ChatGPT Vs Bard: Which Can Better Analyze A Stock?

A few months back, I wrote about six smart ways you can use ChatGPT to analyze a stock. Since then, a more advanced, paid version of ChatGPT (called GPT-4) was released. The developers say it’s more reliable and creative than the previous version, and can better handle nuanced instructions.

Not wanting to be left behind, Google upgraded Bard – a generative AI chatbot meant to be the company’s answer to ChatGPT – and expanded its availability worldwide. So that got me thinking: which of the two now-enhanced chatbots is better capable of helping investors out? Let’s find out by revisiting the six ways to use AI to analyze a stock, and compare the two bots’ responses.

1. Gain a high-level understanding of a company

Prompt 1: “Explain to me Tesla's exact business model.”

Both are pretty good summaries, but I found GPT-4’s answer better and more informative. For example, it tells me that auto sales are Tesla’s primary source of revenue but also mentions other sources that are completely missing from Bard’s response, such as revenue from the sales of services, regulatory credits, and software packages. GPT-4’s response even adds some useful color to these points, explaining what regulatory credits are, for example, and why selling software is a clever aspect of Tesla's business model (it allows the firm to generate additional revenue from its existing customer base).

Prompt 2: “What are Tesla’s economic moats?”

Again, decent answers, but I think GPT-4 has the edge here. While both correctly touch on Tesla’s strong brand and technological leadership, GPT-4’s response discusses economic moats that Bard ignores, such as its vertical integration, software capabilities, and supercharger network. On that last point, GPT-4 again adds some useful information like “this infrastructure is a significant advantage, reducing one of the main pain points for EV owners: range anxiety”. It even touches on potential future economic moats when discussing Tesla’s progress in autonomous driving technology.

Winner: GPT-4 because of its more informative answers.

2. Perform a SWOT analysis

Prompt: “Write me a SWOT analysis on Tesla.”

Both gave detailed answers that had a lot of overlap about the firm’s strengths, weaknesses, opportunities, and threats (SWOT). Personally, I was impressed by both responses, particularly given the fact that the chatbots spat them out in a matter of seconds, whereas an MBA student would likely require several hours to conduct a SWOT analysis.

Interestingly, Bard gave some answers that were more up-to-date – e.g. “the company has been profitable for several years and has a healthy balance sheet” and “the cost of raw materials, such as lithium and cobalt, is rising, which could increase Tesla's production costs”. That’s because Bard’s got access to real-time information whereas GPT-4’s knowledge ends after September 2021.

Winner: Bard inches ahead with its more up-to-date knowledge.

Loser: Those poor MBA students… (joke)

3. Summarize earnings calls

Prompt: “Give me the key takeaways from Tesla’s earnings call. Here’s the URL: https://news.alphastreet.com/tesla-inc-tsla-q1-2023-earnings-call-transcript/”

Unfortunately, this was a painstaking process for GPT-4. Now, the ability to access and read web pages is a feature that’s currently in beta and available only to paying users like me. (And that’s already a disadvantage compared to Bard, which is free and comes with internet access as a standard feature). But it gets worse: after I activated GPT-4’s web browsing beta feature, the AI kept on running into issues when I asked it to give me the key takeaways from Tesla’s earnings call transcript. At times, it just had trouble reading the webpage – even after trying out several URLs from multiple sources.

When it finally did read the page, it took a long time to write a response and it stopped generating an answer halfway through every single time. The GPT-4 reply above is the best I could get after several attempts (notice how it says “based on the information I could extract from the provided link before running into technical issues…” (italics are mine). In terms of the takeaways themselves, they were actually quite good despite not capturing everything said during the call (the technical issues meant that GPT-4 couldn’t analyze the entire transcript).

So how did Bard fare? The chatbot was quick to give me bullet points summarizing Tesla’s performance during the quarter. And at first glance, they seemed pretty decent. But when I started inspecting the actual figures, I realized they were totally wrong. For some reason, Bard was crunching Tesla’s numbers during the first quarter of 2022 despite me providing it with a link to the earnings call transcript for the first quarter of 2023. I tried asking it again, while being more explicit about which quarter I wanted, and I got the same answer. Thinking that it might’ve had some trouble reading the webpage, I tried using several URLs from different sources, but that didn’t fix things either.

