May 4, 2023

10 ChatGPT mistakes you're making and how to fix them

Since ChatGPT is a new technology with new techniques to learn, there are a lot of common mistakes that people make with it. If you're struggling to make it work for you, you might be making some of those mistakes too.
May 4, 2023

10 ChatGPT mistakes you're making and how to fix them

Since ChatGPT is a new technology with new techniques to learn, there are a lot of common mistakes that people make with it. If you're struggling to make it work for you, you might be making some of those mistakes too.
May 4, 2023
Briana Brownell
In this article
Start editing audio & video
This makes the editing process so much faster. I wish I knew about Descript a year ago.
Matt D., Copywriter
Sign up

What type of content do you primarily create?

Videos
Podcasts
Social media clips
Transcriptions
Start editing audio & video
This makes the editing process so much faster. I wish I knew about Descript a year ago.
Matt D., Copywriter
Sign up

What type of content do you primarily create?

Videos
Podcasts
Social media clips
Transcriptions

Open AI’s ChatGPT is quickly becoming a major part of people’s lives — it's definitely a big part of mine. But whenever someone tells me that they tried using ChatGPT and didn’t see the appeal, I usually follow up by asking them what they've tried using it for. Inevitably, their answer reveals why they aren’t getting the results they want.

Since this chatbot is a relatively new AI technology with new techniques to learn, there are a lot of common mistakes that people make with it. If you're struggling to make it work for you, you might be making some of those mistakes too.

ChatGPT mistake #1: You're trying to get facts

The most common mistake I see people making when they're new to ChatGPT is asking it to generate factual information. One person told me they were using ChatGPT for book suggestions — and gave up on ChatGPT when they couldn't find any of them on Amazon.

They shouldn't have been surprised by ChatGPT's failure at the task, though. While the artificial intelligence can provide some interesting ideas, ChatGPT shouldn't be your source for factual information. It's an AI large language model (LLM), not an encyclopedia. Its knowledge is limited to its training data, which might be outdated or incorrect, and it sometimes suffers from hallucinations where it fabricates completely fictional information and presents it with great conviction — like making up titles and authors of books.

Instead, use ChatGPT for brainstorming and inspiration, and rely on traditional research methods for factual information. For instance, if you're looking for the most recent bestsellers or critically acclaimed books in a particular genre, it's better to search for reliable sources like book reviews, award lists, or curated recommendations from trusted experts.

Try this instead: Use ChatGPT for ideas, inspiration, and remixing existing information. For facts, you'll still need to research the old fashioned way.

ChatGPT mistake #2: You're giving up too soon

I've seen people get frustrated with ChatGPT after just one or two attempts, but it's important to remember that prompt engineering is an art that takes practice

If you're not getting the results you want, don't assume your task can't be done. Instead, try rewording your prompt, simplifying the language, or adding more context to help guide how ChatGPT responds. You might be surprised by how much a small tweak to your prompt can change the quality of the generated content. 

If you're stumped, you can often find good example ChatGPT prompts that others have used and tailor those to your needs. Or, ask ChatGPT itself for help! Turns out, it does a fairly good job of coaching you if you ask it to.

Try this instead: Reword your prompt. You can try simplifying the language or adding more context.

ChatGPT mistake #3: You're using the generic tone

ChatGPT often sounds like, well, ChatGPT. Its default tone might not be what you're looking for, especially if you're aiming for a more distinct or personalized style. However, the good news is that ChatGPT can be guided to adopt different tones and styles based on your input.

Instead of settling for the default tone, try specifying the tone or describe a character you want ChatGPT to emulate. For the tone, you can use adjectives like "playful," "formal," or "casual," or better yet, give it some examples to work from. You can also ask ChatGPT to write like someone else, either a character you describe or the name of someone ChatGPT would know from its training set. I've even used my bio and writing to try to get ChatGPT to write more like me. 

You'll still need to edit the results, but by investing time up front you can cut down on the editing time required to add your distinctive tone to the output.

Try this instead: Describe the tone you want, either by using adjectives or by specifying a character that ChatGPT should emulate — maybe even yourself.

image
You can ask ChatGPT to regenerate its response in any tone you want.

