I’m not an Artificial snob. I regularly use ChatGPT, and I’ve written reports about how to create better causes, when to use voice style, and how I nearly won my NCAA March Madness frame using picks from the AI chatbot.

Whether you’re a novice or an expert, I know its limits, and you should also. You don’t want to provide ChatGPT carte blanche, though, which is enjoyable for trying out new dishes, learning a foreign language, or taking a vacation. It’s not very good at all. In truth, it has a lot of potential that is incredibly dubious.

ChatGPT occasionally  hallucinates and passes information off as point, and it may not always had accurate data. Even when it’s absolutely wrong, it’s extremely confident. ( The same can be said, of course, about other generative AI tools. )

The higher the margins are, such as when  taxes, medical expenses, court times, or bank amounts enter the discussion. Below are 11 cases where you should throw down the AI and choose another option if you’re confused about when turning to ChatGPT may be difficult. Use ChatGPT for none of the following.

( Disclosure: In April, CNET’s parent company, Ziff Davis, sued ChatGPT maker OpenAI, alleging that it violated Ziff Davis ‘ copyrights when it trained and ran its AI systems. )

1. diagnosing physical health problems

I’ve undoubtedly given ChatGPT my signs out of interest, but the responses you publish like your worst nightmare. As you look through possible treatments, you might switch from the virus and thirst to a particular type of cancer. I entered a tumor on my neck into ChatGPT with that information. It informed me that I may have cancers, lo and behold. Amazing! In truth, I have a lesion that is non-cancerous and affects one in every 1,000 people. That was explained to me by my qualified physician.

I’m not saying there aren’t some beneficial health applications for ChatGPT&nbsp: it can help you write questions for your next session, convert medical terminology, and create a symptom schedule so you can arrive prepared. And that might reduce the number of physician sessions. AI doesn’t, however, get labs or have you analyzed, and it is undoubtedly not covered by malpractice insurance. Understand its limitations.

2. Taking care of your emotional well-being

Sure, ChatGPT may provide wiring methods, but it didn’t pick up the phone when you’re really struggling with your mental heath. Some people, I’m aware, alternative therapy with ChatGPT. As long as she kept its limitations in thinking, CNET’s Corin Cesaric found it to be somewhat useful for working through grief. However, as someone who has a very real, really human doctor, I can tell you that ChatGPT is still largely just a pale copy and very dangerous at best.

ChatpGPT has no capacity for true empathy, can’t read your body language or tone, and has no living experience. It can only act as a simulation. A qualified therapist adheres to legal guidelines and professional standards to safeguard your rights. ChatGPT is not. Its advice you fail, overlook warning signs, or unwittingly reinforce biases that are embedded in its training data. Leave the more difficult, noisy, human work to a qualified individual with the necessary training to handle it. Choose contact 988 in the US or your nearby line if you or someone you love is in a crisis.

3. making decisions about security right away

Please don’t start ChatGPT and ask for it if you’re in real danger if your carbon monoxide alarms starts chiming. I’d come outside first, then I’d question issues. Big language models are unable to release emergency personnel or smell gas. Every minute you spend typing is a subsequent you’re no evacuating or dialing 911 in a problems. ChatGPT can only function with the bits of information you give it, and it might be too little or too soon in an evacuation. Therefore, treat your robot as a post-accident explainer, not a primary responder.

4. obtaining personal financial or tax advice

ChatGPT you explain what an ETF is, but it is unable to identify your state tax category, filing status, assumptions, retirement objectives, or risk appetite. Since its coaching information may not be accurate for the current tax year or the most recent rate increases, its guidance may be outdated when you press enter.

Friends of mine deposit their 1099 numbers into ChatGPT for a DIY gain. Simply put, the robot can’t change a CPA who is emblem a mistake that may cost you hundreds by uncovering a secret calculation worth a few hundred dollars. Visit a specialist, not AI, when true money, filing deadlines, and IRS penalties are in question. Be aware that any information you give an AI bot will likely be included in its training information, including your salary, Social Security number, and bank routing information.

5. dealing with restricted or confidential data

As a technical journalist, I get news of embargoes every day in my inbox, but I’ve always thought to toss any of these media releases into ChatGPT to get a summary or explanation. That’s because, if I did, the word would leave my handle and travel to a third-party client outside of my nondisclosure agreement.

The same danger exists for consumer deals, health figures, or anything else that is covered by the GDPR, HIPAA, the California Consumer Privacy Act, or simple old trade-secret rules. It applies to your birth certificate, driver’s license, and passport. You can’t guarantee where sensitive information is stored once it’s in the prompt window, who can review it internally, or whether it will be used to train new models. Additionally, ChatGPT is not immune to security threats and hackers. Don’t copy it into ChatGPT if you wouldn’t want to copy it into a public Slack channel.

6. committing any illegal behavior

This one is self-explanatory.

7. cheating at school

If I said I’ve never cheated on my exams, I’d be lying. I didn’t take pride in performing a stunt in high school when I sneaked a peek at a few cumbersome equations in AP calculus using my first-generation iPod Touch. However, with AI, modern cheating’s scale comes off as remarkably tame.

Turnitin and other detectors are improving every semester at recognizing AI-generated prose, and professors are already hearing the” ChatGPT voice” from a distance ( thanks for ruining my beloved em dash ). Real risks include suspension, expulsion, and losing your license. Use ChatGPT as a study companion, not as a ghostwriter, as it is recommended. If you let ChatGPT do the work for you, you’re also cheating yourself out of money on your education.

8. Providing updates and breaking news

Since OpenAI released ChatGPT Search in late 2024 and made it available to everyone in February 2025, the chatbot can instantly retrieve new web pages, stock quotes, gas prices, sports scores, and other real-time data, complete with clickable citations to help you verify the source. However, it won’t continue to stream updates on its own. When speed is important, live data feeds, official press releases, news sites, push alerts, and streaming coverage are still your best bet when it comes to speed.

9. Gambling

I actually had luck with ChatGPT and a three-way parlay while competing in the NCAA men’s basketball championship, but I wouldn’t suggest it to anyone. I’ve seen ChatGPT hallucinate and provide inaccurate data regarding player statistics, misreported injuries, and record wins and losses. I only cashed out because I double-checked every claim before making a wager in real-time, which was even more fortunate. Don’t rely on ChatGPT to get you that win because they can’t see tomorrow’s box score.

10. creating a will or other legally binding contract

ChatGPT is fantastic for presenting the most fundamental ideas. Ask away if you want to learn more about a revocable living trust. You’re rolling the dice, though, the moment you request that it draft the actual legal text. State and county-specific rules for estate and family law apply, so skipping a witness ‘ signature or omitting the notarization provision could have an impact on your entire document. Let ChatGPT assist you in creating a checklist of questions for your attorney, and pay that attorney to turn that checklist into a legal document.

11. creating art

Although this isn’t an objective truth, it’s just my own opinion, I don’t think AI should be employed to produce art. By no means am I opposed to artistic intelligence. I use ChatGPT to brainstorm new ideas and provide headline writing assistance, but it’s only for a brief period of time. Use ChatGPT by all means, but please don’t use it to create works of art that you later pass off as your own. It’s a little gross.