AI shopping is forcing marketers to rethink which digital channels matter most for brand visibility.

For years, marketers have concentrated resources on two questions: Where do consumers spend their time, and where do they spend their money? AI product recommendations add a third: Where do large language models (LLMs) get the information that shapes what they recommend?

“You also need to consider where these LLMs are sourcing recommendations,” said our analyst Blake Droesch during EMARKETER's Future of Digital summit.

That distinction matters because the sources influencing AI recommendations don’t necessarily match the platforms commanding the most consumer attention, ecommerce sales, or ad dollars.

Different LLMs favor different sources

EMARKETER’s AI Visibility Index tracks how brands appear in AI product recommendations across 14 industries. EMARKETER created the index by feeding LLMs thousands of shopping-related prompts and tracking which brands appeared most frequently.

Brands cannot assume the same visibility strategy will work across AI platforms, the data shows.

“Each LLM has a unique method of sourcing product recommendations,” Droesch said.

Both ChatGPT and Gemini lean on brand websites when recommending food and beverage products, but their secondary sources diverge: ChatGPT draws from a broader mix of retailers and publishers, while Gemini favors publishers.

ChatGPT's August citations for food and beverage show that mix at work: Target, Walmart, and Whole Foods Market generated the greatest share, and Consumer Reports and the FDA rounded out the top five. Gemini's leading sources skewed toward industry-specific publishers.

The pattern holds beyond food and beverage. Across the consumer industries EMARKETER tracks, ChatGPT tends to draw on a more varied mix of sources, while Gemini favors publishers.

For marketers, that makes their own websites the starting point, not the entire strategy.

“Your best asset is obviously your website,” Droesch said. “But it is really important that you are designing your website so it's allowing AI agents to be able to get the information that it needs to recommend your product.”

Brands that want ChatGPT visibility should build a presence across relevant retailers and publishers. For Gemini, relationships with category-specific publishers may carry more weight.

AI citation patterns aren’t static

Marketers also need to distinguish between a source that suddenly gains influence and one that consistently shapes AI recommendations.

“LLMs typically cite the same sources, but the citation frequency can be a lot more volatile,” Droesch said.

In personal care and beauty, for example, the American Academy of Dermatology was ChatGPT’s most frequently cited source in August. But across a six-month average dating back to March, Ulta Beauty and Allure emerged as more influential nonbrand sources.

The mix can shift by source type, too. Publishers accounted for about one-quarter of ChatGPT citations in personal care and beauty during May and June. In other months, ChatGPT relied more heavily on brands’ own websites.

That volatility creates two jobs for marketers: Watch month-to-month changes for sudden shifts, including those that may accompany model updates, and identify the sources that maintain influence over longer periods.

“When you're building a long-term strategy, you really want to invest in the sources that are maintaining influence over time,” Droesch said.

AI creates a third bucket for channel strategy

The biggest shift is that AI's most influential sources often sit outside marketers' largest digital investments.

The overlap was limited. When EMARKETER compared major digital platforms by time spent and retailers by ecommerce sales with the domains driving more than 3% of product recommendations on ChatGPT or Gemini, YouTube, Meta's platforms, and Amazon were notably absent.

“These are the channels that command the lion's share of digital ad spending,” Droesch said. “They're also the channels that marketers tend to dedicate the most time and resources.”

That doesn’t make those platforms less important. It means marketers have one more factor to weigh.

Over the past decade, the consolidation of attention and ecommerce around large technology platforms has given brands good reason to prioritize social media visibility and Amazon’s digital shelf. AI shopping could revive the importance of assets and relationships that received less attention along the way, including brands’ own websites and smaller industry publishers.

“There's been a tendency for marketers, rightfully so, to focus more on how does my brand show up on social media? More about how do I optimize for the digital shelf on Amazon?” Droesch said. “But we're now entering this new phase where brands have a renewed reason to consider the power of my brand's website.”

What this means for marketers: Generative engine optimization (GEO) shouldn’t replace existing channel investments. Marketers should treat AI visibility as a third layer alongside attention and commerce, then determine which websites, retailers, publishers, and other sources consistently influence recommendations within their category.

“AI is not replacing those two buckets,” Droesch said. “It's really just adding a third layer of complexity that marketers now need to factor into the equation.”

Watch the full session.