The last time I paid for ChatGPT Plus, it was about $20 a month—at least for me. Depending on where you live or what taxes sneak in, your bill might look a bit different, so I always recommend checking your own billing page just to be sure. As of April 2026, I’d treat that $20 as a rough estimate, not a promise from OpenAI.
If you’re already paying for Plus—or just thinking about it—and actually want it to make a real difference in your daily work, you’re in the right spot. I’m not about to hit you with a bunch of buzzwords. Instead, I’ll walk you through the setup steps I use myself, along with a few habits that have saved me from that dreaded ‘same prompt, fourth try’ frustration.
To be honest, Plus isn’t some magic, super-smart chatbot that’s going to blow your mind the second you sign up. What you’re really getting is a bunch of extra features—but you’ll only notice them if you actually put them to use. We’re talking better access to the latest GPT‑5 models, Projects, Canvas, Deep Research, Memory, Custom GPTs, Codex for devs, Study Mode, quizzes, and image tools right inside your chat. If you mostly use ChatGPT to rewrite emails or just mess around on weekends, you probably won’t need Plus—and honestly, it’ll just feel like a waste. The real loss isn’t the $20; it’s paying for all these tools and never letting them actually help with your real work.
Let me show you exactly how I set things up, step by step—just like I’d do if we were sitting down together. Honestly, menus and limits change all the time, so don’t be surprised if something looks a bit different down the road. I treat the names as rough guides and always double-check in the app. If you’re still on the fence about whether Plus is worth it, I put together a Free vs Plus breakdown that focuses on what you’re actually getting, not just a list of features.
Pick the right model mode (speed vs depth)#
With Plus, you usually get a few GPT‑5‑family options: one that answers fast, one that takes its time to think, and sometimes an auto pick. I try not to overthink it. If I’m just brainstorming or need a quick answer, I go for the fast one. But if I’m doing code review, planning something with lots of steps, or really need the model to catch a mistake, I pick the slower one—or I just tell it up front that I want step-by-step reasoning before it starts answering.
Auto mode works fine—until it doesn’t. If the reply feels thin or just doesn’t cut it on a tough problem, I switch modes before I find myself pasting the same prompt for the fourth time.
When I want better answers, I just write my prompt in plain English: I list what I’m assuming, compare two not-so-great options, then wrap up with a plan and what could go wrong. No jargon needed. What you really want is for the model to slow down and think it through.
Custom Instructions: say it once, reuse everywhere#
Settings → Personalization → Custom Instructions is two boxes. Top box: who you are in one breath (job, stack, what you’re studying). Bottom box: how I want my answers: short or long, bullets or paragraphs, answer first or context first. Fill both out once, and you’ll stop getting those five-paragraph ‘As an AI language model…’ intros every time you start a new thread.once, and new threads stop opening with five paragraphs of “As an AI language model…”
It also helps keep Projects and Canvas from ignoring your preferences every time you open them.
Memory: stable preferences without re‑explaining the project#
If you have Memory turned on, ChatGPT can remember your preferences and little bits of context you let it save. I use it for the boring but important stuff: the tone I need for client writing, the class syllabus I’m stuck with, or reminders like ‘we use Postgres, not MySQL.’ Basically, anything I’d otherwise end up pasting in every week.
Don’t put anything in Memory you wouldn’t share in a public notes app, and check what’s saved every so often. Memory can drift over time.
Projects: turn chats into durable workspaces#
I use Projects for anything where I know I’ll need the files again next week: a messy business plan, a repo I’m learning, or a bunch of image drafts that all share the same brief. You can upload files, pin instructions, and keep your threads from getting cluttered with random questions. I usually keep one active brief per project, and when the scope changes, I just start a new project so old instructions don’t follow me into the next job.
To me, it’s like having a desk drawer with a sticky note on it, instead of just another ’new chat number forty-three.’
Canvas: co‑edit long docs and code#
I use Canvas when I need a side-by-side editor—long memos, code I’m working on, or anything where scrolling through one endless chat makes me dizzy. You can ask for it from the toolbar or just say you want Canvas. Honestly, copy-pasting 800 lines between windows is how I used to waste an hour and then blame the model.
Deep Research: from “search” to “structured briefs”#
I turn to Deep Research when I need more than just a quick Google—like figuring out who competes with X, what’s changed in regulations since last year, or how two frameworks actually differ in practice. I’ve found it helps to narrow down the question, mention the time window, and spell out what ‘done’ looks like.
But I always read the results like a journalist, not a true believer. Even with Deep Research, you can still get confident but wrong answers on the little details.
Study Mode and quizzes: learn, don’t only extract answers#
Study Mode will quiz you with questions instead of just handing over a finished paragraph to memorize. I like to pair it with the quiz-style drills when I’m actually trying to pass something. It’s a supplement, not a replacement for your textbook or your professor’s weird exam focus.
Codex: use it like an engineering ticket#
When I use Codex-style agents, I write up the work like I’m handing it to a contractor: one clear feature, tests if I care about them, and any constraints in plain language. I put the repo context in a Project whenever I can. I always say what ‘finished’ means—never just ‘make it better.’
ChatGPT Images: integrate, don’t isolate#
I find the in-chat image tools are most useful when the picture is part of a back-and-forth: draft some text, make a rough visual, tweak the image, then rewrite the text. If you’re doing serious brand photography or heavy editing, you’ll probably need another app. But for me, the real win is being able to stay in one thread instead of bouncing between four browser tabs.
Prompting habits that scale across features#
When something actually matters, I put the role, goal, constraints, and what kind of answer I want all in one message. I don’t bother with ‘chain of thought’ for every quick email. I’ll outline things messily in chat, clean them up in Canvas, and only add images or slides if they’re really needed for the final result. If the format has to be strict—like JSON or something a linter will check—I just paste a tiny example of what ‘correct’ looks like.
Put the subscription to work#
Find a rhythm that actually fits your style. A couple times a week, I make myself run a real task through something besides plain chat—like a Project, Deep Research, Study Mode, Codex, or an image workflow that would be a pain without Plus. I keep a few long-running projects open, and whenever my work changes, I update my Custom Instructions and tidy up Memory. After a month or so, Plus stops feeling like just a faster version of the free tier. Instead, you start to notice you’re dropping fewer threads and spending way less time repeating yourself.
References#
- ChatGPT — product overview.
- Plans and pricing — confirm before you buy or renew.