📅 Last updated: March 2026

✍️ Written by Jordan Michaels, Writing & Research Tools Editor at AiToolRadar

Overview

OpenAI's flagship conversational AI. The largest ecosystem of custom GPTs, best-in-class general capabilities, and the most recognized brand in AI. From writing to coding to analysis — it handles almost everything reasonably well.

My Experience with ChatGPT

I've been using ChatGPT almost every day since it launched, and my relationship with it has evolved a lot. It's no longer the "wow, this thing can write!" moment — it's become more like a utility, a thinking partner I reach for automatically. Here's how I actually use it and where I think it earns its keep.

Daily workflows: writing, coding, brainstorming

My morning usually starts with a quick ChatGPT session. Not for anything dramatic — it's mostly "help me phrase this email better" or "what's a cleaner way to structure this function." For writing, I use it to get unstuck: I'll dump a rough outline and ask it to suggest a narrative arc, then rewrite everything myself. The drafts it produces are rarely publish-ready, but they break the blank-page paralysis faster than anything else I've tried.

For coding, I treat ChatGPT like a well-read senior developer who can look at a single file and give solid advice, but who hasn't seen your whole codebase. It's great for: writing one-off scripts, explaining what an unfamiliar function does, generating boilerplate I don't want to type, and talking through algorithm choices. Where it falls short is multi-file reasoning — for that I switch to Claude Code or Cursor. Knowing when to switch is half the skill.

Brainstorming is where ChatGPT genuinely shines for me. I'll throw a half-formed product idea at it and ask for 10 angles I haven't considered. It generates fast, the quality varies, but there's almost always one idea in the batch that's actually useful. I've shipped features that started as a throwaway ChatGPT suggestion.

Custom GPTs: the underrated superpower

Most people try ChatGPT's standard interface and stop there. The Custom GPT ecosystem changes the game. A Custom GPT is basically a saved system prompt plus optional uploaded files and tools — but that simple description undersells how useful they are in practice.

I have two Custom GPTs I use constantly. The first is a code review GPT: I've given it a system prompt that describes my team's conventions (naming style, preferred patterns, things we always flag in PRs). When I paste a chunk of code, it reviews it through that lens rather than giving generic advice. It saves me from bikeshedding on PRs and catches things I'd otherwise miss after staring at the same code for hours.

The second is an email drafting GPT. I've uploaded examples of emails I consider well-written and given it instructions about tone — direct but not terse, no filler phrases, always lead with the ask. Now when I need to write a difficult message (declining a partnership, following up on a late payment, giving tough feedback), I paste in the context and get a draft that sounds like me, not like a LinkedIn post.

Building Custom GPTs takes maybe 20 minutes if you already know what you want. The ROI is massive if you do any repetitive type of communication or review work.

GPT-4o vs GPT-4o mini — when to use which

GPT-4o mini is faster and cheaper (essentially free at the Plus tier for most tasks). I use it for anything where I want quick output and I'm not worried about subtle reasoning quality: summarizing a document, generating a first draft I'm going to heavily edit, answering a factual question I'll verify anyway. GPT-4o is what I reach for when the task requires actual thinking — complex code, nuanced analysis, anything where I'd be frustrated by a confidently wrong answer. In practice, I default to mini and upgrade to 4o when the response feels shallow.

Voice mode: better than it sounds

Voice mode is genuinely underrated. I use it for thinking out loud — I'll describe a problem verbally while walking, and the back-and-forth helps me find the flaw in my reasoning faster than typing. It's not about dictating commands; it's about the act of explaining something to another "person" and being asked follow-up questions. There's something about speaking that forces clarity. I've talked through architectural decisions, worked out how to handle a difficult conversation with a colleague, and even used it to prep for client calls. If you haven't tried it for 20 minutes straight, you're missing something.

The system prompt trick for consistent output

One habit that improved my results dramatically: I open every new conversation with a system prompt-style message before asking my actual question. Something like: "You are helping a developer write clear, technical documentation. Be concise. Use bullet points for steps. Avoid filler phrases like 'Certainly!' or 'Great question!'" Then I ask my question. The output quality and consistency improves noticeably. You don't need a Custom GPT for this — just paste a setup block and go.

Honest weaknesses I've run into

Hallucinations on recent events are still a real problem. ChatGPT will confidently cite a paper that doesn't exist, invent a version number, or describe a feature that was deprecated. This isn't a dealbreaker — it just means you verify anything you can't afford to get wrong. For current events or anything time-sensitive, I go straight to Perplexity, which is built for retrieval rather than generation.

