The ChatGPT tips from 2023 that stopped working (and what replaced them)

Type "act as an expert" into ChatGPT today and nothing special happens. That trick, along with dozens of other prompt formulas that circulated in 2023, was designed to coax better answers out of a much weaker model. The models got smarter. The interface got rebuilt. And a pile of settings that used to matter either moved somewhere new or disappeared entirely.

Meanwhile, features that didn't exist two years ago — Projects, persistent Memory, Study Mode, Scheduled Tasks, and a completely rebuilt image generator — now do more heavy lifting than any prompt trick ever did. If you're still writing five-paragraph prompt templates and manually re-explaining your context every session, you're working harder than you need to.

This guide walks through what's actually changed, what still works, and how to use the features that replaced the old hacks.

Why don't the old prompt formulas work anymore?

The elaborate "role-play" prompts, the step-by-step reasoning scaffolds, the exact phrasing rituals — most of that was compensating for models that couldn't infer intent well and had no persistent context. Current models handle plain-English requests directly, and the platform now has structural features (memory, projects, connected tools) that do the context-management work prompts used to fake.

That doesn't mean prompting doesn't matter. It means the type of prompting that helps has shifted from "trick the model into behaving" to "give the model the right structure to work inside of." That's where Projects and Memory come in.

What are ChatGPT Projects and why should you use them?

ChatGPT Projects are dedicated workspaces for a specific piece of ongoing work, instead of one long undifferentiated chat history. OpenAI describes them as smart workspaces: Projects are smart workspaces that keep everything related to a long‑running effort in one place. Group together chats, upload reference files, and add custom instructions so ChatGPT remembers what matters and stays on‑topic. With memory, context, and flexible tools, they're ideal for repeated and evolving work such as writing, research, planning, and more.

Here's how to set one up:

Open ChatGPT and click "New project" in the sidebar
Name it around a specific outcome (a client, a book, a quarter of financial planning)
Upload the reference files that matter — contracts, style guides, past drafts
Add project-specific instructions so every chat inside it follows the same tone and constraints
Start chatting — every conversation in that project can pull from the files and prior chats already there

The part most people miss is memory scoping. Members who choose to use improved memory can use Project-only memory to keep memory context limited to a specific project. With project-only memory, chats can reference other conversations in the same project, but they cannot reference conversations outside the project, and chats outside the project cannot reference conversations in it. That's useful when you're running multiple clients or projects and don't want details bleeding between them.

How does ChatGPT Memory actually work?

Memory used to be a single toggle. Now it's layered. General memory works quietly in the background: when enabled, memory helps ChatGPT automatically remember useful context from your chats, files, and connected apps to personalize your experience, so you don't have to repeat yourself as often. Memory controls are available in Settings > Personalization > Memory.

You're not stuck with whatever it decides to remember, either. You can see the information that ChatGPT remembers in the memory summary. If you want to know if ChatGPT has remembered something, just ask in chat. If it's holding onto something wrong or outdated, you can edit it directly — type what you want changed into the text box at the bottom of the memory summary and it will update accordingly.

Practical move: check your memory summary once a month. Old context (a job you left, a project that wrapped up) can quietly skew future answers if you never clean it out.

What is Study Mode and who is it actually for?

Study Mode is the closest ChatGPT gets to a tutor instead of an answer machine. Instead of handing over a finished solution, study mode in ChatGPT — a learning experience that helps you work through problems step by step instead of just getting an answer. OpenAI built it specifically to address a real concern in education: when students engage with study mode, they're met with guiding questions that calibrate responses to their objective and skill level to help them build deeper understanding.

To turn it on, go to chatgpt.com and click the tools icon in the composer, then select "Study and learn." OpenAI's own instructions confirm the simple flow: Open ChatGPT and log in. Click on the plus sign on the left side of the prompt textbox. Select "Study and learn." Study mode is now active. There's no separate settings page to hunt through — it's a single toggle in the tools menu.

