# 8 ChatGPT Mistakes at Work (and the Fix for Each) > Eight common ChatGPT mistakes at work, from one-line prompts to trusting the first draft and pasting card numbers, each with a simple fix. Source: https://learn.ynteractive.com/content/blog/chatgpt-mistakes-at-work · Updated 2026-10-23 · Free from Stepthrough (https://learn.ynteractive.com) Quick answer The most common ChatGPT mistakes at work are vague one-line [prompts](https://learn.ynteractive.com/content/glossary/prompt), trusting the first draft, letting ChatGPT guess missing details, asking it to "make it better," and pasting private data, and each one has a simple fix. This list is for office staff, owners and anyone who uses ChatGPT for emails and notes, and every mistake comes from a free lesson you can practice in a few minutes, starting with [Prompt Writing 101](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format). Last updated October 23, 2026 [Take the free module →](https://learn.ynteractive.com/courses/ai-at-work/6)Module 6 of the free [AI at Work: Start Here](https://learn.ynteractive.com/courses/ai-at-work) course, about 8 minutes. No account needed. By [Robert Breen](https://learn.ynteractive.com/content/about) · Published October 23, 2026 · Updated October 23, 2026 · 6 min read ChatGPT · Prompts · Email ## Watch: the idea in 0:49 ### ChatGPT mistakes at work Explainer · 0:49 · with captions Eight common ChatGPT mistakes at work, from one-line prompts to trusting the first draft and pasting card numbers, each with a simple fix. **Transcript** Most ChatGPT mistakes at work aren't about ChatGPT. They're habits, and each one has a simple fix. A one-line prompt makes ChatGPT guess everything. You get Dear Valued Customer, emoji and blanks. Instead, write the prompt in four parts. Role, context, task and format. Ask for action items, and it may guess due dates. So add: if an owner or date isn't said, write Not stated. Then you fill in the real dates. Don't ask it to make it better. It may change the parts that were already right. Name the one change, then add a lock. Keep every other fact exactly the same. Have it check its draft against your facts. Leave card numbers and private details out. And read every message yourself before you send it. Practice it step by step, free, in Prompt Writing 101. None of these mistakes are about ChatGPT being bad. They're habits that make good answers less likely or let a bad answer get sent. Here are eight, with the fix for each. ## Mistake 1: Why is a one-line prompt a problem? A one-line prompt makes ChatGPT guess everything you didn't say, so you get something generic. **Fix:** write the prompt in four parts: [Role, Context, Task and Format](https://learn.ynteractive.com/content/glossary/role-context-task-format). In Prompt Writing 101, "Write an email to my customers about our new service." gets back a long email that opens with "Dear Valued Customer," has emoji in the subject line and blanks like [Your Company Name]. The four-part version names the business (Summit Lawn Care, a made-up company), the customers, the service, the $149 price, the dates, one call to action and "under 120 words, friendly and plain, no emoji." The second answer is ready to send. [How to Write Better ChatGPT Prompts](https://learn.ynteractive.com/content/blog/write-better-chatgpt-prompts) explains each part. ## Mistake 2: Why shouldn't you trust ChatGPT's first draft? First drafts can sound careful and still promise something you never said. **Fix:** ask ChatGPT to check its draft against your facts in a table. In the [Reply Faster: Customer Emails](https://learn.ynteractive.com/content/chatgpt-reply-to-customer-emails) lesson, the made-up Cedar Lane Bakery gives ChatGPT a numbered facts list. Fact 4 says the lemon recipe has no nuts, but the kitchen also bakes with nuts. The first draft still calls the cake "completely nut-free." One message catches it: "Check that reply against my facts. Make a table of every promise or claim in it and the fact that backs it up. Flag anything the facts don't support." More in [Check AI Email Drafts for Promises Before You Send](https://learn.ynteractive.com/content/blog/check-ai-drafts-for-promises). ## Mistake 3: Why does ChatGPT invent owners and due dates? If you ask for an action items table without saying what to do about missing details, ChatGPT may fill the gaps with a plausible guess. **Fix:** add "Only use what's in the transcript. If an owner or date isn't said, write 'Not stated'." In [Summarize a Meeting Transcript](https://learn.ynteractive.com/content/chatgpt-summarize-meeting-transcript), that line makes two due dates come back as "Not stated," which is honest: nobody said when the library catering quote or the oven call would happen. You were in the meeting, so you tell ChatGPT the real dates in one sentence and it updates the table. A guessed date in a recap email turns into a missed deadline. [Turn a Meeting Transcript into Action Items](https://learn.ynteractive.com/content/blog/meeting-transcript-to-action-items) covers the full routine. ## Mistake 4: What's wrong with asking ChatGPT to "make it better"? "Make it