And then this happened. Yes, I caught Bard out on its mistake, which it admitted to before giving me another false statement, saying that Tesla hadn’t yet released its Q1 2023 earnings report. (It had been out for a month as of the time of testing/writing).

Winner: While both chatbots had notable shortcomings, GPT-4, with its ability to eventually offer me a few valid points, holds a slight advantage over Bard.

4. Evaluate a company’s ESG credentials

Prompt: “Evaluate Tesla’s ESG credentials.”

Both are pretty good summaries, but GPT-4’s answer had a better overall structure. The chatbot’s analysis looked at Tesla’s environmental, social, and governance (ESG) credentials separately, discussing in detail what the firm is doing well and not very well before providing a neat summary at the end. Bard’s response, on the other hand, had fewer details and was repetitive (the paragraph that begins “Overall, Tesla's ESG credentials are mixed…” is literally repeated twice). It also included some weak points, like “Tesla has a strong commitment to innovation, which could lead to the development of even more sustainable products and services in the future”. Meh…

Winner: GPT-4, for its better-structured and more-detailed analysis.

5. Generate code to backtest buy and sell signals

Prompt: “Write me code to see how Tesla's stock price performs one month after its RSI dips below 30.”

Both chatbots easily performed this task related to Tesla’s share price and its relative stock index (RSI), with both writing similar-looking code in Python. I tested both outputs using Replit, an online tool that allows you to run code live directly in your browser, and both worked as they should. But Bard’s response to my prompt has, in my opinion, one huge advantage. See, it not only wrote the code I requested, but it also ran the code and provided me with the key conclusion I was after. Here’s what it said at the very end of its response: “The results of this code show that Tesla's stock price tends to perform well one month after its RSI dips below 30. The average return over the past 5 years is 15%.” (Italics mine). When I asked GPT-4 to execute the code and tell me the results, it said it didn’t have the ability to do so.

Bard’s response is really convenient because it means I don’t have to run the code myself and worry about any potential bugs that could pop up if I’m trying to backtest more complicated buy and sell signals. Ultimately, as an investor, I’m more interested in the results of the signal than the code itself. And I like that I can ask Bard to directly assess the performance of different buy and sell signals based on technical analysis indicators without even asking it for any code (see the example below).

Winner: Bard, with its ability to simultaneously run the code and provide a summary of the key results.

6. Identify key investment risks

Prompt: “What are the key risks associated with investing in Tesla?”

Pretty good answers, and I found it quite interesting that both mentioned Tesla’s high valuation as a risk. But overall, I found GPT-4’s response better for two reasons. First, it identified twice as many risks. Some of the ones overlooked by Bard are actually quite important, like regulatory and key person risks. Second, it did a better job of explaining the risks in a way that’s more relevant to Tesla. To see what I mean, compare what the two chatbots said regarding technological risks. Bard provided a generic and uninspiring explanation that could apply to any tech firm, whereas GPT-4’s response was firm-specific (“Tesla's future growth is partly dependent on the successful development and deployment of new technologies, such as autonomous driving”).

Winner: GPT-4, for its ability to identify more risks and explain them better.

Overall verdict

Here’s a summary of how the two stacked up across each of the six tasks:

Based on pure numbers, the score is 4-2 for GPT-4, making it the winner. Maybe that’s not surprising considering that the chatbot usually gave clearer and more informative answers. When it came to performing a SWOT analysis, GPT-4 did an impressive job, but Bard came out slightly ahead because of its more up-to-date information. In fact, in all the tests I conducted, I found that Bard's strongest point is its ability to directly evaluate the performance of different buy and sell signals (based on technical analysis indicators).

So if you’re wondering which chatbot you should use to analyze stocks, the best answer is probably: both. Because, why not? After all, Bard is free, and it would take you only a few seconds to copy a prompt from GPT-4 and paste it into Bard to get a different, informed perspective. What’s more, Bard can swiftly scan the internet for current and valuable information. In contrast, GPT-4’s ability to browse the web is not only a paid feature, but also one that’s still very clunky.

What matters more than striving to select the perfect chatbot is embracing the use of AI in the first place. We saw in the examples above how chatbots can help you more easily and efficiently analyze a stock. So while they should never be used in isolation, they can be exceptionally handy tools to make your investing life easier. As the saying goes: work smarter, not harder…