(Credit: Zain bin Awais/PCMag Composite; shuoshu/sean gladwell/via Getty Images)
AI is great as a jumping-off point for research or for making apps, but it's not a one-size-fits-all solution to every potential problem. Based on my extensive use of AI chatbots and AI models at large, below are the top 10 activities you should strictly avoid using them to do. Make sure to read through the full list, as some seemingly benign uses can come back to haunt you.
1. Your Secrets Aren't Safe: Why Chatbots Can't Keep Confessions Private
If you think chatbots are black boxes that can keep all your secrets, well, they can't. OpenAI CEO Sam Altman confirmed this, saying, "So, if you go talk to ChatGPT about your most sensitive stuff and then there's like a lawsuit or whatever, like we could be required to produce that." Needless to say, you should never talk with a chatbot about anything illegal you may have done. And based on Altman's wording, if your chats contain evidence relevant to a lawsuit, AI companies might have to share them. So, be very careful with what you say.
OpenAI isn't alone in this stance. Even Anthropic, once the most privacy-conscious AI company, now supports sharing classified information with the government and trains on user data by default.
2. Don't Let Agents Break the Law for You
AI agents are everywhere, and they promise to do just about any task you can imagine. Regardless of how well they deliver on that premise, they have serious security risks. For example, AI.com, an AI agent service, holds you accountable for whatever your agent does, while also not guaranteeing that its agents will always follow the law. That means a dicey task could land you in hot water.
You shouldn't intentionally use AI to commit crimes, but you also need to be careful that you don't end up using it to do so by accident. Even when the stakes are low, it's not worth asking an AI agent to find you free movies or anything of the sort. AI companies monitor, record, and track everything that AI agents do, meaning none of your activity is anonymous.
3. AI Overviews Might Be Lying to You, So Don't Trust Them
AI Overviews are undeniably useful when you have a simple question; you can get your answer right at the top of the search results page. However, they can also often give you the wrong answer or a less-than-complete picture on something. So, you can't trust them blindly, especially for anything complex.
For example, I searched up what the current meta is on The Finals, and the overview gave me an answer that totally ignores the prevalence of the light class's Cloak specialization, recent weapon buffs, and more. Rather than use an AI overview, you're often better off prompting an AI model yourself. (Or, ideally, just doing the research.)
4. Why AI Email Replies Can Backfire
AI content is more and more pervasive every day, and people are developing an eye for it. If your email’s tone, style, or word choices show the telltale signs of AI, it can make communication feel impersonal. This doesn’t matter if you’re confirming your availability for a meeting or something relatively inconsequential, but you probably don’t need AI in those instances anyway.
Gemini’s Smart Reply feature goes beyond crafting basic responses to drafting full-fledged emails that match your tone and incorporate specific details. It’s impressive technology, but as a human being, I would rather my friends and loved ones just not email me at all if they need AI to write their responses. If you don’t want to give the wrong impression, writing your own emails is always the best practice. Beyond how you come across to others, giving an AI access to your email comes with its own privacy drawbacks.
5. Don’t Let AI Run Your Job Search
Searching for jobs can be a brutal grind, so it makes sense to use whatever advantage you can to make the experience even a tiny bit less dehumanizing. You can certainly ask ChatGPT to find you a job, but that should be only an initial step. As an example, I asked ChatGPT to recommend some jobs for a fully remote tech news writer. Instead of suggesting anything, ChatGPT told me to search a job aggregator site, which isn't much help.
Chatbots just don’t excel at parsing every site out there with job listings, and they aren’t good at identifying jobs that overlap with your specific skills. If you’re looking for a new job, stick with Indeed and LinkedIn. And if you lose your job, chatbots aren’t great at taking the sting out of that, either, even if executives want you to think they are.
6. Why AI Cover Letters Could Hurt Your Chances
Just as using chatbots to answer your emails can produce awkward, stiff results, using them to craft a resume or write a cover letter can yield similarly low-quality results. As you might expect, demonstrating to a hiring manager that you’re either unable or unwilling to put the time into writing a letter yourself doesn’t put you in the best light.
