AppsGamesArticles

How to Use ChatGPT Beyond the Chat Box

Have you ever opened ChatGPT, typed a question, got an answer, and still felt unsure what else it can do? Many people only use the chat box, but the real power is in the extra tools around it. This guide explains the main features you showed in your screenshots, including the menu near the message box and the left sidebar, with clear examples you can copy and use.

ChatGPT started as a smart conversation tool. Today it is closer to a toolbox. You can upload files, analyze data, generate images, search the web, run deeper research, build simple app prototypes, and even use agent style workflows that follow steps to finish a task. If you learn where each feature lives and when to use it, you can save a lot of time every week.

Think of the interface as three areas:

1) The tools menu near the input box (things like Add photos & files, Create image, Deep research, Web search, Agent mode)

2) The More menu (things like Add sources, Study and learn, Canvas, Adobe Photoshop)

3) The left sidebar (things like New chat, Search chats, Images, Apps, Codex, and GPTs)

For each feature, you will see:

What it is in simple words

When to use it so you pick the right tool

How to use it with practical steps

A detailed example prompt you can copy

What it is

This feature lets you upload files directly into the chat. ChatGPT can read and work with the content. That includes screenshots, PDFs, Word documents, PowerPoint files, CSV data, and Excel spreadsheets.

When to use it

Use it when the answer depends on your material, not general knowledge. It is ideal for:

- PDF contracts, reports, manuals, and research papers

- Excel or CSV sales data, ad data, user lists, performance logs

- Images like UI screenshots, design references, ads, dashboards

- Slides when you want a summary, rewrite, or new structure

How to use it

1) Click Add photos & files

2) Upload your file(s)

3) Tell ChatGPT exactly what output you want and what rules matter

4) If you need a structured result, say the format, for example: table, bullet points, JSON, or a step by step checklist

Detailed example prompt

Scenario You upload a messy sales spreadsheet and want a board ready summary.

Copy prompt

Read the uploaded Excel file and do the following. First, check for missing values, duplicates, inconsistent date formats, and obvious outliers. Explain how you cleaned the data in plain language. Second, create a monthly summary table with revenue, orders, and average order value. Third, describe the top 3 changes month to month and what might explain them. Fourth, write an executive summary under 180 words that I can paste into a weekly report.

What you get

You get cleaning notes, a monthly table, a short insight list, and a ready to use executive summary. This is much more useful than a generic explanation.

What it is

This tool generates images from text prompts and can also edit images you upload. It is useful for creative work, UI assets, marketing images, storyboards, and style tests.

When to use it

- You need a new image for a concept, scene, character, or background

- You want to change an existing image, such as making it wider, changing the background, cleaning up details, or matching a style

- You are producing assets for a game, app, quiz, or marketing flow

How to use it

1) Click Create image

2) Write a prompt that includes subject, environment, camera, lighting, style, and any restrictions

3) Add size requirements if you have them, like 1280x720 or 720x1280

4) If you are editing an existing image, be clear about what must stay the same and what can change

Detailed example prompt

Scenario You need a 9:16 background scene with a specific mood.

Copy prompt

Generate a 720x1280 vertical image of an empty small town street at night during heavy rain. Wet asphalt with clear reflections, soft neon signage glow, moody fog in the distance, no people, no text, no logos. Camera is slightly low and centered, leading lines pull the eye forward. Realistic materials and lighting, detailed but not overexposed. Keep the mood tense and quiet.

Tip for better results

Add constraints like no text, no watermark, no extra characters. Mention the camera angle and lighting so the image looks intentional, not random.

What it is

Deep research is a research mode that focuses on finding reliable information first and then writing based on evidence. It is built for reports, comparisons, and content that should not rely on memory.

When to use it

- You want a mini research report, not a quick answer

- You need sources you can check

- You are writing content that needs accuracy, such as market analysis, product comparisons, policy summaries, or technical overviews

How to use it

1) Click Deep research

2) Define the question and the scope, including time range and the type of sources you trust

3) Ask for a structured output, like sections, table comparisons, and a final summary

Detailed example prompt

Scenario You want a credible article about AI agents for business users.

Copy prompt

Use deep research to explain how AI agents are being used in small and mid size businesses in 2025 and 2026. Focus on real use cases like customer support, sales outreach, data reporting, and competitor monitoring. For each use case, include one real example from a product, company, or public case study and provide citations. Then list the common risks and practical ways to reduce them. End with a simple 30 day starter plan for a team with no AI background.

What it is

Shopping research helps you compare products as a buyer. It is designed to narrow options and highlight what matters in real life, like price, tradeoffs, and fit for your needs.