Context window limitations are the other friction point. ChatGPT Plus gives you a 128k context window, which sounds like a lot until you're trying to analyze a large codebase or a book-length document. Claude's context window and how it uses context is genuinely better for long-document work in my experience — it seems to "remember" earlier parts of a long conversation more reliably.

Is ChatGPT Plus worth $20/month?

Honest answer: yes, if you use it three or more times per week for real work. If you're using it occasionally to generate a birthday card, stick with free. But if AI assistance is part of your actual workflow — writing, coding, research, communication — $20/month is cheap compared to what an hour of your time costs. The jump from free (with daily caps that kick in at the worst moment) to Plus (unlimited, faster, with voice and image generation) is meaningful. I've been a Plus subscriber since month two and I've never thought about cancelling.

Pricing

Free tierGPT-4o with daily limits
Paid plans$20/mo (Plus) · $25/user/mo (Team)
ℹ️ Prices shown as of March 2026. Plans and availability may vary by region — check the official website for the most current pricing.

✅ When to use

  • First drafts of any written content
  • Explaining complex topics in simple terms
  • Quick coding questions, debugging snippets, and explaining error messages
  • Generating boilerplate code or scaffolding a new project
  • Brainstorming ideas and generating multiple options
  • Summarizing long documents or articles

❌ When NOT to use

  • Tasks requiring real-time web data — use Perplexity instead
  • Long document analysis over 100k tokens — Claude handles context better
  • Deep codebase work across many files — use Claude Code or Cursor instead
  • Strict instruction-following with complex constraints — Claude is more reliable

💡 Personal Tips

ChatGPT is my default starting point for most tasks — not because it's always the best, but because the Custom GPT ecosystem is genuinely useful. I have GPTs set up for specific workflows that save me hours. For coding, it's great for quick one-off questions and explaining concepts, but for serious multi-file work you'll want Claude Code or Cursor. The free tier is fine for occasional use; pay for Plus only if you're relying on it daily. Underrated tip: use voice mode on mobile for hands-free brainstorming while commuting.

Alternatives

FAQ

Is ChatGPT free to use?

Yes — ChatGPT offers a free tier with access to GPT-4o under daily usage limits. The Plus plan ($20/mo) removes limits and adds priority access, voice mode, DALL-E image generation, and the latest models.

What is ChatGPT best used for?

ChatGPT excels at writing assistance, brainstorming, summarizing documents, explaining complex topics, generating code snippets, and everyday conversational AI tasks. Its Custom GPT ecosystem also allows specialized workflows.

Is ChatGPT good for coding?

Yes, especially for quick scripts, code explanations, debugging single files, and generating boilerplate. For large multi-file codebases, Claude Code or Cursor is generally more capable.

How is ChatGPT different from Claude or Gemini?

ChatGPT has the largest ecosystem (Custom GPTs, plugins, voice mode, DALL-E). Claude has a larger context window and is better at instruction-following. Gemini has deeper Google Workspace integration. ChatGPT is typically the best all-rounder for everyday tasks.

What are Custom GPTs and are they worth using?

Custom GPTs are personalized versions of ChatGPT with a saved system prompt, optional uploaded files, and specific tool access. You can build one in about 20 minutes. They're genuinely worth the effort for any repetitive task — a code review GPT tuned to your team's conventions, an email drafting GPT that matches your tone, or a research assistant primed with your domain context. Available to Plus subscribers only.

Should I use GPT-4o or GPT-4o mini?

Use GPT-4o mini for speed and quick tasks where accuracy is easy to verify — summaries, first drafts, simple Q&A. Use GPT-4o when the task requires real reasoning: complex code, nuanced analysis, or anything where a confident wrong answer would cost you time. In practice, start with mini and upgrade if the response feels shallow or misses the point.

Does ChatGPT have access to the internet?

ChatGPT Plus has a web browsing tool that can search the internet for recent information. However, it's not as reliable or fast as a dedicated search-first tool like Perplexity. For up-to-date research or anything requiring current data, Perplexity is the better choice. For analysis and generation tasks, stick with ChatGPT.

How do I get consistent output from ChatGPT?

Start every conversation with a setup message that defines tone, format, and constraints before asking your actual question — treat it like a system prompt. For example: "You are helping a developer write concise documentation. Use bullet points for steps. Skip filler phrases." This dramatically improves consistency without needing a Custom GPT. If you use the same setup often, turn it into a Custom GPT to save time.