Top Credit Cards in India via An LLM

I wanted to upgrade my credit card and decided to use an LLM to arrive at an answer. Goodbye SEO and hello AI Search Optimization?

I will use -
1. ChatGPT 4o
2. Perplexity Free Version
3. Claude 3.5 Sonnet

Initial Prompt - What are the top credit cards in India that have high reward points? Assume a moderate spender with diverse spending habits. (I then used a few more prompts in Claude and ChatGPT to get a table and ranking. Your prompting may result in different results).

I then collated all this info into a google spreadsheet and limited it to the top 5. See results below (add HTML via Claude)

Complete Colorful Credit Card Comparison

.credit-card-table {
border-collapse: collapse;
width: 100%;
font-family: Arial, sans-serif;
font-size: 14px;
box-shadow: 0 0 20px rgba(0, 0, 0, 0.1);
}
.credit-card-table th, .credit-card-table td {
border: 1px solid #ddd;
padding: 12px;
text-align: left;
vertical-align: top;
}
.credit-card-table th {
background-color: #4CAF50;
color: white;
font-weight: bold;
text-transform: uppercase;
}
.credit-card-table tr:nth-child(even) {
background-color: #f9f9f9;
}
.credit-card-table tr:hover {
background-color: #f5f5f5;
}
.source-perplexity { background-color: #e6f3ff; }
.source-chatgpt { background-color: #e6ffe6; }
.source-claude { background-color: #fff0e6; }
.credit-card-table td:first-child {
font-weight: bold;
color: #333;
}
.credit-card-table .card-name {
font-weight: bold;
color: #0066cc;
}
.credit-card-table .score {
font-weight: bold;
color: #ff6600;
}

Subsequent prompting by asking each LLM to compare the output and recommendation of the others and spotting any errors - Prompt- "Attached is the recommendations of other AIs/LLMs including yours (in CSV and PDF). Do you agree or stick with your recommendation? Is there anything incorrect with the other LLM's recommendations"

Response from ChatGPT -

For High Spenders: My original recommendations are valid and focus on maximizing rewards and premium features. The high annual fees and joining fees are justified by the benefits.

For Moderate Spenders: Perplexity's recommendations are more appropriate. They emphasize cards with lower fees and balanced reward structures, which might be better suited for users with diverse but moderate spending habits.

Claude responded by providing an explanation of why the other LLMs might give a different response. Perplexity hallucinated

I then downloaded all my Credit Card statements and fed it to Claude to analyze my spending patterns. I used the new project feature and fed into the knowledge base. The analysis in Claude was disappointing. It had errors in the total spends and the spends per category which I had to disregard.

Other Observations -

They rely on historical information and don't take into account latest information. This is expected given they are trained on old data. If you have a similar query, you should consider an LLM that is more recent or has access to recent information. The reward points above, annual fees etc aren't 100% accurate. They are close and you will need to go to the bank's website to verify.

After noticing the above, I decided to use other traditional methods to arrive at a decision. As newer models develop and the choices expand, it gets more important and confusing on which LLM to use for which task. This was the biggest limitation. I did not select a card listed above as there was a newer card not captured above which I found better. However, it still remains a good starting point as it gives you options to select from. The cards listed above are in the top from a historical perspective but there are newer releases that one has to factor in.

I ended up visiting blog posts via google and found this website quite helpful. On a side note, finding a good card really depends on your spending habits and the type of rewards you're looking for.