It's genuinely useful for three situations:

Homework and test prep — it won't just solve the problem, it'll ask what you've tried
Reading dense material — upload a PDF or lecture slide and ask it to walk you through the argument
Learning a skill from scratch — it adapts explanations based on what you already seem to know

If you just need the answer fast, turn it off and use regular chat. Study Mode is deliberately slower — that's the point.

How do Scheduled Tasks replace manual reminders and follow-ups?

Scheduled Tasks got a full rebuild in mid-2026, replacing the older "Pulse" system. The new version lives in its own sidebar page and handles far more than reminders. ChatGPT can run one-time or recurring tasks, monitor for changes, and respond to supported events when available.

Setting one up doesn't require special syntax — you just ask. To create a task, ask ChatGPT to complete an action, such as "Let me know when my package is delivered." From there, manage everything from one place: go to Scheduled to view and manage tasks. You can also open the task list mid-conversation — to open the task list from a conversation, select the more options menu (•••), then select See scheduled tasks. Select the task title in its conversation.

There's a genuinely powerful upgrade here that most people haven't discovered: monitoring tasks that watch your actual tools. Event-triggered (webhook-based) tasks run in Work and respond to supported Gmail, Slack, or GitHub activity. They are available to eligible users on Plus, Pro, Business, Enterprise, and Edu plans, and in ChatGPT for Healthcare workspaces. That means a task can watch your Gmail for a specific type of message and only notify you when something actually needs attention — instead of you checking manually.

One limit worth knowing before you build a workflow around this: a scheduled task cannot run more than once per hour. Plan usage limits also apply to task runs. Also worth noting — if you create a task in a project, it cannot access uploaded files or files stored in that project, so don't expect project-scoped tasks to read your project documents automatically.

What changed in ChatGPT's image generation update?

If the last time you generated an image in ChatGPT was more than a few months ago, the tool you remember doesn't exist anymore. The two biggest wins are aspect ratio control and character consistency, and together they turn image generation from a novelty into something you can actually use for real design work.

Aspect ratio used to be a fixed set of presets. Now it's a continuous range: aspect ratios now run from 3:1 to 1:3, with native 16:9 and 9:16. That means one prompt becomes a YouTube thumbnail, an Instagram Story, and a wide LinkedIn banner. No cropping. No reframing. Instead of generating a square and manually cropping it into a banner, you just tell the model the ratio you need up front — or ask it to reframe an existing image without changing the composition.

Character consistency has also jumped noticeably. The newer model handles reference inputs and iterative edits far better than earlier versions: better character and scene consistency. Reference inputs and prior outputs are honored more carefully. Doing iterative edits (same character, new pose; same poster, new headline) no longer requires re-rolling the whole composition. That's the difference between "generate a mascot" being a one-off image and being an actual reusable brand asset.

The model can now also handle genuinely complex layouts without falling apart. Editing and Format Flexibility Are First-Class: Conversational refinement, selective area edits, up to 8 consistent multi-image outputs with character continuity, and any aspect ratio from 3:1 to 1:3 — the loop matches how real creative teams iterate. If you're producing a batch of product shots or a comic-style storyboard, ask for the full set in one prompt with a locked character description — you'll get more consistency across frames than generating them one at a time.

Practical prompt structure that works well with the new model:

Generate [subject], aspect ratio 16:9, keep the character's
face, hair, and outfit consistent across all versions.

Skip the old-style prompt scaffolding ("You are a professional photographer with 20 years of experience..."). It doesn't move the needle the way it used to — plain, specific instructions do.

How should you actually rebuild your ChatGPT workflow?

Stop treating ChatGPT as one long chat window and start treating it as four separate tools stacked on top of each other:

Projects for anything ongoing — a client, a course, a book
Memory for personal context that should follow you everywhere else
Scheduled Tasks for anything you'd otherwise have to remember to check manually
Study Mode for anything you're trying to actually understand, not just get answered

None of this requires a prompt formula. It requires picking the right feature for the job before you start typing.

The tools changed. Most tutorials didn't catch up. The people getting the most out of ChatGPT right now aren't the ones with the best prompts — they're the ones who stopped guessing and started reading what actually shipped.