better" doesn't say what's wrong, so ChatGPT may change the parts that were already right. **Fix:** ask for the specific change and lock the rest. The bakery lesson's fix request: "Use that allergy wording. Then make the whole reply a little warmer and shorter, under 120 words. Keep every other fact exactly the same. Reply with just the email." [How to Revise a ChatGPT Draft](https://learn.ynteractive.com/content/blog/how-to-revise-a-chatgpt-draft) has more examples. ## Try it free, step by step **Prompt Writing 101: Role, Context, Task, Format**. Ask ChatGPT for the same customer email twice, once vaguely and once with a four-part prompt, and see exactly why structure gets a better answer. You do every step yourself in a practice copy of ChatGPT, it checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, and nothing touches your real accounts. [Start the free lesson](https://learn.ynteractive.com/modules/prompt-writing-101) [Or read the full step-by-step guide](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format) ## Mistake 5: What shouldn't you paste into ChatGPT at work? Card numbers, account numbers, ID numbers, health details and anything else the reply doesn't need. **Fix:** remove private details before you paste, and follow your company's policy on AI tools. The Reply Faster lesson says it at the copy step: if there is anything private in the thread, like a card number, leave it out. The meeting lesson adds that you should check your company's rules before sharing a transcript with private details. For work, a business ChatGPT plan that doesn't train on your data by default is the safer choice. [Is It Safe to Paste Client Emails into ChatGPT?](https://learn.ynteractive.com/content/blog/is-it-safe-to-paste-client-data-into-chatgpt) goes industry by industry. ## Mistake 6: Why does ChatGPT answer before you finish pasting? Pressing Enter sends the message, so a request sent before the pasted text gets an answer to half a question. **Fix:** press **Shift+Enter** for a new line, then paste. The meeting lesson has you type the request first, then paste the transcript underneath. In real ChatGPT, press Shift+Enter to start a new line, then Ctrl+V. Plain Enter would send the message. If a file is easier, click **+**, then **Add photos & files**, and attach it instead. ## Mistake 7: Should you keep going in a chat that went wrong? Usually not. A bad answer stays in the conversation and can steer what comes next. **Fix:** click **New chat** and ask again, properly. In Prompt Writing 101, after the vague answer comes back, the next step is **New chat** before the four-part prompt. Starting fresh means the generic email isn't sitting in the conversation. Keep going in the same chat when the earlier context helps, like asking for a follow-up email after a meeting summary. The meeting lesson does exactly that: ChatGPT still has the whole meeting, so you don't paste anything again. ## Mistake 8: Why read a ChatGPT email before you send it? Because you're the one sending it, and ChatGPT doesn't know your business the way you do. **Fix:** one last read of the dates, amounts and hardest answer before you click Send. Every reply lesson ends with that read. In the bakery lesson, you check the date, the price and the allergy answer, then click **Send** in Gmail. The meeting lesson says to always check the owners and dates before the recap goes out. ## How do you avoid these mistakes as a team? Write the fixes down as a few shared rules, and have everyone take the same three short lessons. A simple set of rules: - Use the four-part prompt for anything that goes to a customer. - Give ChatGPT numbered facts, and have it check its draft against them. - Write "Not stated" instead of guessing. - Ask for specific changes, not "make it better." - Leave private numbers and details out. - A person reads every message before it's sent. The three lessons take 8 to 9 minutes each, are free, run in a practice copy and work with any ChatGPT plan, including the free one, within its limits. They're bundled in the [ChatGPT Jump Start path](https://learn.ynteractive.com/content/paths/chatgpt-jump-start). ## Key takeaways - A one-line prompt makes ChatGPT guess; a four-part prompt with Role, Context, Task and Format gets a usable answer. - ChatGPT's first draft can overpromise, so ask it to check every claim against your numbered facts. - Adding "If an owner or date isn't said, write 'Not stated'" stops ChatGPT from inventing details in meeting recaps. - Ask for specific changes and add "Keep every other fact exactly the same" instead of "make it better." - Remove private details before pasting, start a new chat when an answer goes wrong, and read every message before you send it. ## Frequently asked questions **What is the most common ChatGPT mistake at work?** Writing a one-line prompt and expecting a specific answer. Give ChatGPT the role, the context, the task and the format you want. **Why does ChatGPT make things up?** ChatGPT fills gaps with plausible guesses. Give it the facts, tell it to write "Not stated" when something is missing, and check its draft against your facts. **How do I type a new line in ChatGPT without sending?