Today’s generation has a new idol – generative AI. And honestly, who wouldn’t be seduced? Ask an AI tool to explain a balance sheet and it responds like your favourite finance professor. Ask it to summarise a 200-page annual report and it does so with the patience of a diligent chartered accountant. Want stock ideas? It generates them in fluent, confident prose. It is tireless, smooth, and seemingly wise. Of course we’re tempted.

But hidden beneath this fluency is a serious problem. Generative AI isn’t expanding our mental models—it’s narrowing them. It is quietly reinforcing one of our most persistent investing biases: familiarity. And if we’re not careful, we could find ourselves with portfolios that look eerily similar, carry the same risks, and collapse in unison – not because AI knows too much, but because it knows too little, and hides that ignorance behind feigned authority.

The repetition machine

As investors, we imagine AI as a diligent analyst that scans filings, evaluates fundamentals and forms views. But that’s not what it does. Generative AI doesn’t “think” – it predicts. It arranges the most statistically likely words in the most statistically likely order. And what dominates its training diet? Firms with the biggest digital footprints—big banks, telecom companies, IT giants and other large caps.

Big names in these sectors appear everywhere, from earnings blogs and finance threads to case studies and podcasts. That’s what the model sees. That’s what it harnesses. So, when you ask it, “What should I buy for the long term?”, it doesn’t weigh fresh data or live valuations. It pulls from ‘memory’. And because frequency breeds familiarity, the most visible names become the most suggested. The result? Comfort cloaked as insight, wrapped neatly in a smooth paragraph that’s just polished enough to be persuasive, just hollow enough to avoid scrutiny.

Real opportunities remain hidden

The Indian market isn’t a neatly packaged Nifty 50 or Sensex. It’s chaotic, uneven, and full of edge cases. The real growth stories are often hidden in the shadows—specialty chemical exporters in Vapi, niche engineering firms in Coimbatore, logistics networks in Bhiwandi, and small-town non-bank lenders with stellar books. These companies don’t get mainstream press. They don’t make LinkedIn lists or YouTube thumbnails. They rarely show up in AI training data, so they rarely show up in AI answers.

Ask for “undervalued small caps” or “emerging mid-cap bets”, and the AI model still circles back to the same old suspects using new language. Meanwhile, the real compounders—the ones building quietly in Rajkot or Ludhiana—remain invisible. It’s a savage irony: the tool that’s thought to democratise discovery ends up reinforcing what’s already known.

Algorithmic consensus

And here’s the scary part. If enough investors lean on the same AI assistants, and those assistants lean on the same training data, we could all end up owning the same portfolio. Instead of diverse perspectives, we’ll have algorithmic consensus. Instead of healthy debate, we’ll have statistical convergence.

This isn’t just lazy investing, it’s dangerous. When too many portfolios hold the same set of familiar stocks, corrections become crashes as the exits get crowded. What looks like safety is in fact synchronized fragility. This isn’t diversification, it’s herd behaviour.

Western lens distorting the Indian view

There’s a deeper issue at play here. The bulk of content AI models are trained on is Western, especially American. AI speaks ETF fluently; it thinks in FAANG. Its mental model is shaped by Wall Street narratives and S&P 500 logic.

So when an Indian investor seeks guidance, it often responds with American frameworks — index-heavy allocation models, dividend strategies, passive investing mantras. But Indian markets don’t follow that script. We have promoter risk, capex booms, micro-cap turnarounds, and regulatory shocks. What works in New York doesn’t always map on to Mumbai. Yet we keep seeing our questions answered through this foreign filter, and wondering why the advice feels off.

AI isn’t the enemy, blind faith is

Generative AI has potential. It can simplify complex reports, translate jargon, and make financial learning more accessible. But it is not a stock-picker. It is not a strategist. It is certainly not a shortcut to outperformance. It will only make us smarter if we treat it as a tool, not a tutor.

Ask it to challenge your views, not endorse them. Ask for risks, not just trends. Ask it to explain the plumbing of industries, not pick the best faucets. Use it to think better. But don’t outsource your thinking.

Dr. Chandni Rani is assistant professor at Galgotias University; Dr. Simarjeet Singh is assistant professor at GLIM-G.