'Artificial scare'? 65% Bengalureans prefer chatbots over search engines, finds survey

Bengaluru: A recent survey conducted in Bengaluru has revealed that over 65% of respondents now rely on AI chatbots for information, surpassing the use of traditional search engines like Google.

The survey highlights a growing trend in AI adoption, extending beyond information retrieval to applications in traffic management, healthcare, and research. However, despite the increasing usage, experts caution against the limitations and risks of this early-stage technology.

Daily searches

Titled ‘How Bengaluru Uses AI’ and conducted by LocalCircles, the survey polled 2,133 respondents.

The findings indicate that a significant 33% of users now turn to ChatGPT for everyday searches instead of traditional search engines. In addition, 11% prefer Perplexity, while 7% and 5% use Co-Pilot and Gemini, respectively.

Explaining this shift, AI enthusiast and entrepreneur Siddharth Das noted that search engines primarily collect vast amounts of data based on keywords, but do not necessarily solve problems or provide direct answers.

In contrast, AI chatbots analyse user queries and offer tailored insights. He pointed out that while a search engine can present multiple TV options, an AI chatbot can compare models and recommend the best one based on specific user needs.

Voices of concern

While acknowledging AI’s efficiency, Srinath Srinivasa, Dean of Research and Development at IIIT Bangalore, cautioned against hallucinations — a phenomenon where AI generates false or misleading information.

He explained that since AI chatbots function as next-word prediction models, they are designed to please users, sometimes fabricating information in the process. He warned that users must be cautious not to accept such content at face value.

The survey found that 44% of respondents considered AI-generated information to be "somewhat accurate." Meanwhile, 24% believed it was "mostly accurate," and 9% found it "extremely accurate."

However, concerns persist about AI perpetuating societal biases. A report submitted to the United Nations by Ashwini KP, UN Special Rapporteur on Contemporary Forms of Racism, revealed that AI models trained on biased data can reinforce prejudices, including racism and xenophobia.

The survey also explored how people are using AI in various applications.

It revealed that 51% of the respondents rely on AI for general information retrieval, while 32% seek information not easily accessible through standard search engines.

Document summarisation emerged as another popular use case, with 27% leveraging AI for this purpose.

Meanwhile, 15% use AI for writing tasks, and 11% and 10% utilise it for image and video creation or enhancement, respectively.

While AI-powered summarisation has been widely acknowledged as one of the technology’s most successful applications, experts continue to warn against sharing personal images and videos with AI tools.

Siddharth Das downplayed privacy concerns, attributing them to a broader lack of legal safeguards in India. However, Srinivasa was cautious as he said AI models could retain and use such data for training purposes, potentially leading to privacy risks.