ChatGPT mistake #4: You haven't developed your point of view

One common complaint is that ChatGPT's output is kind of dull. In order to make it more insightful, you need to give ChatGPT specifics on what you want it to write. In other words: it needs a point of view

My workflow looks something like this: I start by developing my own outline (sans AI) for the content I want to create. Then I spend some time brainstorming with ChatGPT about things that I could add. I evaluate the ideas and decide whether they’re compelling enough to include in the outline. I've found this method works a lot better than trying to ask ChatGPT to write an outline first and then having it edit that one to include the things I want.

That's actually the method I used for this article. ChatGPT's suggestions were much different than my ideas — no surprise since my outline came from actual conversations I've had with people. But I did end up including one of its suggestions that I hadn't thought of (Mistake #7: Your session is too long — ironic, since I'd experienced it myself.)

Try this instead: First, use your squishy human brain instead of a mechanical one. That is, you'll need to decide on the main points you want to make. Only then should you start brainstorming with ChatGPT to think your points through and to get suggestions for things you might not have considered.

ChatGPT mistake #5: You didn't give enough context

It's tempting to dash off a short prompt and see what comes out, but generally, the more information you can provide to ChatGPT, the better. The time you take on the front end to think through what you’re really asking for will pay off with a better response.

When crafting your prompt, aim to offer as much information as possible about the context. This could include background details, specific examples, or stylistic preferences. That way, by providing a richer context, you're equipping ChatGPT with the necessary tools to tailor its response to your specific needs. At the end of the day, a well-crafted prompt is the foundation of a successful ChatGPT interaction — and making one often takes some time to get right.

Try this instead: Offer as much relevant information as possible when crafting your prompt to help ChatGPT understand your requirements.

ChatGPT mistake #6: You mix your topics or tasks

I often come to ChatGPT with a handful of tasks I want to get done, like drafting an email or two alongside generating some ideas for what to make for dinner this week. Although it's tempting to prompt all of those tasks in one chat window, having multiple tasks in the same thread can make the quality of the generated results worse. 

ChatGPT remembers the context from your previous prompts, so it's better to separate the sessions by topic as much as possible to maintain focus. As an added bonus, that makes it easier to keep your generated content organized because it’s already separated into different threads.

Try this instead: Separate your sessions by topic or task to get more focused and accurate responses.

ChatGPT mistake #7: Your session is too long

When I first started using ChatGPT, I found myself caught in lengthy sessions, partially because I was new to chat-based prompting and I wasn't getting the output I wanted. My beginner-level prompting caused more problems than I realized. Eventually ChatGPT would forget previous information that I provided and I had to re-insert it into the prompt, leading to worsening results.

Instead, it would have been better if I had restarted the conversation when things got muddy. It's better to focus on one aspect of your project at a time and just start a new session when things get lengthy or when experimenting with a new prompt style. This ensures that ChatGPT maintains the necessary context and provides accurate results.

Try this instead: Break your session into smaller components and use separate chat windows for each part of your project. Restart in a new window if your prompts start to give you bad results or if you have to repeat your prompt too many times.

ChatGPT mistake #8: You take the first output

I'm often surprised at how quickly some people give up on ChatGPT. They might try one or two things and then give up, believing that the AI just can't provide what they're looking for. Most of the time, you’ll need to put in some effort, maybe by adjusting the results with a targeted follow-up prompt, suggesting changes and edits to the output, or asking ChatGPT to refine its first response based on additional parameters. 

Instead of settling for the first output, try regenerating the response or prompting ChatGPT to tweak the previous output to systematically improve the results. 

Remember that ChatGPT's output has a random element, so if you think the prompt is good, you can simply ask ChatGPT to generate a different response. This gives the AI another chance to come up with something closer to what you’re looking for.

Try this instead: Regenerate the response or ask for tweaks to the output to improve the results.

ChatGPT mistake #9: You're being too polite

When I was debugging code with ChatGPT, I encountered a situation where I could see the error in the code it generated, but ChatGPT struggled to identify the issue. When prompting to get ChatGPT to fix the code, I worded my request the way I would talk to a junior developer: "Do you think the line where we're defining the variable could be an issue?" To my surprise, ChatGPT said no! It then tried to fix a different line (for the fourth time). Spoiler: The original line was the issue.

The problem was that I defaulted to how I would coach a person — by providing suggestions of where to look for the issue rather than directly telling them what to do. A colleague would catch what my polite wording really meant (the line I identified is the issue!), but the nuance was lost on ChatGPT.

