How ChatGPT is Turning CD Depositors Into Rate Hackers
Consumer ignorance once shielded banks and credit unions from flaws in their deposit products. But now, AI is quietly arming depositors with the tools to outsmart outdated CD structures. With just a simple prompt, anyone can uncover loopholes that cost banks big time. The era of passive depositors is over. Are you ready?
By Neil Stanley
Simple Subscribe
Subscribe Now!
While banks decide if and how they will use artificial intelligence, they face another more urgent question: How will AI affect depositors’ ability to pursue their best interests?
Previous advances in the age of information – Google, social media, mobile phones – enhanced access to information. Now AI can use that information for you, including all the features and unintended options you put in your deposit accounts.
For some, those questions will be simple, such as "If I have a $100,000 CD at 1% interest and it will mature on May 20, 2026. How much will I earn?" The day I tried this question in ChatGPT, instead of telling me how to get the answer, it accurately said: Total balance at term: $100,000 + $1,000 = $101,000.
That’s a contagious result because it’s very convenient; it removes the need for depositors to overcome the unknown. What happens when AI can answer questions that affect the returns they receive on their deposits? What happens if those questions take advantage of deposit product design flaws made exploitable by AI?
Combined with other technology, AI advances enable a new kind of "AI-driven yield harvesting" that is extremely dangerous for banks’ and credit unions’ cost of funds. Others have covered why it threatens savings and money market accounts, so let’s look at the traditional certificate.
What Certificates Are ‘Exploitable’ by AI?
The certificate’s penalty structure was created before calculators. When Walter Wriston designed the certificate in 1961, the CD was given a penalty of "X number of days of interest" because it was simple to calculate…without a calculator.
Depositors have viewed the CD penalty as something to be avoided. No matter how big or small the actual dollar amount, conservative CD investors don’t want to pay it. More active investors may use CDs but have never attempted to calculate the penalty, even with the help of programs like Microsoft Excel or online calculators. When "penalty equals bad," why investigate further?
This assumption has prevented depositors from making a dangerous realization for their bank or credit union: In a rising rate environment, like the one the industry saw in 2022 and 2023, I could triple my return by paying my CD penalty.
Now, all depositors have to do is ask an AI platform, and it will tell them how to find and execute on that 300% increase in return if it’s available.
So, how big is this risk? And, is it probable, rather than just possible, that it will result in losses for banking institutions?
How Many CDs Offer Depositors a 300% Return Increase?
Aside from no-penalty CDs, or those not as widely used, most banking institution certificates use the "X number of days of interest" structure for penalties. Industry regulators do not report on this type of information, nor do they report certificate volumes stratified by term and rate.
Still, we can get a sense by analyzing a group of client institutions as if they were one bank to size up the deposit profits at risk from AI. Combined, these 10 institutions have $4.04 billion in certificates.
How much CD volume has a low rate and a maturity of more than 12 months?
Add all the CD volumes with a rate of less than 3% across all terms; that’s $530 million for which depositors can achieve higher returns. For the volume below 1% – some $112 million – depositors can achieve a 300% increase in return, especially when wholesale and U.S. Treasury rates are where they are today. These are the most profitable deposits in the entire CD portfolio.
The issue here for banks’ cost of funds is that if a 1% CD leaves, they have to replace the funding. If wholesale funding is now 3% higher, as often happened during the past three years, the replacement rate will likely be close to 300% higher for the institutions.
Now, if the depositor paid an early withdrawal penalty of 90 days on $100,000 at 1%, the bank collected $250. That penalty, though, is small because it’s based on the interest rate rather than the market value of the CD. It neither prevented the CD’s attrition nor did it take a meaningful bite out of the cost of funds increase for the bank.
It’s not self-interest, nor is it CD’s design, that keeps depositors from gaining better returns – passing in turn the cost increase to their bank – it’s know-how.
Dig deeper:
- From Hot Money to Core Deposits: How Serving Savers Can Power CDs
- Timing, the New Way to Outcompete Peers for Time Deposits
- Five Ways to Strengthen Your Institution for a Falling Rate Cycle
Will Depositors Game CDs Using AI?
Before AI, it was difficult for deposit markets to behave like stock or bond markets. The limiting factor, of course, was not institutions, which have utilized U.S. Treasuries for a long time. It was depositors, because most have little or no experience with financial math, and generally have low comfort with calculating present value for their assets, even though they still buy bonds and purchase stocks without a complete understanding.
Anyone who uses Google now gets advice from AI, and that changes the deposit picture dramatically when know-how is the only barrier to better depositor returns.
Having read a news headline on interest rates, a depositor can easily obtain guidance on what to do about it. They only need to enter the question into Google or ChatGPT; the latter even guides them when they don’t have all the information.
For example, if they ask: "I have a $100,000 CD earning 2.5% APY scheduled to mature on 7/15/2026, and it has an early withdrawal penalty of 90 days’ interest. If I can reinvest the funds after paying the penalty, should I take the penalty and reinvest?"
Here’s how ChatGPT answers:
- Current rates (as of May 2025) for 12–15 month CDs are often in the 4–5% APY range.
- Conclusion: If you’re eligible for a CD at ≥3% APY, it’s almost certainly profitable to pay the penalty and reinvest.
Would you like help finding a current list of competitive CD rates to act on this?
This is just one small portion of the guidance provided by ChatGPT, and 3.07% is actually the true break-even. However, it still frames CD options in ways that motivate a decision by making it easier to compare. It even offers to help the depositor find CDs that fit its framework for a higher return.
Profits at Risk During the End of Ignorance and Passivity
Banks and credit unions should not see the rise of AI as an Armageddon for the profitability of the deposit business. But the rise of AI does require executives and owners to confront a facet of the funding marketplace that has gone unaddressed:
Any depositor can call Fidelity, Charles Schwab, or Edward Jones for a better deal than their bank offers. The income treasuries generate is tax-exempt, and depositors don’t need a Treasury Direct account, nor must they understand bond investing. Someone just does it for them.
Some people were already getting around their lack of know-how by taking their money out of the banking industry. Those left buying banks’ and credit unions’ certificates at rates below 3% are the "sleeping depositors," who could be nudged awake at any moment by a Google search.
These facets are the reason executives already try avoiding CDs as a safe harbor, and now prefer high-yield savings accounts and money markets. By teasing depositors in with high HYSA rates, the institution gets deposit volumes for which it can change the price when rates decline. Unfortunately, the volumes obtained this way may be even hotter hot money than CDs because of technologies, such as MaxMyInterest, that automatically moves deposits between institutions.
With AI and software allowing this kind of yield harvesting, what was a convenient design for CDs and for HYSAs before calculators is now, in the age of AI, like running a casino with no modern rules about card counting.
Can you imagine being the casino owner, looking across his floor at blackjack tables full of people using AI on their mobile phones to tell them when to hit and when to stay? If the owner said, "Well, we’ve always done it this way," beating the house would become very easy indeed.
Institutions need funding that cannot be gamed by technology. In today’s landscape, that’s a CD for which early withdrawal is priced according to the damage or benefit it brings to a banking institution.
