There’s no question that earning and spending credit card rewards has become far more complicated in recent years.
Case in point, while planning a simple eight-day trip to Portland, I took all of the following steps to ensure that I was maximizing the current suite of cards in my wallet:
- Moved my Chase Ultimate Rewards points from my Freedom Unlimited and Freedom Flex accounts onto my Sapphire Preferred account so that my points would have 25% more value when booking through the Chase travel portal.
- Cross-checked the cost in points to book a Delta flight through the Chase Travel Portal, Capital One Travel and Citi Travel – as well as current point-to-mile transfer ratios to see which overall option was cheapest.
- Applied for the Marriott Bonvoy Boundless card in order to score instant status and a three-night welcome bonus (after $3k worth of spending) to cover the first chunk of my hotel stay.
- Split the rest of the hotel bill between my Citi Premier card to score a $100 statement credit and my Chase Sapphire Preferred (CSP) to score a $50 statement credit (yes, the front desk thought I was nuts).
- Booked my rental car using my CSP instead of my Citi Premier – because even though the Premier offers 3X on travel, the CSP offers 2X worth 2.5X towards travel plus a free Auto Rental Collision Damage Waiver.
All told, my calculated approach saved me over $1,000+ but the time and effort involved felt like planning the Normandy invasion. Plus, I certainly wouldn’t expect the average person with limited time, energy and obsession with credit card rewards to spend hours finding the most point-efficient path forward.
In other words, I doubt that many others spend their Sunday afternoons poring over points, statement credits and transfer ratios while The Price is Right echoes in the background.
Does AI provide another way?
That’s why my colleagues and I started wondering: Is rewards management something you can outsource to AI? Can an algorithm like ChatGPT advise you on things like:
- How to spend your credit card rewards points?
- Which card you should use to book certain travel categories, like hotels?
- What your credit card points are worth in real-world value?
And finally, can an AI advise you on how to book a vacation based on the current cards you have in your wallet?
Such abilities don’t seem outside the realm of possibility for AI. After all, AI algorithms tend to base their answers off of data scraped from the internet – and there’s certainly no shortage of blog posts, fact sheets, reviews and Reddit discussions about virtually every rewards card on the market.
So to find out if robots could help us, I had a literal conversation with the most recent version of OpenAI’s ChatGPT, GPT-3.5, to test its “knowledge” of credit card rewards. The paid version (GPT-4.0) is slightly newer, but I wanted to approach this from the perspective of an everyday credit card user who probably isn’t paying a subscription.
On that note, I’m also not an AI Prompt Engineer. Thanks to various TEDx Talks I’ve grown to appreciate that there’s a special “way” to speak to AI for optimal results—and for my part, I really only know the basics (be clear, specific, etc.).
So like Domhnall Gleeson in the movie Ex Machina, I’ll be approaching my first major encounter with AI like a normal, everyday guy: curious, excited and keen for answers to life’s biggest questions:
So without further ado, let’s start the interview.
Can an AI tell you how to spend your credit card points?
To break the ice on my date with ChatGPT, I gave it a softball question: what’s the best way to spend Chase Ultimate Rewards points?
You can see its response below. First impressions were good – seemingly pulling data from Chase.com itself, our AI friend was keenly aware that the Chase Sapphire Preferred/Reserve cards offer a nice bonus multiplier when redeeming points for travel (1.25X and 1.5X, respectively).
So far, I’d say ChatGPT is passing the Turing test—meaning if I got this answer in an email, I wouldn’t be able to tell you whether it was written by an AI or by someone else at Fortune Recommends.
Now that we’ve warmed up the servers, let’s try a tougher question.
Can an AI tell you which card is better for use on certain spending categories (e.g. hotels)?
Next, it’s time to test GPT-3.5’s knowledge of travel rewards. Which card does it recommend for specifically hotel stays: the Chase Sapphire Preferred or the Capital One Venture Rewards?
Here’s what Wi-Fi WALL-E had to say:
Overall, I was a little less impressed by ChatGPT’s answer this time around.
Credit where credit’s due, Chase does have a slightly better selection of hotel transfer partners (IHG Rewards, Marriott Bonvoy and World of Hyatt compared to Capital One (Wyndham Rewards and Choice Privileges).
However, I disagree with the notion that the Capital One Venture Rewards offers more “simplicity and flexibility” for travel redemption. In truth, the CSP offers 5X on all travel booked through Chase while the Venture card only offers 5X on hotels and rental cars.
But hey, maybe that one’s on me. Maybe I could’ve phrased the question a little better to solicit a more helpful and conclusive answer. So I gave it a second try, clarifying to ChatGPT that I wanted the best card for booking hotels, not redeeming points for them.