Instead, try being clear and direct about what you want ChatGPT to do. Remember that the AI is a tool, not a person, and it doesn't have feelings that can be hurt by direct language. (Just don't get too used to it and start being rude to your colleagues.)

Try this instead: Be clear and direct about what you want ChatGPT to do without worrying about manners.

ChatGPT mistake #10: You're not checking the answers

It's tempting to save time by copying and pasting whatever ChatGPT generates for you, but it's also a super-fast way to make major mistakes. Remember, ChatGPT is an incredibly powerful tool, but it's not infallible. It can sometimes provide outdated, incorrect, irrelevant, or even nonsensical information.

Before using the generated content, take a moment to review and verify the answers or suggestions provided by ChatGPT. Cross-checking the information with reliable sources and using your own judgment will help you avoid errors and make sure the content is accurate and relevant. It's an extra step, but it lets you use ChatGPT's output confidently, knowing that it meets your expectations and it’s free from any glaring mistakes. A little vigilance goes a long way in getting the most out of your AI assistant.

Try this instead: Always double-check the generated content for accuracy and relevance before using it in your work.

Conclusion

Mastering ChatGPT is all about patience, persistence, and playing around with your prompts. It's a powerful tool, but it's not perfect, and it's up to you to fine-tune its output. 

Just remember to give enough context, be direct, separate sessions by topic, and double-check the results. By doing this, you'll greatly enhance the quality of ChatGPT's responses and, ultimately, your own work. 

Briana Brownell
Briana Brownell is a Canadian data scientist and multidisciplinary creator who writes about the intersection of technology and creativity.
Share this article
Start creating—for free
Sign up
Join millions of others creating with Descript

10 ChatGPT mistakes you're making and how to fix them

Podcast microphone bent over a huge keyboard with an even huger computer monitor looming over it, emblazoned with the ChatGPT logo

Open AI’s ChatGPT is quickly becoming a major part of people’s lives — it's definitely a big part of mine. But whenever someone tells me that they tried using ChatGPT and didn’t see the appeal, I usually follow up by asking them what they've tried using it for. Inevitably, their answer reveals why they aren’t getting the results they want.

Since this chatbot is a relatively new AI technology with new techniques to learn, there are a lot of common mistakes that people make with it. If you're struggling to make it work for you, you might be making some of those mistakes too.

Remove all your “ums” and “uhs” with a click, correct your voiceover by typing, and get studio-quality sound wherever you record. Check out our
Tools that work for creators.

ChatGPT mistake #1: You're trying to get facts

The most common mistake I see people making when they're new to ChatGPT is asking it to generate factual information. One person told me they were using ChatGPT for book suggestions — and gave up on ChatGPT when they couldn't find any of them on Amazon.

They shouldn't have been surprised by ChatGPT's failure at the task, though. While the artificial intelligence can provide some interesting ideas, ChatGPT shouldn't be your source for factual information. It's an AI large language model (LLM), not an encyclopedia. Its knowledge is limited to its training data, which might be outdated or incorrect, and it sometimes suffers from hallucinations where it fabricates completely fictional information and presents it with great conviction — like making up titles and authors of books.

Instead, use ChatGPT for brainstorming and inspiration, and rely on traditional research methods for factual information. For instance, if you're looking for the most recent bestsellers or critically acclaimed books in a particular genre, it's better to search for reliable sources like book reviews, award lists, or curated recommendations from trusted experts.

Try this instead: Use ChatGPT for ideas, inspiration, and remixing existing information. For facts, you'll still need to research the old fashioned way.

ChatGPT mistake #2: You're giving up too soon

I've seen people get frustrated with ChatGPT after just one or two attempts, but it's important to remember that prompt engineering is an art that takes practice

If you're not getting the results you want, don't assume your task can't be done. Instead, try rewording your prompt, simplifying the language, or adding more context to help guide how ChatGPT responds. You might be surprised by how much a small tweak to your prompt can change the quality of the generated content. 

If you're stumped, you can often find good example ChatGPT prompts that others have used and tailor those to your needs. Or, ask ChatGPT itself for help! Turns out, it does a fairly good job of coaching you if you ask it to.

Try this instead: Reword your prompt. You can try simplifying the language or adding more context.

