ChatGPT for Beginners: Two Prompts Worth Learning First
Most people’s first week with ChatGPT goes the same way. You type in a question, you get back something that sounds confident and says almost nothing useful, and you quietly decide the whole thing is overhyped. I have watched that happen in class more times than I can count. The tool is not the problem. The problem is that ChatGPT for beginners gets taught as “just ask it anything,” which is bad advice.
Asking it anything is how you get a paragraph of filler. What works is boring and specific: tell it who it is, tell it what you want, hand it your real information, and say what shape the answer should come back in. Then teach it your preferences once instead of retyping them in every new chat.
Those are the two tips below. Neither takes more than ten minutes to try, and neither needs a paid plan.
ChatGPT for Beginners, Tip 1: Role, Goal, Data, Format
A prompt that gets a usable answer has four parts, and beginners usually supply only one of them. Here is the whole pattern with a real job attached. Say it is Friday afternoon and you have a page of scratch notes that has to become a status update:
You are a project coordinator writing for a VP who skims. Below are my rough notes from this week. Write a five-bullet status update covering what moved, what slipped, and what needs a decision from her. Use only what is in the notes. If something is missing, write "not stated" rather than filling it in. Notes: [paste your notes]
- Role – “a project coordinator writing for a VP who skims” sets the vocabulary and the length without you having to ask for either.
- Goal – “what moved, what slipped, what needs a decision” is the structure you actually want. Vague requests get vague output.
- Data – your messy notes, pasted in. This is what makes the answer yours instead of generic.
- Format – five bullets. Say the number. “Brief” means nothing to a language model.
What comes back is a draft you edit for two minutes instead of a blank page you stare at for twenty. The awkward sentences are already written and you are only fixing them.
The gotcha: that line about “use only what is in the notes” is the one everybody skips, and it matters most. Left to itself, ChatGPT smooths over a gap with something plausible – a completion date you never wrote down, an owner’s name it inferred. Telling it to write “not stated” turns those gaps into something you can see and fix. I would rather read “target date not stated” than a bullet that confidently invents the 15th.
Worth adding to the end of any prompt where you are not sure you gave enough context: Before you answer, ask me up to three questions you need answered to do this well. It asks, you answer in a sentence each, and the result beats its first guess every time.
Tip 2: Set Your Custom Instructions Once and Stop Repeating Yourself
Almost no beginner finds this on their own, and it changes every conversation afterward. In the browser, click your profile picture in the top right, then Settings, then Personalization, then Custom instructions. You get two boxes.
In the box asking what traits ChatGPT should have, something like this works well:
Answer in plain English. Lead with the answer, then the reasoning. Skip the preamble and the compliments. If you are not confident, say so instead of guessing. When you give me steps, number them and name the exact menu path. Expand and explain more when I ask why.
In the box about what it should know about you, give it the context you would otherwise retype daily:
I am an operations manager at a 60-person manufacturing company. I use Excel, Outlook and Teams daily and I am not a programmer. Most of what I write goes to department heads. Use US spelling and MM/DD/YYYY dates.
Save it, and every new chat starts from there. No more “shorter, please” on the third try, no more re-explaining what your company does. If you use ChatGPT a few times a week, this one screen saves more typing in a month than any clever prompt trick.
The gotcha: custom instructions only apply to new chats, so the conversation open in front of you will not change – start a fresh one to test it. And go easy on “no preamble” – write it too forcefully and you get clipped, telegraphic answers on the days you wanted the explanation, which is why that “expand when I ask why” sentence stays in. Also, whatever you type in those boxes travels into every chat, including ones you share with a colleague, so keep client names and anything confidential out of it.
Plenty more on the calendar
If you would rather watch this done live than read about it, there is a ChatGPT 101 beginners session on Thursday, November 19. It is the ground-up version of the above: getting set up, writing prompts that hold together, custom instructions, uploading a file and asking questions about it, and where it still gets things wrong. Bring a real task and we will work on it live.
The full current schedule lives on the front page of PCWebinars.com, and new dates go up as they are confirmed. Register for this one if the timing works, or browse the others – there is usually something on Excel, Power BI, Copilot or Claude within the next couple of weeks.
About your trainer
I’m Tom Fragale. I have been a computer trainer and consultant for more than 30 years, and I have taught well over 30,000 people how to get more out of the software they already own – Excel, Access, Power BI, Word, Outlook, and now the AI tools that sit on top of all of it: Claude, ChatGPT, and Microsoft Copilot. My background is applied, not theoretical. Alongside teaching I build real databases and reporting systems for clients in Excel, Access, and Power BI, so the examples I bring to class are problems I ran into last week, not textbook exercises. I also publish free tutorials on YouTube and longer courses online.
If this was useful, there is plenty more of it – tutorials, recordings and the live schedule – over at PCWebinars.com.