Unfortunately, my reframing of the question seemed to flummox our friendly neighborhood robot even more, resulting in a boilerplate description of both cards:
Once again, I do have to give ChatGPT credit for providing a helpful and cohesive string of information – even if it didn’t answer the original question.
Three questions in and we’re getting mixed results. The AI seems to know how to spend rewards points effectively, but struggles to compare two cards directly (at least, with the limited prompts I’ve given it thus far).
But let’s change the subject a bit and see if ChatGPT can help us with another tedious and confusing part of redeeming credit card points: assigning a value to them.
Can an AI tell you what your credit card points are worth?
The days are long gone when credit card points were worth a penny each. Now they can be worth anywhere from 0.2 cents (as is the case when converting IHG One Rewards to Delta SkyMiles) all the way up to 2.8 cents (when converting AmEx points to ANA Mileage Club).
Most banks are pretty upfront about point values—but only if the point values are above a penny each. When they’re worth less, their weak cash value is often buried in the terms and conditions.
That’s why it would be mighty helpful if our AI assistant could tell us what our points are worth, on average, before we mistakenly redeem them for poor value (or apply for the wrong card in the first place). So let’s ask:
Not bad, ChatGPT, not bad!
Presumably, the algorithm pooled together estimates from multiple personal finance sites to find an overall average. I’m also impressed how it’s able to provide value estimates by category, giving us the heads up that we’re probably best off redeeming our AmEx points for travel (which is 100% true).
Now let’s see how well it does with hotel points, such as the Marriott Bonvoy points you’ll earn with the excellent Marriott Bonvoy Boundless card:
Again, a solid and helpful response. ChatGPT knows that your redemption values start to fall as you look at Category 7s and 8s like the Ritz-Carlton. More pertinently, it knows that the secret to spending Marriott Bonvoy points is to book during off-peak pricing whenever possible.
So far, ChatGPT seems great at:
- Giving you the highlights of a rewards card you might be considering
- Estimating the cash value of certain rewards points
- Offering suggestions on how to redeem your points for maximum value
It seems less effective at comparing the merits of two distinct rewards cards, but hey, we can just keep doing that on our own.
Now for the final exam.
Can AI help you plan a vacation around what cards are in your wallet?
Naturally, there are tons of ways I could ask this question. And since ChatGPT-3.5 is free, I’d encourage you to go try a few combinations yourself.
But in my case, even my very first attempt at a prompt resulted in a rather intriguing—and revealing— answer from our AI friend:
The most impressive part of this response is that ChatGPT seemingly knows that the Chase Sapphire Preferred has travel insurance and the Citi Premier doesn’t. That may seem like a minor footnote, but it’s a distinction that could save you up to $20,000 per trip.
I’m also a fan of the General Tips the AI provided, which could also be huge time- and money-savers.
But keen-eyed credit card nerds will notice something… off… about this response. The AI didn’t mention any statement credits, and more tellingly, it got the Citi Premier card’s reward categories totally wrong. According to Citi, the card gets 3X on dining, not 2X, and there’s no rewards category for “Entertainment.”
So where did ChatGPT get this information?
The massive caveat to all this: ChatGPT’s data is old
ChatGPT may pull data from every corner of the internet, but it’s not today’s internet. It’s an archive from January, 2022.
To the AI’s credit, if you ask about something current it’ll often warn you of this limitation:
So what does this mean within the context of using AI to spend rewards points?
Well, considering just how fast the credit card industry moves, it means ChatGPT’s usefulness as a modern day assistant is limited. Without knowing the latest rates, fees, rewards, transfer rates, or even which rewards cards are currently available, there’s not a whole lot of guidance the chatbot can provide without some degree of uncertainty.
You can always pay $20 per month to upgrade to GPT-4 Turbo, which pulls from data as recent as April, 2023. But even then, your AI assistant may not be aware of the latest and greatest benefits, such as the World of Hyatt Business card’s temporary 75,000-point welcome bonus. So until we get an AI ravenous for real-time data, we’ll have to settle for something that’s helpful—with an asterisk.
The final verdict
ChatGPT’s ability to synthesize data from multiple sources, extract the most helpful nuggets and tie it all together in a cohesive, legible response is beyond remarkable. I threw it some moderately tough questions for this article and many of its responses looked like email responses I might’ve gotten from a colleague (though wordier, and with fewer emojis).
But the chatbot’s inability to incorporate data from the last nine months is a severe and subjectively deal-breaking limitation when it comes to offering credit card advice. With benefits and rewards constantly coming and going on a weekly basis, you’re still going to need someone (or something) who can process that data in real-time, producing hot-and-fresh recommendations that can save you time, money and points.
So until the machines rise up to replace us, Fortune Recommends is here to help.