ChatGPT mistake #3: You're using the generic tone

ChatGPT often sounds like, well, ChatGPT. Its default tone might not be what you're looking for, especially if you're aiming for a more distinct or personalized style. However, the good news is that ChatGPT can be guided to adopt different tones and styles based on your input.

Instead of settling for the default tone, try specifying the tone or describe a character you want ChatGPT to emulate. For the tone, you can use adjectives like "playful," "formal," or "casual," or better yet, give it some examples to work from. You can also ask ChatGPT to write like someone else, either a character you describe or the name of someone ChatGPT would know from its training set. I've even used my bio and writing to try to get ChatGPT to write more like me. 

You'll still need to edit the results, but by investing time up front you can cut down on the editing time required to add your distinctive tone to the output.

Try this instead: Describe the tone you want, either by using adjectives or by specifying a character that ChatGPT should emulate — maybe even yourself.

image
You can ask ChatGPT to regenerate its response in any tone you want.

ChatGPT mistake #4: You haven't developed your point of view

One common complaint is that ChatGPT's output is kind of dull. In order to make it more insightful, you need to give ChatGPT specifics on what you want it to write. In other words: it needs a point of view

My workflow looks something like this: I start by developing my own outline (sans AI) for the content I want to create. Then I spend some time brainstorming with ChatGPT about things that I could add. I evaluate the ideas and decide whether they’re compelling enough to include in the outline. I've found this method works a lot better than trying to ask ChatGPT to write an outline first and then having it edit that one to include the things I want.

That's actually the method I used for this article. ChatGPT's suggestions were much different than my ideas — no surprise since my outline came from actual conversations I've had with people. But I did end up including one of its suggestions that I hadn't thought of (Mistake #7: Your session is too long — ironic, since I'd experienced it myself.)

Try this instead: First, use your squishy human brain instead of a mechanical one. That is, you'll need to decide on the main points you want to make. Only then should you start brainstorming with ChatGPT to think your points through and to get suggestions for things you might not have considered.

ChatGPT mistake #5: You didn't give enough context

It's tempting to dash off a short prompt and see what comes out, but generally, the more information you can provide to ChatGPT, the better. The time you take on the front end to think through what you’re really asking for will pay off with a better response.

When crafting your prompt, aim to offer as much information as possible about the context. This could include background details, specific examples, or stylistic preferences. That way, by providing a richer context, you're equipping ChatGPT with the necessary tools to tailor its response to your specific needs. At the end of the day, a well-crafted prompt is the foundation of a successful ChatGPT interaction — and making one often takes some time to get right.

Try this instead: Offer as much relevant information as possible when crafting your prompt to help ChatGPT understand your requirements.

ChatGPT mistake #6: You mix your topics or tasks

I often come to ChatGPT with a handful of tasks I want to get done, like drafting an email or two alongside generating some ideas for what to make for dinner this week. Although it's tempting to prompt all of those tasks in one chat window, having multiple tasks in the same thread can make the quality of the generated results worse. 

ChatGPT remembers the context from your previous prompts, so it's better to separate the sessions by topic as much as possible to maintain focus. As an added bonus, that makes it easier to keep your generated content organized because it’s already separated into different threads.

Try this instead: Separate your sessions by topic or task to get more focused and accurate responses.

ChatGPT mistake #7: Your session is too long

When I first started using ChatGPT, I found myself caught in lengthy sessions, partially because I was new to chat-based prompting and I wasn't getting the output I wanted. My beginner-level prompting caused more problems than I realized. Eventually ChatGPT would forget previous information that I provided and I had to re-insert it into the prompt, leading to worsening results.

Instead, it would have been better if I had restarted the conversation when things got muddy. It's better to focus on one aspect of your project at a time and just start a new session when things get lengthy or when experimenting with a new prompt style. This ensures that ChatGPT maintains the necessary context and provides accurate results.

Try this instead: Break your session into smaller components and use separate chat windows for each part of your project. Restart in a new window if your prompts start to give you bad results or if you have to repeat your prompt too many times.

ChatGPT mistake #8: You take the first output

I'm often surprised at how quickly some people give up on ChatGPT. They might try one or two things and then give up, believing that the AI just can't provide what they're looking for. Most of the time, you’ll need to put in some effort, maybe by adjusting the results with a targeted follow-up prompt, suggesting changes and edits to the output, or asking ChatGPT to refine its first response based on additional parameters. 

