Tackling scams, misinformation: Deep dive into Perplexity’s WhatsApp bot proposition
The underlying methodology defining Perplexity WhatsApp bot is the same as Perplexity’s AI app or the interface on the web — access to multiple models
500 million users and counting, that’s the estimated number of WhatsApp users in India (globally, the number hovers around 2 billion), and that’s the barometer of scale and relevance, since no official numbers from Meta are available. For Perplexity, to find a way to be at the fingertips of this user base, is an achievement that is perhaps theoretically matched only by Meta’s own Meta AI. This week, Perplexity CEO Aravind Srinivas announced the Perplexity WhatsApp bot (save +1 833 436-3285 as a contact on your phone) without any subscription, with a rather detailed feature set.
The company tells HT it is still early days in terms of sharing usage metrics or trends, but they hope to tackle misinformation and an increasingly worrying trend of scams by allowing the bot to fact-check and verify dubious claims users often find going around in WhatsApp forwards, by using the latest artificial intelligence (AI) models. “WhatsApp forwards are dangerously filled with click-baits and propaganda. And AI is the only way to end it once and for all,” says Srinivas, in a statement. This should give Perplexity a stronger footing as it competes with everything rivals including OpenAI and Google are doing, in a rapidly evolving AI ecosystem.
The current arsenal that underlines the Perplexity WhatsApp bot, in addition to fact-checking messages, includes a voice mode for conversations with the Perplexity AI models, decoding videos for context and accuracy of information, as well as well as generating media including memes, for sharing in chats. The intention is to crack group chats as well, but that’s proving to be a hurdle. “This one is hard, as the official API doesn’t support this. We will figure out something here hopefully together with Meta,” Srinivas remains confident.
The fact-checking ability in particular is quite seamless, and Perplexity hopes to be useful particularly for senior citizens, in decoding the truth in messages they receive. The way to use this is simple — forward a message that you’ve received, to the Perplexity WhatsApp bot. This can include text messages, images, links or videos. Perplexity will cross-reference any claims being made, with real-time web data through search — complete with citations for the information. This will also help protect against scams, that often include malicious links, for instance.
At this time, the Perplexity WhatsApp bots supports 20 Indian languages, including Hindi, Telugu, Tamil, Bengali and Urdu. There is a likelihood that a message a user forwards to the bot, will be understood, with a response in generation. There have been recent reports of scams particularly targeting elders, rampant on messaging platforms, including examples of a 75-year-old who lost ₹11 crores to scammers who posed as a financial services company on the platform, and fake trading app, and a 77-year-old losing ₹3.8 crores after finding herself at the receiving end of a digital arrest scam.
There are some limitations too using Perplexity within WhatsApp. It will, not play along with any news-oriented searches via WhatsApp. A least from HT’s conversations with the bot, it was perplexing stumped when we asked it to detail today’s IPL games and the current points table in that context. Neither would it detail the latest updates on the latest situation between India and Pakistan on the Kashmir issue — the Perplexity app does, though. It is these little details which will define whether a user sticks around, or switches to another AI app.
AI beyond an app, and a need to build
The underlying methodology of the Perplexity WhatsApp bot is the same as the Perplexity app or the interface on the web — access to multiple models. There’s Anthropic’s Claude 3.7 Sonnet and Sonnet Thinking, OpenAI’s GPT 4.1 as well as o4-mini, Google’s Gemini 2.5 Pro, Grok 3 Beta and the R1 1776, which Perplexity calls an “unbiased reasoning model”, available to users as part of the bot’s proposition. Unlike the app or the web where a user can choose a specific model to interface with for any query, the Perplexity WhatsApp bot will make the selection automatically based on the sort of query it has received. This, in a way, simplifies things for the WhatsApp audience Perplexity wants to find relevance with.
The Perplexity WhatsApp bot isn’t the first integration of its kind with messaging platforms, but there seems to be an intent to see this through, unlike efforts seen previously. OpenAI launched a ChatGPT WhatsApp integration late last year, but for a large part, they are more general-purpose conversations and don’t often have citations the way Perplexity integrates with responses. xAI’s Grok finds integration in Telegram for Premium users — but Telegram’s user base in a market such as India, is comparatively smaller.
There is of course WhatsApp owner Meta’s AI push, called Meta AI, which was unlocked on WhatsApp, Instagram, Facebook and Messenger, last summer. It doesn’t need users to save a specific contact, and is integrated within the app at various touch-points, including the search bar. There are, two distinct elements to Meta AI. First, it has faced criticism for the permanent button approach which still cannot be turned off or hidden (there is also a risk a user may accidentally trigger it). Perplexity, and indeed OpenAI’s WhatsApp bots, need to be manually called upon. Secondly, Meta AI is more focused on casual queries including image generation, and isn’t optimised for research depth and source transparency, unlike Perplexity’s approach.
The key to success in coming months will be the promised group chat support, as well as building the fact-checking capabilities as scams and malicious intent evolves. The feature set delivers on the promise, for now, but it will be crucial to build broader language support for the multiplicity of Indian languages and dialects, to be able to tap most of WhatsApp’s user base in the country. And of course, avoid the sort of stumbles that we faced when we asked the bot about IPL.
