

Some companies signing the pledge include Google, Amazon, Apple, IBM, Pearson, NVIDIA, OpenAI, Microsoft, Oracle, Adobe, Cisco, Dell, Intel, McGraw-Hill, Workday, Booz Allen, and Magic School AI.

These organizations pledge “to make available, over the next four years, resources for youth and teachers through funding and grants, educational materials and curricula, technology and tools, teacher professional development programs, workforce development resources, and/or technical expertise and mentorship,” working alongside the White House Task Force on Artificial Intelligence Education.

“The Pledge will help make AI education accessible to K-12 students across the country, sparking curiosity in the technology and preparing the next generation for an AI-enabled economy. Fostering young people’s interest and expertise in artificial intelligence is crucial to maintaining American technological dominance,” added.

Michael Kratsios, Director of the White House Office of Science and Technology Policy and Chair of the White House Task Force on AI Education, invited other organizations to join the pledge.

“AI is reshaping our economy and the way we live and work, and we must ensure the next generation of American workers is equipped with the skills they need to lead in this new era,” said Secretary of Labor Lori Chavez-DeRemer. 

Brian Stone, performing the duties of the National Science Foundation (NSF) director, said that his institution will fund cutting-edge research, support teacher development, and expand access to STEM education.

As of June 30, 2025, these were the organizations supporting the Pledge:

“There’s a shift in the customer behavior, which is playing a big part, in terms of the overall adoption of these tools,” Kundu said.

Kendra Scott is one of many brands using AI-powered tools to soup up the online customer experience. About 57% of retail leaders told EMarketer that chat-based customer support is the area of retail that will be most heavily influenced by AI through 2026. While these services were once deemed “clunky” and could turn customers away, they’re becoming a more popular must-have as technology improves.

For its part, Kendra Scott is two years into a three-year digital strategy transformation that’s so far delivered 50% sales growth. Kundu said the strategy relies on three main pillars: a better mobile experience, AI-powered personalization and what he calls “experiential unified commerce,” ensuring brand and experience consistency across channels.

In March, Kendra Scott launched a standalone website for its Yellow Rose brand, and it will unveil a full redesign for both brands’ sites in the third quarter of 2025. They’ll be powered with a brand-new tech stack, Kundu said, to make sure that they’re able to load faster. In general, a 10% improvement in site speed translates to a 1% uptick in conversion, Kundu said. So far, the Yellow Rose site has seen a 15-20% speed improvement.

“We wanted the digital experience to carry the warmth and emotional connection that our stores are known for. And that meant migrating to a more modern, progressive web app architecture and a more composable tech stack,” he said.

But building up a site in this way requires attention to detail — in particular, how users are already interacting with the site, Kundu said. That included examining entry traffic and checkout starts, for example, versus checkout completion.

“We have a very clear understanding of the overall customer funnel. When we see a dropoff in a certain section on the website, we become more curious, in terms of what may be the driver for that. And then the team will line up different tests and learn about those friction points in the journey,” he said.

But some of the biggest impact is coming from AI-powered tools. Beyond AI Copilot, Kendra Scott is using predictive AI for marketing. Kundu said it uses a third-party tool to read through over 500 customer behavioral signals that feed into personalized marketing messages. This can result in more personalized calls to action and category banners on product listing pages, and add more trending pieces to product display pages. So far, the company has seen a 5% RPV lift from aligning those changes.

For Kundo, these changes come in concert with stronger imagery and brand storytelling, and adjustments to the supply chain that ensure the company can keep up with demand.

“All these AI applications allow us to deliver relevance at scale — not just to improve conversion, but also to build trust, emotional connection and long-term loyalty by showing up in the right moment with the right tone and a clear understanding of who the customer is,” Kundu said.

Tim Glomb, vp of digital, content and AI at marketing platform Wunderkind, said brands today are charged with making meatier e-commerce experiences because of both customers’ expectations and the way they discover sites. More expansive product descriptions, for instance, can help brands stand out in the databases used by AI engines like ChatGPT. “SEO is breaking down, in the traditional sense,” he said. “It’s not just about traffic. It’s about having the right metadata and product descriptions that answer real customer questions.”

From a consumer perspective, Glomb said Kendra Scott is poised to keep growing if it can help send shoppers in the right direction. He recently headed to the brand’s site when shopping for his teenage daughter’s birthday. He knew the brand was popular with her age group after seeing a brand activation at a volleyball tournament, but he didn’t know what styles to look at or which might be a good fit.

The right-rail AI Copilot will ask what kind of finish he’s looking for, whether sterling silver or 18-karat gold vermeil. It will also dig into sizing and budget.

“What do I need, based on what I know? I have a 14-year-old who’s tall and wants jewelry,” he said. “For someone like me who’s not fluent in jewelry or fashion, the AI bot knows the filters and features [that can help].”

Glomb anticipates that prediction, personalization and recommendation tools will get even better at knowing who a customer is and what they may want as soon as they land on a site. “It would be the same way you’d walk into a store and the counter person says, ‘I could show you 500,000 things in this case, but here are three things based on our conversation,’” he said.