How to Use ChatGPT Like a Pro
Getting consistently useful work out of ChatGPT comes down to how you write prompts, how you provide context, how you check what comes back, and how you keep a session's useful outcomes from disappearing. Here is a repeatable workflow that makes every session faster and more reliable.
Updated 6 min read
Using ChatGPT well is a skill that improves with practice. The people who get consistent, reliable results share a few habits: they write specific prompts, they give context upfront, they iterate rather than accept the first answer, and they check anything that matters against a real source.
The gap between a good session and a frustrating one is often in how the question was asked, not in what ChatGPT can do. Tools like fenbs extend this by keeping the useful work — decisions, tasks, outstanding questions — in a shared record that any connected assistant can read at the start of the next session.
Write specific prompts
A vague prompt produces a vague answer. The more precisely you describe what you want — the format, the length, the audience, the constraints — the closer the first response will be to what you need.
Compare "write a summary" with "write a three-paragraph summary for a non-technical audience, covering what the decision was, who it affects and what the next step is." Both ask for the same thing, but the second one gives ChatGPT a clear target.
Useful dimensions to specify: audience, format, length, tone, what to include, and what to leave out. You do not need all of these every time. Add the ones that matter for the task in hand.
Give context upfront
ChatGPT does not know your project, your constraints or your history unless you tell it. Starting a new chat and asking a question without context is like asking a colleague who just joined the call to answer before anyone has explained the situation.
Spend one message at the start of each session setting the context: what the project is, what has been decided so far, any constraints, and what you need from this session. This front-loading makes every subsequent message in the session faster and more relevant.
For recurring projects, the context that matters is the same every time. Writing it once and saving it — in ChatGPT Projects, in custom instructions, or in a project record — means you do not have to re-type it each session.
Use ChatGPT Projects for ongoing work
ChatGPT Projects (available on paid plans) attach instructions and files to a set of conversations. Every chat within a project starts with those instructions already loaded, so you do not have to re-establish context each time. Create a project for any recurring body of work with a stable set of rules and references.
Keep the project instructions focused on what genuinely stays the same. A project instruction that covers every possible scenario becomes as long as re-explaining the project manually. Short, stable facts work better than long, conditional ones.
Iterate instead of accepting the first draft
ChatGPT rarely produces the best answer in a single pass for complex tasks. The first response is a starting point. Ask for changes — "make it shorter," "remove the third point," "rewrite this for a more technical audience" — and the next version is usually closer.
Ask it what it left out. "What assumptions did you make?" and "what are the main weaknesses of this approach?" often surface important gaps faster than re-reading the response yourself.
For complex tasks, break them into steps. Ask ChatGPT to complete the first step and confirm it before moving to the next. This catches misunderstandings early rather than after a long response has gone in the wrong direction.
Check what matters against real sources
ChatGPT can produce confident-sounding answers that are wrong or out of date. For anything where accuracy matters — a legal detail, a technical specification, a current figure — verify the specific claim against the primary source. Ask ChatGPT to provide its source, then open that source and check.
ChatGPT with web browsing can retrieve recent information, but it can also hallucinate citations. If it gives you a specific URL or publication, visit the page and confirm the information is there before relying on it.
Save good prompts for reuse
When a prompt works well — the exact framing that gets the right kind of code review, or the framing that produces useful feedback on a document — save it. Writing it again from memory next time means rebuilding something that already works.
ChatGPT does not provide a native prompt library outside of custom instructions and Projects. Keeping a list of useful prompts in a notes app or a shared record means they are retrievable without searching through old conversations.
Start every session with the context already there
A fenbs project record holds the rules, decisions and Where we left off that ChatGPT reads at the start of a session. Instead of pasting context into each new conversation, the connected assistant reads it directly. Create a free account, add the project and write the first set of standing rules and decisions so the next session starts informed.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
How to do this with fenbs
- Create a free account and add a board for each ongoing project.
- Write the standing rules and decisions for each project as Decisions and rules. These appear first in the AI context at the start of each connected session.
- After each ChatGPT session, ask it to save progress in fenbs: the task states, any new decisions, and where the next session should start.
- Save useful prompt templates in fenbs's Skills and prompts section so any connected assistant can retrieve them.
- Connect at Integrations so ChatGPT reads the project context at the start of each session instead of waiting for you to re-paste it.
For comparison with other models, the page on Claude vs ChatGPT explains where each performs differently. The guide on how to use ChatGPT efficiently covers workflow habits for getting more done in fewer messages.