When to use it

- You are buying hardware or software and want a shortlist

- You want a simple comparison table and recommendations

- You need help translating requirements into product specs

How to use it

1) Click Shopping research

2) Tell it your budget, region, must have features, and what you want to avoid

3) Ask for a comparison table plus a clear recommendation

Detailed example prompt

I want a lightweight 14 inch laptop for travel. Budget is 1000 to 1500 USD. I will edit 1080p video, do basic Python work, and I care about battery life and weight. Give me 5 options with a comparison table including CPU, RAM, storage, weight, battery estimates, ports, and a short reason for each. Then tell me which two are best for my needs and why.

What it is

Agent mode is a way to let ChatGPT act more like a worker that follows steps, checks progress, and finishes a full workflow. Instead of answering one question, it can plan and execute multiple actions to produce a final deliverable.

When to use it

- You want a repeatable workflow, like a weekly report

- You have a multi step task, like competitor monitoring plus summary

- You want a complete output, not advice

How to use it

1) Click Agent mode

2) Provide the goal, inputs, rules, and output format

3) Ask it to confirm assumptions and list the steps it will follow

4) Review the result and refine the instructions for next time

Detailed example prompt

You are my weekly competitor monitor. Here are 8 competitor URLs. Every Monday, check their pricing pages and product announcement pages. Compare any changes against my own pricing table that I will paste here. Output a report with these sections: Summary, Price changes, New features, Risk level, Suggested response actions. Keep it short and clear. If you find a major discount or a new plan, highlight it at the top.

Why this matters

This is the difference between a tool that talks and a tool that produces.

What it is

Add sources is about giving ChatGPT a reliable knowledge base to use. It can be your own docs, policies, writing guidelines, product notes, or other reference material you trust.

When to use it

- You want ChatGPT to answer based on your internal materials

- You want consistent outputs that follow your style and rules

- You are building a personal knowledge library

How to use it

1) Open Add sources

2) Connect or upload materials

3) Set a simple rule: use my sources first, and say when something is missing

Detailed example prompt

I will provide a folder of product docs and writing rules. For every response, use my sources first. If the answer is not in my sources, clearly say that and then give general guidance. When you use my sources, include a short note that points to the relevant section so I can confirm quickly.

What it is

This mode is designed for learning. It breaks topics into small steps, checks understanding, and helps you practice. It is great for coding, language learning, or any skill you want to build steadily.

When to use it

- You want a structured learning path

- You want exercises and feedback, not just explanations

- You want the model to slow down and teach clearly

Detailed example prompt

Teach me Selenium step by step as a beginner. Each lesson should include one concept, one short code example, and one small practice task. After I respond, correct me and then move on. Keep the language simple and avoid long explanations.

What it is

Web search is the fast way to look things up online. Use it when you need current information, official documentation, or the latest news.

When to use it

- You want the newest steps for a tool setup

- You want official links

- You want to confirm facts that change over time

Detailed example prompt

Use web search to find the current official setup steps for PayPal sandbox testing for a website checkout flow. Summarize the steps in plain English and list the common mistakes developers run into.

What it is

Canvas is a workspace for longer projects. It is useful when you want to write, edit, and refine a document or code over multiple iterations. It feels closer to working in a document editor than a chat thread.

When to use it

- Long guides, scripts, proposals, and content templates

- Rewriting and polishing large documents

- Building structured outputs that need consistent formatting

How to use it

1) Open Canvas

2) Create a document with a clear title

3) Tell ChatGPT the formatting rules and the goal

4) Iterate section by section

Detailed example prompt

Create a Canvas document called Weekly AI Tools Report. Use a fixed structure: Highlights, New releases, Useful workflows, Risks and notes, Next week watchlist. Every week, I will paste links and short notes. You will fill the sections and keep the structure consistent.

What it is

This option connects to Photoshop style editing workflows. It is best when you need more precise image editing than a simple prompt, such as clean cutouts, layered thinking, and detailed fixes.

When to use it

- You need clean transparent PNG assets from a UI screenshot

- You want to remove backgrounds, repair edges, or polish a composite

- You want consistent assets for production use

Detailed example prompt

Extract the UI elements from this screenshot into separate transparent PNGs. Include buttons, icons, progress bars, and cards. Keep original sizes, remove the background cleanly, and name files by element type so I can reuse them in my design system.

What it is

This is a reflection tool. It helps you review your past activity and patterns. People use it to understand what they worked on, what topics mattered, and what they might improve.