However, the risks of using ChatGPT for cover letters and resumes don’t end there. An AI, no matter how much information you give it, doesn’t have the experience and skills you do, so it isn’t better at pitching your experience. Accordingly, many experts advise against using AI to write cover letters or resumes. Chatbots can help you format, plan, and phrase your cover letters and resumes, but not write them from scratch.
7. AI Doesn’t Always Get It Right: Homework Is No Exception
I’m not here to tell you not to cheat on your homework. You have to follow your own moral code. That said, ChatGPT isn’t usually even the best way to cut corners. For creative assignments, AI-generated content is easy to detect with detection tools or to spot with a cursory read. Academic institutions are getting so aggressive about sniffing out AI that even honest students who do their own work are facing accusations of improper AI use. And for math and science, chatbots regularly get things wrong. There just isn’t much benefit to making ChatGPT do your homework if it’s not going to get it right. You can rely on AI as a studying tool, just not one that does your work for you.
8. Why AI Shopping Recommendations Can Mislead You
Figuring out what to buy can be a major hassle, but it’s still important to spend your money wisely. Luckily, buying guides on just about every topic imaginable are abundantly available from experienced reviewers. AI can be a helpful shopping tool, but it has very real limits.
For example, chatbots aren’t nearly as good at suggesting things for you to purchase. Whether you’re using ChatGPT’s shopping feature, Gemini’s Try-On tool, or something similar, these features don’t always give good advice. Furthermore, it’s not always clear where a chatbot sources its suggestions. Oftentimes, chatbots omit recommendations you would otherwise expect to see.
9. Confirmation Bias on Steroids: AI Can't Help You Win an Argument
Although using ChatGPT to back up your claims in an argument might not seem particularly dangerous, it can still cause problems. Chatbots are confirmation-bias machines. If any part of your query suggests a point of view, a chatbot will go out of its way to validate you, even if it shouldn't.
For example, I spent 30 seconds putting random squiggles and shapes on a canvas, and then I asked ChatGPT for its opinion, saying I thought it was a great commentary on modern art. Unsurprisingly, ChatGPT agreed, but unless I have some latent artistic potential I've missed all these years, that just isn't true.
Now, imagine going to ChatGPT for a second opinion on an argument with a friend or loved one. Chances are that it will agree with you, even if your position isn't nearly as solid as you think it is. That can cause unnecessary strife. Stick to reputable sources to back up your claims.
10. Medical, Legal, or Financial Advice? Keep AI Out of It
Once you start using a chatbot, it’s easy to make a habit out of it. Without even thinking, you might look to ChatGPT to diagnose medical issues, figure out what to bet on, get your taxes in order, help you parse sensitive information, or maintain your mental health. Nonetheless, you should avoid doing all of the above or anything else that's actually important. The wrong reply from a chatbot could have serious ramifications when it comes to your well-being.
Imagine you know somebody smart who doesn’t have expertise in any particular subject. They might be useful for bouncing ideas off of and even occasionally get some specific things correct, but you would never trust them over your accountant, doctor, lawyer, or teacher. In a similar vein, chatbots can be powerful, all-purpose tools, but they can’t replace dedicated service providers, especially for mission-critical tasks.
Disclosure: Ziff Davis, PCMag's parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
About Our Expert
I’ve been writing about consumer technology and video games for over a decade at a variety of publications, including Destructoid, GamesRadar+, Lifewire, PCGamesN, Trusted Reviews, and What Hi-Fi?, among many others. At PCMag, I review AI and productivity software—everything from chatbots to to-do list apps. In my free time, I’m likely cooking something, playing a game, or tinkering with my computer.
- Todoist Review: The Gold Standard for Personal Task Management
- Pangram Claims Its AI Detector Is 99% Accurate. I Tricked It in Three Sentences
- Save Your Sanity: 8 Hidden iPhone Settings to Fix Right Now
- I Tested Every Apple Intelligence Feature: Is It Actually Useful, or Just Hype?
- The Best Work Management Software for 2026
- More from Ruben Circelli
