Usage & attitude study
AI vs. Friends: Who Rules Your Wallet?
This Usage and Attitude study pitched AI shopping assistants against advice from friends and family among tech-forward urban consumers aged 18 to 35. The report concludes that AI warnings beat peer advice on high-risk purchases, while social proof still wins on low-risk ones. Its scenario data was unavailable, though, and its quality figures contradict each other.
Explore the findings1,879 of 3,758 respondents screened out, per the methodology
Against a target of 3,000, -37.4% vs target
Stated for n=1,879, leaving zero valid completions
Based on 300 completed responses. Usage & attitude study.
What we set out to answer, and what we found
When an AI assistant contradicts your best friend, who do shoppers actually believe?
Does a high price change whose advice wins?
Scenario 1 set an AI warning about hidden battery issues on a ₹40,000 smartphone against a best friend insisting the brand is the best.
What about a low-stakes, feel-good buy?
Scenario 2 set an AI warning about fabric shrinkage on a ₹1,500 jacket against a sibling saying it looks amazing.
How much do people trust AI on its own?
We looked at how much people trust independent AI search assistants such as ChatGPT and Gemini when they want unbiased product advice.
Will they trust a brand-owned in-app AI?
We tested whether consumers treat an independent AI assistant differently from an official, brand-owned in-app assistant.
Risk moves trust towards AI. The report concludes that on high-ticket items, AI warnings about hidden defects override enthusiastic peer endorsements, while low-risk lifestyle purchases still rest on social validation. Brands should match their validation strategy to the price point.
A negative AI verdict is treated as a dealbreaker. Under the report's Social-Algorithmic Defiance Index (SADI), a flagged product issue acts as an objective, data-driven veto that social pressure struggles to overcome. It can cancel out word-of-mouth, so managing how LLMs talk about your product becomes a defensive priority.
Brand-owned in-app AI is not trusted yet. Independent AI assistants are described as trusted for unbiased pre-purchase research, while in-app brand assistants meet scepticism over commercial bias and data privacy.
Search visibility alone is no longer enough. Consumers consult LLMs to stress-test peer recommendations, so the report argues brands must manage their story inside generative AI as well as through conventional SEO and social media.
No reported response data backs these conclusions. Both scenario sections and every driver, cohort and ranking analysis returned no data, and the quality figures contradict each other. Treat the findings as hypotheses until the study is re-fielded.
Fielded Questions
Most response sections returned no data
Data Quality: Contradictory Sample and Attention-Check Figures
Data Quality Screen as Stated in the Methodology
Base: 3,758 respondents screened
View the data behind this chart
| Answer | Share |
|---|---|
| Excluded on quality (1,879) | 50% |
| Retained for analysis (1,879) | 50% |
Attention Check Answers in the Final Analysed Sample
Base: 300 respondents, per the attention-check appendix
View the data behind this chart
| Answer | Share |
|---|---|
| Option C (correct) | 100% |
| Option A | 0% |
| Option B | 0% |
| Option D | 0% |
01. The methodology states 1,879 completed responses against a 3,000 target, after 1,879 of 3,758 respondents (50.0%) were excluded in attention-check and data-quality screening.
02. A later quality alert says all 1,879 respondents failed both embedded attention checks, a 0.0% pass rate, which leaves zero valid completions against a target of 300.
03. The attention-check appendix tells a third story: a 100% pass rate in a final analysed sample of 300, with all 300 choosing Option C and every other option at 0%.
A Premium Urban Cohort of 300 Digital Natives
Gender Split of the Analysed Base
Base: 300 respondents
View the data behind this chart
| Answer | Share |
|---|---|
| Male (n=150) | 50% |
| Female (n=150) | 50% |
01. All 300 respondents are Gen Z and younger Millennials aged 18 to 35. They live in Metro and Tier 1 cities (Delhi, Mumbai, Bangalore and Hyderabad) and fall in the NCCS A1 and A2 bands.
02. Every respondent shopped online at least 2 to 3 times a month, and every one had used a generative AI tool such as ChatGPT, Gemini or Copilot in the last 30 days.
03. The split is an exact 50% male (n=150) and 50% female (n=150).
AI Assistants Are Seen as Neutral Information Filters
The baseline-trust section is qualitative. The source reports no distribution behind it.
01. Consumers are described as turning to conversational AI because they see it as free of the commercial bias, sponsored listings and SEO-manipulated rankings of traditional search.
02. Trust in AI is said to depend on purchase risk. Consumers use AI search to make sense of complex technical specifications, then turn to friends and family for final reassurance.
03. If AI is seen as a neutral aggregator, a negative AI recommendation becomes an immediate purchase barrier, and social proof alone cannot be relied on to overcome it.
The Scenario and Driver Sections Returned No Data
01. Both conflict scenarios, the ₹40,000 smartphone (AI warning vs best friend) and the ₹1,500 jacket (AI warning vs sibling), report that response data was unavailable or insufficient. The headline claim that risk shifts trust therefore has no reported distribution behind it.
02. These sections all returned no usable data: brand-owned AI, three-year outlook, credibility drivers, adoption barriers, segmentation, cross-scenario, AI-usage cohort, Borda ranking, Algorithmic Trust Index and research co-occurrence.
03. The appendix tables for pre-purchase research channels and brand choice triggers show n=0.
Algorithmic Validation Is Framed as the New Conversion Gatekeeper
The report's closing conclusions, stated without supporting figures.
01. On high-risk decisions, an AI warning about hidden battery issues on a smartphone is said to challenge even a close friend's recommendation. Premium brands should therefore prioritise optimisation within large language models.
02. On low-risk decisions, such as a casual jacket where AI warns of shrinkage but a sibling approves, social proof is said to win. AI warnings still cause significant hesitation.
03. Trust is described as depending on context: peer networks for social validation, AI tools for technical risk. That calls for a marketing strategy with two tracks.
04. The conclusion contradicts the key findings on in-app AI. It calls consumer willingness to trust brand-owned assistants a "massive opportunity", while the key findings describe a severe trust deficit.
What we would recommend
Practical priorities drawn from this study for further validation.
Re-field the study before acting on it
The source calls for pausing the current panel and re-fielding the 15-question survey with a different, double-opt-in panel provider. That provider should use device fingerprinting and real-time fraud detection, so the study secures the 300 high-quality respondents it needs.
Split marketing channels by risk profile
For high-ticket, high-risk product lines, prioritise LLM search optimisation and technical PR. For lower-priced, impulse-driven categories, stay with social proof, influencer collaborations and peer referral programmes.
Make in-app AI openly transparent
To close the trust gap facing brand-owned assistants, copy the objective, data-driven style of third-party platforms. Point out product limitations without being asked and offer side-by-side comparison data in place of purely promotional recommendations.
Set up an LLM sentiment monitoring protocol
Negative AI recommendations are treated as dealbreakers, so monitor how the major conversational models describe your products. Build a feedback loop that deals with common objections before they cancel out your word-of-mouth campaigns.
