Struggling With ChatGPT? You’re Not Alone — Try This
I’ve been using OpenAI products for years — even before ChatGPT. My “company of one” has a $200,000 per month usage cap. It’s safe to say I rely on OpenAI models more than most people.
Many of my friends and family treat ChatGPT like Google: type in a search, cross their fingers, and hope something useful pops out. Sometimes it works. Often it doesn’t. Meanwhile, I’m here sounding like a conspiracy theorist, insisting there’s a super-intelligent model hiding behind the website.
I’ve already written about AI’s adoption problem from both the enterprise and product perspectives. If ChatGPT frustrates you, don’t worry, you’re in good company. The AI labs still have a lot of work to do. I’ll know they’re making progress when they stop naming their models after scrapped Star Wars droids.
In the meantime, here are some tips to get the most out of your ChatGPT experience.
Tip 1: Write Longer Prompts
Just because the ChatGPT interface looks like a Google search bar doesn’t mean that’s how you should use it. You can type thousands of words into that box. The models behind ChatGPT thrive on context. The more you provide, the better the results.
Here’s an example. I’m heading to DC this weekend. Here’s a typical prompt I see people type into ChatGPT.
Short ChatGPT prompt
Here are the suggestions that prompt produces:
WorldPride 2025 Kickoff
Music & Festival Highlights: Project Glow Festival, Liberation Weekend, Sound Scene at the Hirshhorn
Arts, Theater & Culture: “Andy Warhol in Iran,” Union Market Street Fest
Food & Drink: Takoma Trukgarten, Providencia
Weather & Safety: Tornado Watch in effect until Saturday, 12:00 AM EDT
Getting Around: Due to WorldPride events, expect road closures
Not bad, but those are the same suggestions ChatGPT would give anybody who asked. Let’s see what happens when I provide more context.
Longer ChatGPT prompt
Here are the suggestions I get with that prompt:
Coffee Spots with Character: Grace Street Coffee Roasters, Yellow the Café, Baked & Wired
Quirky Walking Adventures: The Exorcist Steps, Georgetown Spy Tour
Hidden Gems & Offbeat Stops: Dumbarton Oaks Gardens, Old Stone House, Smith Row
Boutique Browsing: Book Hill
Evening Entertainment: Blues Alley
Weather Outlook: You’re in luck with the weather. Perfect conditions for walking and outdoor activities.
I’m more likely to do the things on the second list than the first. It’s not that the first list was bad. It just failed to capture all the stuff on my mind when I typed “What should I do in DC this weekend?”
When in doubt, word vomit whatever’s on your mind. It’s better to provide the model with too much context than not enough. Context still beats computation in most scenarios.
If you want to learn more about prompting, this article is excellent.
Tip 2: Click “New Chat” Often
If you started a chat by asking which laptop to buy and are now asking about that weird rash — STOP! Click “New Chat” and start fresh.
ChatGPT predicts its next response based on everything in your current conversation. You don’t want it blending tech advice with dermatology. When you switch topics, switch to a new chat.
If your chat’s going off the rails, don’t waste time trying to fix it — just bail. ChatGPT often stumbles before it shines. It’s a probabilistic model, not a perfectionist. Just because it whiffed once doesn’t mean it won’t crush it next time. Copy your prompt, start a new chat, and try again.
Tip 3: Pick the Right Model
The AI companies don’t make it easy to figure out which model to use. Is o3 better than GPT-4o? Is Claude Sonnet 4 better than Claude Opus 4. AI marketing is a train wreck.
If you’re early on the learning curve, here’s a quick rundown of how to use the current options:
Base models (e.g., GPT-4o) — I used this for the DC prompts. Use base models when there’s no single right answer. They’re fast, so you can quickly iterate until you land on something that works.
Reasoning models (e.g., o1, o3) — Use these when there’s a right answer that needs careful thought (e.g., “Should I take this new job or stay where I am?”). Reasoning models are slower because they break complex questions into smaller steps.
Depending on your subscription, you might see a bunch of other models (I currently have access to eight). If you’re finding this article helpful, most of those probably aren’t for you. Just pick the highest-numbered base model and reasoning model you can access, and stick with those.
My current ChatGPT model options
You’ll also see a “Tools” menu in the chat window. Use a tool if it fits, but don’t overthink it. ChatGPT usually knows what it needs. The one exception is “Deep Research.” If you want ChatGPT to spend 10–15 minutes crafting a multi-page response, go ahead and click it (e.g., researching neighborhoods when you’re moving to a new city).
Tip 4: Talk More, Type Less
Most people can speak at double or triple the words per minute that they type. One way to make your prompts longer is to speak them instead of typing them. Speech is underrated as an AI modality.
You can use the dictation button on your phone or in the ChatGPT window to speak your prompt, but that’s not the same as voice mode. Dictation turns your voice into a prompt. Click on the voice mode icon, and you’ll meet a glowing blue orb that is ChatGPT.
Active voice mode in ChatGPT
Voice mode isn’t like Siri. It processes your voice directly rather than converting it to text first. That means you can have a conversation, even interrupting ChatGPT if it starts to wander. I use voice mode to think through complex ideas and draft long emails.
If you haven’t tried voice mode, you’ll be surprised how much better the experience feels once you get used to it. Unfortunately, this is one feature that depends on your subscription tier. That brings me to my last tip.
Tip 5: Pay
I applaud OpenAI for making some impressive capabilities free. That said, the free version of ChatGPT isn’t the same experience as the paid one. Most people I know who are underwhelmed by ChatGPT are still using the free tier. I get it — I’m underwhelmed by that version too.
$240 a year isn’t pocket change. However, that buys you a version of ChatGPT that feels smart. If you’re curious, try it for a month. Worst case, you’re out twenty bucks. Best case, you finally understand why people like me won’t shut up about it.
It’s Still Early…But
If you feel like you’ve missed the AI bandwagon, you haven’t. Even people who talk confidently about AI often don’t fully understand it. You’ve got time to catch up.
That said, you’ve got to start. If you’ve been waiting for AI to get “good enough” to take seriously, now’s the time to revisit that decision. And if you’re still frustrated by AI, there’s a good chance the problem isn’t the technology. It may be how you’re using it.
Good luck!