Amazon, Target, Nordstrom and Costco dominate AI recommendations on delivery, returns and support, while challenger brands make inroads in gift categories.
AI assistants largely agree on which retailers to recommend for holiday shopping, but their gift suggestions vary widely, according to a new study from BrandRank.AI. The findings also point to customer service policies and reputation as factors shaping which retailers appear in AI-generated recommendations.
The BrandRank 2026 Holiday AI Recommendation Share Index examined 950 responses from ChatGPT, Claude, Gemini, Grok, Perplexity and Meta AI. The questions covered gift ideas, retailers, delivery, returns, customer service and trust.
Four or more engines named the same retailer in 65% of retailer questions. For product questions, the same level of agreement occurred in 31%. The Amazon Fire HD Kids tablet was the only product to receive unanimous recommendations.
Service separates retailer recommendations
Amazon, Nordstrom, Target and Costco accounted for 80% of responses to questions about delivery, checkout, pickup, returns, complaints and trust. Across the full study, the four retailers represented 32% of recommendations.
The engines tended to associate Amazon with delivery and checkout, while Nordstrom appeared more often in answers about returns and resolving problems. On eight core customer service questions, Nordstrom received 25 of 48 first-place recommendations; Amazon received one.
Pete Blackshaw, co-founder and CEO of BrandRank.AI and author of The Answer Economy, said:
Pete Blackshaw, co-founder and CEO of BrandRank.AI
“I’ve spent much of my career listening to customer complaints. Now AI is reading return policies and service stories before deciding who deserves the recommendation. Amazon wins the service that prevents problems. Nordstrom wins the service that fixes them. That’s a wake-up call for every brand.”
BrandRank found that “return” appeared 28 times in 72 customer service rationales it examined. References to AI, chat or automation did not appear in those rationales, despite retailers’ investment in AI-powered support.
Challenger brands appear in gift recommendations
The study found strong agreement around several brands outside the major retail groups. All six engines recommended Quince for affordable cashmere and Bellroy for men’s wallets. Five of the six named Aura digital frames for grandparents and Furbo cameras for pet lovers. Claude was the only engine to name Pop Mart’s Labubu as the toy of the season.
By contrast, gift questions about mothers, fathers and coworkers, as well as gifts under $25 and luxury gifts over $250, drew six different answers from six engines.
The study’s findings suggest that recommendations can vary by platform. Gemini named Amazon 26 times and did not name Walmart, while Perplexity recommended Nordstrom more often than Amazon.
Blackshaw added:
“Every AI engine sees the market differently. A brand that wins in one can be nearly invisible in another. As AI agents increasingly influence purchases, brands need to understand not just whether they’re recommended, but why they’re trusted, where they’re missing and what happens when consumers need help.”
BrandRank also flagged 33 responses involving outdated products, questionable pricing or recommendations that did not fit the question. Four engines continued to refer to Amazon’s shopping assistant as Rufus, which the study says had been retired earlier in 2026.
Implications for customer service
The results highlight a growing challenge: AI recommendations may depend not only on the service a company provides, but on whether that service is clearly documented. As AI becomes another discovery and decision-making channel, the clarity and credibility of a company’s service information could influence which brands customers consider before they ever visit a website or contact support.
BrandRank said it plans to revisit the service questions during the January returns season to assess whether holiday shopping and service experiences change retailer recommendation patterns.
Research details
The October 2026 study used a consistent protocol across six AI engines and assessed 950 generated responses and their supporting rationales. It compared brand visibility, recommendation patterns, cross-engine agreement and potential accuracy issues.
The findings are a snapshot of AI recommendations, not independent product tests or a measure of actual retailer performance. The questions, responses and research details are available in the full report.