Instead of settling for the first output, try regenerating the response or prompting ChatGPT to tweak the previous output to systematically improve the results. 

Remember that ChatGPT's output has a random element, so if you think the prompt is good, you can simply ask ChatGPT to generate a different response. This gives the AI another chance to come up with something closer to what you’re looking for.

Try this instead: Regenerate the response or ask for tweaks to the output to improve the results.

ChatGPT mistake #9: You're being too polite

When I was debugging code with ChatGPT, I encountered a situation where I could see the error in the code it generated, but ChatGPT struggled to identify the issue. When prompting to get ChatGPT to fix the code, I worded my request the way I would talk to a junior developer: "Do you think the line where we're defining the variable could be an issue?" To my surprise, ChatGPT said no! It then tried to fix a different line (for the fourth time). Spoiler: The original line was the issue.

The problem was that I defaulted to how I would coach a person — by providing suggestions of where to look for the issue rather than directly telling them what to do. A colleague would catch what my polite wording really meant (the line I identified is the issue!), but the nuance was lost on ChatGPT.

Instead, try being clear and direct about what you want ChatGPT to do. Remember that the AI is a tool, not a person, and it doesn't have feelings that can be hurt by direct language. (Just don't get too used to it and start being rude to your colleagues.)

Try this instead: Be clear and direct about what you want ChatGPT to do without worrying about manners.

ChatGPT mistake #10: You're not checking the answers

It's tempting to save time by copying and pasting whatever ChatGPT generates for you, but it's also a super-fast way to make major mistakes. Remember, ChatGPT is an incredibly powerful tool, but it's not infallible. It can sometimes provide outdated, incorrect, irrelevant, or even nonsensical information.

Before using the generated content, take a moment to review and verify the answers or suggestions provided by ChatGPT. Cross-checking the information with reliable sources and using your own judgment will help you avoid errors and make sure the content is accurate and relevant. It's an extra step, but it lets you use ChatGPT's output confidently, knowing that it meets your expectations and it’s free from any glaring mistakes. A little vigilance goes a long way in getting the most out of your AI assistant.

Try this instead: Always double-check the generated content for accuracy and relevance before using it in your work.

Conclusion

Mastering ChatGPT is all about patience, persistence, and playing around with your prompts. It's a powerful tool, but it's not perfect, and it's up to you to fine-tune its output. 

Just remember to give enough context, be direct, separate sessions by topic, and double-check the results. By doing this, you'll greatly enhance the quality of ChatGPT's responses and, ultimately, your own work. 

Featured articles:

No items found.

Articles you might find interesting

Product Updates

How Descript Uses AI To Match Room Tone For Seamless Edits

Few recordings of human speech are recorded “perfectly” — that is to say, inside an acoustically treated studio, with a high-quality microphone, and without external noise or interference. Most are recorded imperfectly, with varying degrees of background noise and unwanted sound. Maybe there’s an air conditioner in the background or line-level fuzz. This is known as room tone, and when it comes to editing audio, dealing with room tone can be challenging and costly, requiring the use of dedicated plugins and software.

For Business

How to Create an Amazing Product Demo Video

Product demo videos are an excellent opportunity for you, the maker. Your product demonstration video should aim to answer a viewer’s burning question: “What can this product do for me?”

Product Updates

High Fidelity Remote Recording Just Got a Whole Lot Easier

Earlier this week, Zoom released update 5.2.2, which includes a new feature they’re calling “High Fidelity Audio” mode. Peter Kirn over at Create Digital Music covered the update, writing that it would be excellent for musicians. Zoom has always been the easiest way to record remote interviews, but it came with a cost: the audio quality wasn’t as high as a local recording. We wondered how High Fidelity Audio mode would sound for podcasts and remote interviews, so we tested it ourselves — and the results are incredible. Listen for yourself.

Podcasting

Interview recording equipment: The essentials

Editing podcasts becomes easier when you're more comfortable with your equipment. Learn how the right gear will allow you to devote more energy to your content.

Podcasting

Using ambience and sound effects for podcast soundscapes

There are a range of ways to think about and implement sound effects for podcasts. Here's where to find that music and how to use it in a way that benefits your show.

Video

7 tips for making better educational videos

Whether you're teaching physics or summer styling tips, it's easy for educational videos to fall flat. These 7 tips will keep viewers engaged.

Related articles:

Share this article

Get started for free →