When to use it

- You want a yearly recap of your usage

- You want to spot habits, strengths, and gaps

- You want a short summary you can save as notes

What it is

This is a playful area for interactive quiz experiences. It can inspire quiz formats, question styles, and result page structures.

When to use it

- You want ideas for quiz pacing and results

- You want to test user engagement flows

What it is

This is where you can browse different built in apps and tool experiences. If you are not sure which feature is best for your task, this is a good place to explore options.

What Codex is 

Codex is ChatGPT's code-focused working environment.

In normal chat, ChatGPT can write code, but it behaves more like a "helpful explainer."
In Codex, ChatGPT behaves more like a junior software engineer sitting next to you, whose main goals are:

  • correctness

  • structure

  • readability

  • maintainability

Codex is optimized to read existing code carefully, understand how different parts connect, and then modify or extend it without breaking things.

Think of Codex not as "AI that writes code from scratch," but as:

a tool for working with real code, not just generating examples.

How Codex is different from normal ChatGPT

Many users do not realize this difference, so it is important to be clear.

Normal chat mode is best for:

  • learning concepts

  • small snippets

  • explanations

  • brainstorming ideas

Codex is best for:

  • editing real projects

  • refactoring existing files

  • fixing bugs without changing unrelated logic

  • keeping layout, styles, and structure consistent

In other words:

If you care about not breaking what already works, Codex is the safer choice.

When to use Codex

You should switch to Codex when at least one of these is true:

  • You already have working code and want to change layout, style, or logic

  • You are building a small but real tool, not a demo snippet

  • You need ChatGPT to return complete files, not fragments

  • You want changes to be minimal and controlled, not creative

Typical use cases include:

  • Building a simple web tool with HTML, CSS, and JavaScript

  • Adjusting layouts for mobile or tablet

  • Debugging automation scripts (Python, JavaScript, Selenium)

  • Refactoring messy code into cleaner functions

  • Adding features without rewriting everything

How to use Codex effectively

Codex works best when you treat it like a developer, not a chatbot.

A strong Codex request usually includes:

  1. The full code (or at least the full file being changed)

  2. What must stay the same

  3. What must change

  4. How careful it should be

Avoid vague instructions like "improve this" or "make it better."

Instead, be explicit.

Detailed example explained step by step

Original example prompt:

Here is my interactive web page code. Convert the layout to a mobile first vertical design. The main image area should take about 80 percent of the screen height. Place the narration box on top of the image with a frosted glass look. Keep the fonts and all other styling as close as possible. Return the full updated HTML and CSS.

Let's break down why this prompt works well:

  • "Convert the layout to a mobile first vertical design"
    → clear goal

  • "The main image area should take about 80 percent of the screen height"
    → measurable requirement

  • "Place the narration box on top of the image with a frosted glass look"
    → concrete UI instruction

  • "Keep the fonts and all other styling as close as possible"

  • "Return the full updated HTML and CSS"
    → defines output format

This is exactly the kind of instruction Codex handles best.

A more advanced Codex example

Here is a more advanced prompt that shows Codex's real strength:

This is a working interactive story page.

Please do the following carefully:

  • Convert the layout to a mobile-first vertical design (9:16).

  • The background image must remain unchanged.

  • Character images should scale responsively but never exceed 80 percent of screen height.

  • Move the dialogue box above the image and apply a semi-transparent frosted glass effect.

  • Do not change font families, font sizes, or animation timing unless necessary.

  • Keep all existing JavaScript logic exactly the same.

Return the complete updated HTML and CSS.
Also list the specific sections you modified.

This kind of prompt turns Codex into a safe refactoring assistant, which is extremely valuable if you are not a professional developer.

What GPTs are 

GPTs are custom versions of ChatGPT that you design for a specific job.

Instead of repeating the same instructions every time, you can:

  • define rules once

  • lock in output formats

  • enforce tone, structure, and constraints

After that, you just use the GPT like a tool.

Think of GPTs as:

"Your personal AI tools, built without coding."

How GPTs are different from normal chats

In a normal chat:

  • rules are easy to forget

  • formatting may drift

  • tone may change

  • structure may vary

In a GPT:

  • rules are permanent

  • formatting is consistent

  • output is predictable

  • mistakes repeat less often

This is especially important for people who:

  • write content at scale

  • generate prompts repeatedly

  • follow strict formatting rules

  • work in teams

When GPTs are most useful

You should create a custom GPT when:

  • You repeat the same workflow more than 3–5 times

  • You care about consistency more than creativity

  • You have strict rules (wording, punctuation, structure, tone)

  • You want to reduce prompt writing time

Typical GPT use cases include:

  • Image prompt generator

  • SEO article writer with fixed structure

  • Quiz content generator

  • Data extraction and formatting assistant

  • Brand-specific copywriting tool

What you can define in a GPT

When creating a GPT, you can define:

  • Its role (for example: “You are an image prompt engineer”)

  • Its rules (for example: punctuation, banned words, structure)

  • Its output format (tables, sections, JSON, HTML)

  • Its tone (formal, casual, user-like, neutral)

  • Its language behavior (single language or multilingual)

This is why GPTs are powerful: they encode behavior, not just knowledge.

