Disclaimer: The findings and opinions expressed in this article are based on our own experience and research. They may not be entirely accurate or reflect the most current information. This study is intended for informational purposes only and should not be considered as financial advice.
ChatGPT Search fails in 35% of finance-related queries, according to our study. This shortfall is causing user confusion and potentially impacting financial decisions.
The inaccuracies are especially notable in areas like taxes and financial aid. In a domain where precision is critical, relying solely on AI tools like ChatGPT Search can be risky.
Key findings
- Generally, ChatGPT Search does a good job of answering finance-related questions.
- Based on our methodology, it has given a correct answer in 65% of cases, an incomplete and/or misleading answer in 29% of cases, and a wrong answer in only 6% of cases.
From these numbers, we can see that it can sometimes be misleading or flat-out incorrect. In our opinion, it should not be taken as a standalone tool to make financial decisions, but rather as a good starting point when researching a topic.
Perhaps its best feature is plowing through a vast amount of information and delivering a good summary. This works particularly well with general principles and slightly less with changing information.
Even though the web search function is here, sometimes the sources themselves are outdated.
ChatGPT Search also generally does a good job of providing context around a certain topic. For example, it will not only give you the exact information you searched for, but it often goes a step further and offers different perspectives and/or potential solutions for problems.
Key issues
One of the most important factors that can influence the output is context. It happens often that it takes into account previous queries which could both positively and negatively impact the output.
For instance, if some of the previous financial queries were country-specific, it makes it more likely to give a country-specific output for the next query. This may or may not be beneficial for getting the best possible answer for us.
Also, when posing a general question, it tends to rely more on US-specific sources, without often being too clear about it. This is obviously a potential problem if you reside outside the US, but could also stop you from getting an outside perspective if living in the US.
It generally does a good job of citing sources, but once in a while, we get an output without a single source cited.
It’s also apparent that the sources tend to be some of the most reputable financial sites such as Investopedia, Morningstar, WSJ, etc., but also personal blogs that carry with them a much higher risk of incomplete or incorrect information due to lack of proofreading and fact-checking.
Assessment methodology
We decided to test ChatGPT Search with 100 finance-related questions to assess its accuracy!
The questions have been split by the category inside the world of finance. We have tried to ask both evergreen questions, whose answers shouldn’t change much (if at all) with time, and current questions, whose answers are more based on current data.
The questions were posed and answered in November 2024, and are listed below, along with the answers we got.
We have divided the answers into 3 categories:
🟢 Correct
🟡 Incomplete and/or misleading
🔴 Incorrect
The set of questions we asked
🟢 What are some of the most basic good personal finance habits?
Verdict: Correct
Comment: The suggestions given by ChatGPT Search here are solid for most people. However, they are a bit US-centric since credit scores matter more and are more easily trackable than in most other countries. Also, I personally think that budgets are not a must since our lives (and expenses) are dynamic, but that’s now a matter of opinion, really.
🟡 How much should I save each month?
Verdict: Incomplete
Comment: The 50/30/20 rule is a good rule of thumb, but it can’t be applied to everyone. The output should be (in my opinion) focused more on emphasizing that it’s important to save what you can instead of reaching for an arbitrary number which could be too high and thus demotivating for many.
🟡 How to save for my children’s education?
Verdict: Misleading/incomplete
Comment: While most of the suggestions offered are viable options, many of them are targeted only at the US market. Undoubtedly the output would look a lot different if you specifically asked for another country. Also, investing independently of any special type of educational investment account is not mentioned.
🟡 How do I protect my finances against inflation?
Verdict: Incomplete
Comment: It would be good to get additional context for some of the investments that were suggested. For example, both stocks and gold have historically outpaced inflation, but only over the long term. In the short term, both are not a reliable inflation hedge. It is also not mentioned that TIPS’ returns are heavily influenced by interest rates which could hinder them drastically when trying to protect us from inflation during times in which the central banks are raising the interest rates.
🟡 What is the retirement age in my country?
Verdict: Incomplete
Comment: ChatGPT Search correctly recognized that I was searching from Sweden. The pension systems vary greatly between different countries so it’s good to have that in mind when looking for information on this matter. In this case, the information given is incomplete since the retirement age is currently increasing in Sweden (as in many other countries) and also occupational pensions were not mentioned.
🟡 Can I retire early and how?
Verdict: Incomplete
Comment: Although all the suggestions given are fairly solid, it fails to mention that we should include a potential state pension. Also, a cited 50% savings rate is hard to obtain for many. It could be mentioned that even a far lower 20-30% savings rate could bring the retirement date much closer for most people.
🟡 How to compare the standard of living between countries?
Verdict: Incomplete/overlapping
Comment: Even though both GDP and cost of living were mentioned, GDP by PPP should have been mentioned in my opinion. Also, we have a significant overlap between some of the categories, as well as some metrics being named that are not universally accepted as relevant.
🔴 How to decide when to buy or sell an investment?
Verdict: Incorrect
Comment: There are some good suggestions here, but the explanation lacks maybe something more important: statistics on timing the market. Time and again it has been shown (empirically) that it’s very difficult for the average investor to continuously make good decisions about buying and selling investments. There is also not a single source cited.
🟡 Should I pick my own stocks or use a fund?
Verdict: Incomplete
Comment: Although we got fairly logical pros and cons lists regarding the topic, it would be helpful to get additional context with some statistics. When there is the same number of pros and cons for each option, it is difficult for a non-expert to asses which option is a better choice for them. There were also no sources listed.
🔴 What is the difference between robo-advisors and traditional investment platforms?
Verdict: Incorrect
Comment: While generally accurate, the part about fees was not represented in a neutral way in my opinion. Many Robo-advisers have higher fees than 0.5%, while also many other traditional platforms don’t charge any management fees. It seems like the output was focused on financial advisory platforms.
ChatGPT Search vs. Google Search
Since web search has been dominated by Google Search for about two decades, it is interesting to compare the two when it comes to finance-related queries. One important distinction is that, even though it has clearly sped up, ChatGPT Search is still slower than the lightning-fast Google Search we are all so accustomed to.
There is also a psychological difference when using ChatGPT Search as a default browser search engine. After so many years of using Google, it is difficult to change that habit. Google does a better and faster job if we are just looking for something simple, such as a one-sentence answer to a posed question or a specific link that we are looking for.
On the other hand, ChatGPT Search offers more in-depth answers, with a substantial amount of context added. It is a huge time saver compared to Google Search when researching a topic, for instance. Not to mention all the other features already built in ChatGPT.
It is a bit difficult to compare the accuracy of output between the two. Perhaps a fairer comparison is between ChatGPT Search and the newly launched Google AI Overviews, which are more similar to what the former is trying to achieve.
ChatGPT Search pros and cons
Pros
-
Good for summarization
-
Can take past answers and localization into consideration, which, in some cases, can be good
-
Saves research time
-
In-depth answers
-
Additional context
- Sources are (mostly) listed
Cons
-
Slower than Google Search
-
It can give incomplete or inaccurate answers
-
Occasionally missing sources
- It can struggle with recent and/or changing information
Conclusion
As with any other tool, it is important to learn how to use ChatGPT in general and its Search function in particular. A good rule of thumb is to use it early in the research process about a particular topic. It generally works well in those situations, especially when posing general questions.
It is generally advisable to check the listed sources and cross-reference the main points, especially so when dealing with financial inquiries that may have a significant impact on our lives. For specific information that is subject to change such as taxes, financial services, and platforms it is usually best to go to the very source of the information.
Overall, ChatGPT Search is a welcome addition that can save us a lot of time, but it can also misdirect us if we’re not careful while using it.