** Press Shift+Enter. Plain Enter sends the message. **When should I start a new chat in ChatGPT?** Start a new chat when an answer went wrong or you're moving to a different task or customer. Stay in the same chat when ChatGPT needs the earlier context. **Is Cedar Lane Bakery a real business?** No. Cedar Lane Bakery, Summit Lawn Care and the bakery team in the meeting lesson are made-up examples. ## Free lessons in this post - [Prompt Writing 101: Role, Context, Task, Format](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format): Ask ChatGPT for the same customer email twice, once vaguely and once with a four-part prompt, and see exactly why structure gets a better answer. - [Reply Faster: Turn a Messy Customer Email Thread into a Clear, Friendly Reply](https://learn.ynteractive.com/content/chatgpt-reply-to-customer-emails): Turn a messy customer email thread into a clear, friendly reply: give ChatGPT the thread and your facts, check every promise, fix the tone, and send it from Gmail. - [Summarize a Meeting Transcript with ChatGPT](https://learn.ynteractive.com/content/chatgpt-summarize-meeting-transcript): Paste a long meeting transcript into ChatGPT and get the decisions, an action items table with owners and due dates, and a follow-up email ready to send. ## Watch it in a free lesson: Prompt Writing 101: Role, Context, Task, Format ### Prompt Writing 101: video walkthrough Full walkthrough · 2:20 · with captions Ask ChatGPT for the same customer email twice, once vaguely and once with a four-part prompt, and see exactly why structure gets a better answer. The whole lesson narrated step by step, recorded in Stepthrough’s simulated practice environment. [Watch it with chapters and the full transcript](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format#watch) From the free lesson [Prompt Writing 101: Role, Context, Task, Format](https://learn.ynteractive.com/content/prompt-writing-101-role-context-task-format) ## Related terms Plain-English definitions, each with an example from a free lesson. [Prompt](https://learn.ynteractive.com/content/glossary/prompt) · [Role, Context, Task, Format](https://learn.ynteractive.com/content/glossary/role-context-task-format) · Call to action (CTA) · [Action items](https://learn.ynteractive.com/content/glossary/action-items) · AI literacy · [AI note-taker](https://learn.ynteractive.com/content/glossary/ai-note-taker) · AI summarization · [Canned response](https://learn.ynteractive.com/content/glossary/canned-response) [The full AI glossary](https://learn.ynteractive.com/content/glossary) ## Related prompts - [Action items table prompt](https://learn.ynteractive.com/content/prompts/action-items-table-prompt): Paste a meeting transcript or rough notes and get a table of every task, owner and due date, with "Not stated" wherever nobody… - [Buyer inquiry reply prompt](https://learn.ynteractive.com/content/prompts/buyer-inquiry-reply-prompt): Reply to a buyer's listing inquiry from your facts sheet: fees, rules and showing times, with sources instead of opinions on… - [Candidate rejection email prompt](https://learn.ynteractive.com/content/prompts/candidate-rejection-email-prompt): Write a short, respectful rejection email that thanks the candidate for the step they reached and promises nothing you won't do… - [Check reply promises prompt](https://learn.ynteractive.com/content/prompts/check-reply-promises-prompt): A ChatGPT prompt that checks a draft before you send it: a table of every promise, the fact behind each, and safer wording for… [All free prompt templates](https://learn.ynteractive.com/content/prompts) ## Visual guides The same ideas as quick visual guides: swipe through the slides, or read each one as text. - [Summarize a meeting transcript with ChatGPT](https://learn.ynteractive.com/content/visual/meeting-recap-steps): A meeting transcript is useful for about five minutes, and then nobody reads it again. - [What is an AI hallucination?](https://learn.ynteractive.com/content/visual/term-hallucination): AI term of the day: hallucination. - [Meeting notes myth: “summarize this” is all ChatGPT needs](https://learn.ynteractive.com/content/visual/summarize-this-myth): Myth: paste a transcript, type “summarize this”, done. Fact: ask for the decisions, an action items table with owners and dates, and open questions. [All visual guides](https://learn.ynteractive.com/content/visual) ## Keep reading - [How to Write Better ChatGPT Prompts: 4 Simple Parts](https://learn.ynteractive.com/content/blog/write-better-chatgpt-prompts): Vague prompts get generic answers full of [Your Company Name]. Role, Context, Task and Format fix that. A real before-and-after, plus a template. - [Is It Better to Start a New Chat in ChatGPT? When and Why](https://learn.ynteractive.com/content/blog/chatgpt-new-chat-vs-same-chat): Yes, start a new chat in ChatGPT when you switch tasks or customers, or after a bad answer. Stay in the same chat to build on the last reply. Examples. [All blog posts](https://learn.ynteractive.com/content/blog) · [All free tutorials](https://learn.ynteractive.com/content)