AI Shopping Assistance: Is It Enabling User Convenience Or Cementing Brand Bias?

Hyderabad: After influencing the world in almost every imaginable way, from helping students write essays to assisting military operations, artificial intelligence now wants to impact how you shop. Tools like ChatGPT’s Shopping Research have been in existence for months, and now Google is enhancing its Gemini-powered shopping assistance feature by combining it with the Shopping Graph.

With the same end goal as ChatGPT, the new feature moves beyond simple text recommendations to actual product listings, enabling users to find specific items and compare them without leaving AI’s stronghold, which spans across the Gemini app, AI Mode in Search, and Circle to Search.

Google says that these tools have been designed to help users in India shop with “more confidence and less effort”. But who decides which products make it to AI's recommendation engine? The AI-powered shopping experience raises some important questions that, if left unanswered, could give birth to a whole set of new problems—threatening the level playing field, fostering monopolies, eroding consumer choice, reinforcing brand bias, and more.

Product Prioritisation in AI Shopping

Talking about who controls product prioritisation in AI-led shopping, Jaspreet Bindra, co-founder & CEO, AI & Beyond, highlights the complexity of the power equation. “It’s not just the algorithm—it’s the platform’s commercial model, advertiser influence, and data signals all working together. The concern is that this decision-making layer is invisible to users, even though it directly shapes what they see and buy,” he says.

Ritwik Batabyal, CTO & Innovation Officer, Mastek Group, explains that product prioritisation is shaped by a combination of all three—algorithms, advertiser inputs, and publisher data. “However, the algorithm acts as the final decision-maker, synthesising these signals based on user intent, relevance, and behavioural insights,” he says. “The key challenge will be ensuring that commercial interests do not disproportionately influence recommendations at the cost of user trust.”

Adding further, Batabyal talks about a fourth, often overlooked, actor in this equation: the user’s own historical self.

“Behavioural data from past purchases creates a feedback loop that can trap users in a preference prison — repeatedly surfacing what they have previously bought rather than what they might genuinely need or discover next,” he adds, emphasising that true user-centricity would require AI systems to distinguish between revealed preferences and aspirational intent, and to occasionally surface serendipitous recommendations that break the loop. “That is a significantly harder design challenge than optimising for click-through rates,” he says.

Transparency in AI recommendations

Commenting on the need for transparency in the AI-recommendation system, both Bindra and Batabyal emphasise that it is no longer optional but fundamental to building trust when AI becomes the intermediary between users and the internet.

Bindra cautions that users should be clearly informed whether a recommendation is based on relevance, past behaviour, or commercial influence. “Without that clarity, trust in these systems will erode quickly,” he says.

Batabyal also echoes the thought and adds that explainability frameworks will become a critical differentiator for platforms looking to build long-term credibility.

OpenAI claims to have transparency measures set in place for the Shopping Research feature. In a blog post, it explained that user chats are not shared with retailers and that the results are organic, drawn from publicly available retail sites by reading product pages, citing sources, and avoiding low-quality or spam-heavy sites. It also provided a registration link for merchants who wished to have their products appear in Shopping Research.

The Shopping results in Gemini, AI Overview, and Circle to Search are powered by the Shopping Graph, which updates billions of listings hourly. Google says that instead of just keyword matching, the products are selected based on data quality, price competitiveness, and review scores. The three-step process includes product feeds from Merchant Centre, ingested into the Shopping Graph with added signals like pricing, reviews, and taxonomy. Gemini AI then interprets shopper queries semantically, focusing on intent rather than exact keywords. Finally, results are ranked by feed quality, relevance, competitiveness, and accuracy.

Monopolistic risks in AI-powered shopping

The AI-powered shopping experience, if not governed properly, also poses a risk of establishing monopolies. Experts believe that it could reinforce existing market dominance by consistently favouring well-established brands with richer data signals.

“The fight for the digital front door is also a fight for control. If a few platforms dominate this layer, they can influence not just discovery, but also commerce. That creates a real risk of reinforcing monopolies—where a handful of ecosystems decide what products, brands, or services get visibility at scale,” Bindra cautions.

AI-Powered Shopping Tools Risk Reinforcing Monopolies, Experts Warn (Getty Images)

Batabyal, while emphasising the same risks, believes that with the right regulatory frameworks and ethical AI practices, these systems can be designed to promote fairness and competition. “Ensuring diversity in recommendations should be a conscious design principle, not an afterthought,” he says.

Environmental Cost vs Convenience

Google does not want you to look at Search results. Even though it has a dedicated Gemini app and AI Mode built in Search—for times when you actually want artificial intelligence to save you time or help you understand complex topics—it is also hellbent on using planet's resources to serve you AI-generated snippets (called AI Overviews) for the majority of queries, irrespective of their significance or whether you prefer traditional web results.

As the most popular search engine in the world wants to reshape how you experience the web, there's little you can do—other than ignoring the AI Overviews, which still leaves you with a sense of guilt over wasted resources.

Artificial Intelligence requirements are amping up electricity consumption of data centres (ETV Bharat)

Experts caution about the environmental impact of large AI models and call for a future where AI scales with sustainability. Batabyal and Bindra both believe that while the convenience and efficiency gains of AI are significant, they cannot come at an unsustainable cost. They say that the industry must focus on developing energy‑efficient models, optimising infrastructure, and adopting greener data practices.

Google and OpenAI once revealed the environmental cost of a single text query—0.34 Wh of energy + 0.32176 ml water for ChatGPT and 0.24 Wh of energy + 0.03 gCO2e + 0.26 ml of water. However, Batabyal highlights that this cost is almost entirely invisible to the end user.

“A traditional web search query consumes a fraction of a watt; an AI-generated overview of comparable information can consume orders of magnitude more energy—yet the experience feels identical. There is no carbon signal at the point of consumption,” he says, highlighting that this cost is almost entirely invisible to the end user.

He proposes a system where AI platforms are required to disclose energy costs for every query the user makes, just like food products are required to display nutritional information on every packet. “Until AI platforms are required to disclose per-query energy costs, users and regulators will have no basis on which to make informed demands,” he says. “The sustainability conversation must move from voluntary corporate pledges to structural disclosure requirements."

In the context of AI-powered shopping, Jaspreet Bindra says that convenience alone cannot justify unchecked energy consumption. “We need innovation that is both powerful and sustainable,” he says.