74.6% of Consumers Ask AI Before Buying: Is Your Brand Mentioned?

For Indonesian consumers, AI is starting to become one of the entry points for finding product recommendations. The "The Rise of AI-Assisted Consumer Information Search" research, involving 1,596 Indonesian respondents, found that 74.6% of consumers have used AI to look for product information before buying, while 26.4% have made it a routine part of their purchase decision process.

This shift is changing how a brand gets discovered. When someone asks ChatGPT, for example, "which air fryer brand is the most worth it?" or "which local sunscreen for oily skin is good?", AI generally names only one to three brands. Unlike Google, which shows many search results, AI summarizes the answer, so only a handful of brands make it into the initial recommendation.

This pattern also shows up in various international studies. Eight Oh Two found that 47% of consumers say AI influences which brands they trust. Meanwhile, McKinsey's analysis shows that in categories such as credit cards, hotels, electronics, and apparel, brands ranking at the top of Google do not always appear in AI answers. In other words, a high position on the search results page does not automatically guarantee a brand gets recommended by AI.

This finding is even more relevant given that Indonesia is one of the largest ChatGPT markets in the world. In August 2025, Indonesia was recorded as the country with the fifth-largest number of ChatGPT visits globally, reaching around 216 million visits in a month. This means more and more Indonesian consumers could potentially discover a brand through a conversation with AI, not just through traditional search engines.

Most digital marketing strategies in Indonesia are still built on SEO logic: keyword optimization, building backlinks, chasing rankings. That logic isn't wrong. But it isn't enough for AI search.

Princeton University's research, presented at ACM KDD 2024, is one of the earliest studies to systematically test the factors that make a piece of content more likely to be cited by generative AI engines. The research, conducted together with researchers from Princeton, Georgia Tech, IIT Delhi, and the Allen Institute for AI, built a benchmark of 10,000 queries across 25 domains, then tested nine strategies for optimizing content.

One of the most notable findings relates to the use of data. Content that includes specific statistics has up to a 41% higher chance of being cited compared to content without similar data. Not because the article is longer or repeats more keywords, but because AI can more easily identify information that is concrete and verifiable.

The difference can be seen in a simple example. A skincare article that only writes, "oily skin needs a lightweight moisturizer," gives correct information, but it is still general. In contrast, an article that states, "a survey of 2,400 respondents found that 68% of people with oily skin experienced breakouts after using an oil-based moisturizer," offers a more specific fact. In such conditions, AI tends to choose the second source when composing its answer.

Then, adding expert quotes to an article raises visibility by 28%. In SEO, quotes barely affect ranking. In AI search, quotes become a credibility signal that the model responds to directly.

Content that cites trusted sources raises visibility by 30% to 40%. There's a circular logic here: to be cited by AI, content needs to cite other parties that AI already trusts.

This research also shows that excessive keyword repetition no longer provides meaningful benefit. Unlike traditional search engines, generative engines understand context using a language model, not simply keyword matching.

This research also found an interesting pattern related to a website's position in Google search results. The "Cite Sources" strategy gave the biggest visibility boost to sites ranked fifth on the SERP, at 115.1%. In contrast, for sites already ranked first, visibility actually dropped by 30.3%.

This finding shows that the chance of appearing in an AI answer does not always follow the ranking order on Google. Websites that have not reached the top position still have a chance of being cited as long as they present information that is easy to verify and backed by clear sources.

One of the biggest shifts in AI search lies in the sources of information used to compose answers. While SEO strategy has largely focused on optimizing a company's own website, AI actually draws references from a wide range of sources.

McKinsey's analysis shows that in many product categories, a brand's official website accounts for only around 5% to 10% of the sources AI cites. The rest comes from media outlets, review sites, discussion forums, affiliate websites, and user-generated content.

This finding shows that brand visibility in AI search is no longer determined solely by the quality of an official website. Consumer conversations, media coverage, and independent reviews all play a role in shaping the answers AI gives.

