Phase 1 · ChatGPT · Level 2 · Practitioner
ChatGPT: Practitioner Quiz · 10 min
Why it matters A quick check of your judgement before you move on. There's no penalty for a low score, and you can retake it, or take it first to test out of a level you already know.
You've moved past the basics into the features that make ChatGPT useful for real work: files, memory, Projects, Canvas and research. These questions aren't trivia; they're the judgement calls that decide whether a slick-looking output is actually right and safe to send. Pick the best option for each scenario, and you'll see an explanation after each one.
Passing this checkpoint: work through the level first, then score 70% or more here.
Skipping this level: already confident? Take this cold and score 80% or more to test out and jump ahead. Below that, nothing is lost, you just study the level as normal.
1. You upload a sales spreadsheet and ask ChatGPT for the total revenue. It replies instantly with a confident figure and no mention of doing any calculation. What's the wisest response? Use the figure; it's from the actual file, so it must be accurate. Ask it to calculate the total by running code rather than estimating, and to show the working, then spot-check a few rows. Ask the same question again and average the two answers. Round the figure up slightly to be safe and use that. 2. A messy CSV lists regions as "South", "Sotuh" and "South". You ask ChatGPT to total revenue by region and it returns a tidy table with a separate small line for "Sotuh". What has gone wrong? Nothing; "Sotuh" is clearly a distinct region that needs its own line. It grouped by the literal text, so a typo became a phantom region; the real South total is understated and needs the typo corrected first. The table is fine; you should just delete the "Sotuh" row to tidy it. ChatGPT invented the "Sotuh" line and it isn't in the data at all. 3. You want ChatGPT to always reply in UK English, in short bullet points, for your non-technical team. Where should that go? Typed into every new chat, so you stay in control of it. Into custom instructions, since it's a durable preference that should apply everywhere. Into a memory you ask it to save mid-conversation. Nowhere; you can't change how ChatGPT writes. 4. You open your Memory list and find an item: "User is leaving their job soon." You mentioned it once, weeks ago, and it's no longer true. Why does this matter? It doesn't; old memories are ignored automatically. A stale saved fact can quietly colour future replies across chats, so out-of-date or oversharing items are worth deleting when you review the list. You should turn Memory off permanently; there's no other fix. You should keep it; more memory always means better answers. 5. You're running two pieces of work in ChatGPT: a sensitive HR matter and a general marketing campaign. What's the cleanest way to keep their context and files apart? Keep both in one big "Work" Project so everything's in one place. Put each in its own Project, so their files and project memory stay compartmentalised. Run both in ordinary chats and rely on ChatGPT to keep topics separate. Use custom instructions to tell it never to mix the two. 6. You've drafted a two-page report and need to rework the tone of just one section, then shorten the whole thing, over several rounds. Which approach fits best? Keep asking in plain chat and re-read the whole reprinted document each time. Open it in Canvas and make targeted edits to specific sections, using version history to compare. Start a new chat for each change so nothing gets confused. Paste it into email and edit it entirely by hand. 7. You need to compare three suppliers' current pricing and data-protection terms for a decision you'll act on, an involved question. Which mode fits, and what should you do before it runs? A quick web search; there's nothing to set up. Deep Research, and read its proposed plan, correcting the scope (UK-only, official sources) before it starts. Deep Research, but skip the plan step to save time. ChatGPT's built-in knowledge, since it knows about suppliers. 8. A Deep Research report comes back polished, well-structured and full of citations. What should you do before relying on it? Trust it; the citations prove every claim is backed by a real source. Open the key citations and confirm each source actually says what the report claims, and cross-check any load-bearing figure. Assume the length and formatting mean it's been checked for you. Trust it if it sounds confident and the links all open. 9. You set up a Project with the launch brief as a project file, and a new chat answers "the launch is on 3 September" by citing the brief. It's for a message going to the whole company. What's the right move? Send it; the answer came from your own uploaded file, so it's reliable. Ask it to quote the exact line, then glance at the brief yourself to confirm the date before it goes out. Change the date to whatever you remember it being. Trust it because Projects don't make mistakes with their own files. Answered 0 of 9.
Answer every question to see your score