AI assistants are emerging as ‘parallel consumers’ for brands

New Delhi: Brands must now understand and influence not only consumers but also artificial intelligence assistants that are increasingly shaping purchase decisions and brand perceptions, according to Faisal Khan, Managing Director of TheSixth.AI. Speaking at the Market Research Society of India’s webinar, Quality, Clarity, Action: How AI is Re-writing Marketing Research Today, Khan described AI assistants as a “parallel consumer” that marketers and researchers could no longer afford to ignore. The webinar, held on July 22, brought together Khan; Soham Chakravorty, CMI Manager, Skin Cleansing, Hindustan Unilever Limited; and Ashwani Kumar Singh, Strategic Business Advisor at TrayiStats AI Technologies. Focusing on how AI is transforming brand-health research, Khan said brands traditionally measured awareness, consideration, reputation and loyalty by questioning consumers. The emergence of platforms such as ChatGPT and Gemini, however, had added another layer to the decision-making ecosystem. “Earlier, the job of brand custodians was to understand, measure and influence consumers directly. Now, in addition to doing that, they must create a mechanism to measure and influence AI assistants,” Khan said. He said marketers needed to examine whether their brands appeared when consumers sought recommendations from AI platforms, how these platforms presented their strengths and weaknesses, and what overall perception of the brand they generated. “They hold an opinion about you. When a consumer goes online and asks an AI assistant for a recommendation, does the brand show up or not? How is it projecting the brand?” he said. Traditional brand tracking leaves a blind spot Khan argued that conventional brand-health studies could no longer provide a complete picture because consumers continuously left behavioural signals across search engines, social platforms, reviews and AI assistants. Search activity, he said, represented active consideration and could, in some categories and markets, emerge as a leading indicator of future market share. Social media showed how narratives and perceptions were created, while reviews captured the gap between promises made by brands and consumers’ actual experiences. AI could bring these otherwise independent signals together, analyse them at scale and identify how they influenced one another. “Traditional brand health, as we understand it today, leaves us with a huge blind spot in the ecosystem,” Khan said. He added that the importance of AI in research should not be reduced to faster analysis, automated presentations or cost efficiencies. “Speed and cost efficiencies are realities, but they are not what makes AI important. What makes it important is trust, robustness and reliability,” he said. Khan also pointed to a shift in research deliverables. Instead of ending with a static report or presentation, research projects could now produce interactive platforms through which marketers could ask follow-up questions, generate charts, explore data and prepare summaries independently. “The data becomes much more actionable and democratised. It becomes a living document or a living process that can continue throughout and even after the research,” he said. Human insight continues to give AI meaning Chakravorty said AI was helping large organisations combine years of reports, studies, dashboards and consumer insights into searchable knowledge systems. This could make insights accessible not only to research and marketing teams but also to employees working across functions such as supply chain and sales. “AI helps amplify research functions, and human insight gives it meaning,” Chakravorty said. According to him, AI could synthesise information and identify connections, but researchers would still need to understand the emotions behind consumer behaviour and decide how much importance to give conflicting evidence. “Not all truths hold equal weight. Contradictions are the nature of our business, and the final judgement will still have to be taken by the researcher,” he said. Poor data can produce confident but incorrect answers Singh focused on protecting research quality as AI-generated responses and automated survey fraud become more difficult to distinguish from genuine human participation. He cautioned researchers against relying only on writing style, polished language or particular phrases to determine whether an open-ended response had been produced by AI. “The job is not to claim with certainty whether an answer was written by AI. The job is to flag low-effort or inconsistent answers and place them before a human for a decision,” Singh said. He added that automated detection tools could sometimes incorrectly reject genuine responses, particularly those written by people whose first language was not English. “The real risk is that bad data does not disappear with AI. It comes out more polished, more confident and harder to challenge,” Singh said.