This shift is also making activities like PR, community management, and presence in discussion forums increasingly relevant. These channels have the potential to become reference sources for AI when composing answers for users.

When someone types "local skincare recommendations for sensitive skin" into ChatGPT, which brand shows up and why? Do Wardah, Somethinc, or other local brands that already have high offline brand awareness automatically get mentioned more often? Or is there another factor that matters more?

The Princeton GEO Paper tested English-language content. It's not yet clear whether its findings apply to Indonesian-language content, and whether the AI models predominantly used in Indonesia are exposed enough to quality Indonesian-language sources to accurately assess local brands.

The most practical question will be: how big is the gap between a brand's position on Google Indonesia and its visibility in AI search?

Answering this requires an approach that hasn't been widely done yet. Send dozens of product search queries to several AI platforms, record which brands appear and how consistently, then compare the results with conventional brand awareness data. Dedy Budiman's research (2026) has already mapped the demand side, namely how Indonesian consumers are starting to use AI for product research. What's still missing is the supply side: how visible are Indonesian brands within those AI answers?

McKinsey projects that brands failing to adapt could lose 20% to 50% of traffic from traditional search as consumer decisions increasingly get made inside AI interfaces, before the first click to any website even happens.

Consumers have already changed how they ask questions. The question now is: has your brand changed how it answers?

For Brands: Start to Show Up in AI Answers, Not Just in Rankings

When AI mentions only one to three brands, and 47% of consumers say that answer shapes their trust, a brand that isn't mentioned loses its chance before evaluation even begins. McKinsey projects that brands slow to adapt could lose 20% to 50% of traffic from traditional search. The risk isn't just dropping in ranking. A brand could disappear from the consideration set entirely.

What determines visibility in AI differs from SEO. Princeton's research shows content with specific statistics rises by 41%, expert quotes rise by 28%, and citations to trusted sources rise by 30% to 40%. For brands, materials like white papers, consumer data, and data-driven releases now carry new value. Not just PR material, but material ready to be cited by AI.

McKinsey also found that a brand's website accounts for only 5% to 10% of the sources AI uses. The rest comes from media, forums, and review platforms. A brand that puts nearly all its investment into its own website is competing for only a small slice of what AI actually reads. Community management and presence in trusted third-party sources now also help determine whether a brand gets mentioned.

For the Indonesian market, the question remains open. Indonesia is the fifth-largest ChatGPT user base in the world, but there is not yet data mapping how visible local brands are within AI answers. One simple way to start reading your position: ask ChatGPT, Gemini, and Perplexity the kinds of questions consumers typically ask, then see whether your brand name comes up and how consistently. The results often differ significantly from that same brand's position on Google.

References

Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative Engine Optimization. KDD '24: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 5-16. https://doi.org/10.1145/3637528.3671900 (Preprint: https://arxiv.org/abs/2311.09735)

Silliman, E., Boudet, J., & Robinson, K. (2025, October 16). New front door to the internet: Winning in the age of AI search. McKinsey & Company. https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/new-front-door-to-the-internet-winning-in-the-age-of-ai-search

Eight Oh Two Marketing. (2026, February 4). The 2026 AI + Search Behavior Study: AI Is Now the First Stop for Search. https://eightohtwo.com/blog/2026-ai-search-behavior-study-ai-now-first-stop-for-search/

OpenAI via Reuters. (2026, February 27). ChatGPT reaches 900 million weekly active users. https://techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users

GoodStats. (2025). Indonesia Jadi Pengguna ChatGPT Tertinggi Ke-5 di Dunia. https://goodstats.id/article/indonesia-jadi-pengguna-chatgpt-tertinggi-ke5-di-dunia-mjoj8 (Primary data: Similarweb, August 2025)

Budiman, D. (2026). The Rise of AI-Assisted Consumer Information Search. Fundamental and Applied Management Journal. DOI 10.66314/famj.v4i2.597. SSRN: https://ssrn.com/abstract=6837778