Detailed example explained

Original example prompt:

Create a custom GPT called PromptFactory. It should generate image prompts in two languages, English and Chinese. For every request, it outputs: a main prompt, a negative prompt, camera and lighting notes, and three variants optimized for different image tools. It must use only straight apostrophes and avoid emojis. It must keep prompts clear and practical.

Let's explain what this GPT actually does:

  • It eliminates repeated instruction writing

  • It enforces bilingual output automatically

  • It standardizes prompt structure

  • It prevents common formatting mistakes

  • It becomes a reusable production tool

Once built, usage becomes extremely simple.

How using a GPT changes your workflow

Without a GPT:

  • You rewrite long prompts every time

  • You correct formatting mistakes repeatedly

  • You remind ChatGPT of rules again and again

With a GPT:

  • You write one short input like:

    "Rainy cyberpunk street background, no characters, 9:16"

  • The GPT automatically outputs:

    • English prompt

    • Chinese prompt

    • Negative prompt

    • Camera notes

    • Lighting notes

    • Three optimized variants

This is not convenience. This is workflow compression.

A more advanced GPT example

Here is an example of a high-value custom GPT setup:

Name: RSOC Article Builder

Purpose: Generate SEO articles with strict structure

Rules:

  • Always output HTML using h2 and p tags

  • Bold important keywords

  • Begin with a question in the first paragraph

  • End with a Conclusion section

  • Avoid specific banned phrases

  • Use only straight apostrophes

Behavior:

  • First output a list of similar titles

  • Then generate the full article

  • Keep language natural and human-like

Once created, every article follows the same standard automatically.

Goal Convert raw data into a clear story.

Tools Add photos & files plus basic analysis.

Steps

1) Upload the file

2) Ask for cleaning, summary tables, and charts

3) Ask for a short executive summary and action suggestions

Copy prompt

I uploaded a CSV. Clean it and create a simple report: 3 key charts, 5 insights, and 3 actions I should take next week. Keep the language simple and business friendly.

Goal Avoid outdated facts and improve trust.

Tools Deep research plus Canvas.

Steps

1) Run deep research with a clear scope

2) Ask for an outline and key points with sources

3) Move the writing into Canvas and edit it section by section

Goal Create a working prototype fast.

Tools Codex plus Create image if you need visuals.

Steps

1) Describe the tool you want, inputs, outputs, and UI style

2) Ask for a single file prototype

3) Test locally and ask for fixes

Copy prompt

Build a simple calculator web page in one HTML file. Users enter daily ad spend and average cost per click. Output estimated clicks and a simple weekly total. Make it mobile friendly and clean. Include basic input validation and a reset button.

Goal Spend less time checking the same sites.

Tools Agent mode and scheduling if available.

Steps

1) Define the sources to monitor

2) Define what counts as an important change

3) Define the report format

4) Review the first output, then tighten the rules

Good results come from clear instructions. A practical prompt usually includes four parts:

Goal What you want at the end

Inputs Files, links, or text you will provide

Rules Style, tone, formatting, word limits, what to avoid

Output format Table, checklist, report, code file, bullet points

Example of a strong prompt

Goal: create a one page report I can send to my team. Inputs: I will upload an Excel file. Rules: keep language simple, no long paragraphs, use only English punctuation. Output: a short summary, a table of key metrics, and 3 actions.

ChatGPT can do far more than answer questions in a chat box. Once you use the surrounding tools, you can turn it into a practical assistant for real work. Use Add photos & files to work with your own data and documents. Use Create image for new visuals and image edits. Use Web search and Deep research when accuracy and current information matter. Use Canvas for longer projects that need revision. Use Codex for building and fixing code. Use Agent mode when you want a workflow that produces a finished deliverable. Finally, use the left sidebar tools like Search chats, Images, Apps, and GPTs to stay organized and reuse what works. If you learn these parts of the interface, you will know what ChatGPT can do, how to trigger the right feature, and how to get outputs you can actually use.

Editor's Choice