GPT-5: Adapting to the New AI Model's Quirks and Capabilities

GPT-5 works differently than GPT-4o -- here are 5 expert tips to make it click for you

These smart tips will help you adapt and unlock the model's full potential Now that we've had a full week with GPT-5, hopefully you've had the opportunity to play around and test it. Undoubtedly, you've noticed that it's a bit different from GPT-4o. Millions of people have had mixed feelings about the new model; some love the added depth and intuition, while others feel frustrated by the sudden shift in behavior and tone. I agree that GPT-5 takes a little getting used to. It doesn't always respond the way you expect, especially if you're used to the more literal style of GPT-4o. But once you start to understand how it "thinks" and adjust your prompting style accordingly, you may find it's actually more powerful, creative and conversational than anything that came before. Here are some of the more annoying features about GPT-5 and workarounds to make them less aggravating. GPT-5 is more autonomous and context-aware. Up until now we've had to overexplain our intent when prompting chatbots. Those days are gone because GPT-5 can infer more from less. You don't always need super-specific instructions because the AI just "gets" it. You can now express your intent with fewer words. Prompting tip: Use more natural, open-ended language as if you're talking to a colleague you're collaborating with and not a search engine. For example: Try: "I'm prepping for a job interview. Can you help me sound more confident when talking about my weaknesses?" Not: "Write 3 bullet points describing weaknesses in a professional tone for a job interview in human resources." GPT-5 has thinking modes that activate based on your prompt. Don't worry about the complexity of your question because the AI will automatically switch between a fast, surface-level mode and a slower, more analytical mode depending on the difficulty of your prompt. When prompting, if you want deeper reasoning, ask it to "think step-by-step" or make a point to use prompts that trigger analysis (like comparisons, strategy or abstraction). Try this: "Think step-by-step: how could a small business use AI to improve customer retention?" This model is better at retaining coherence across longer interactions. You may notice that GPT-5 asks you more questions to ensure it understands. At the same time, it handles follow-ups and evolves the conversations much more naturally than GPT-4o, so you can build on what you've already discussed. GPT-5 has an upgraded memory, so it retains even more. For example, if you're in a brainstorming session, try a variety of prompts. Try: "Let's keep going with that idea but make it funnier." Then "Now shift the tone to be more emotional." This model is more multimodal and better at integrating different inputs. The latest upgrade means the AI can analyze images, charts and even tone of voice more effectively than ever before. For that reason, you can fully combine inputs more freely. If you're using Voice or Vision, you'll want to include more contextual cues now. Try: "Here's a photo of the inside of my fridge, can you help me meal plan?" or even "I'm feeling overwhelmed, talk to me like a friend." GPT-5 interprets vagueness or ambiguity more confidently. Remember the "blank space" prompt? That one works even better now because GPT-5 will often "fill in the blanks" more aggressively. Keep in mind that while this can be useful, the update can also be risky if you're not clear. If you don't want the AI to make assumptions, you'll need to be explicitly clear. Try: "Don't assume tone, wait for me to specify." Or, "Ask me follow-up questions before writing." If GPT-5 has left you scratching your head, you're not alone. The shift from GPT-4o to GPT-5 was a little rough for me, too, and I'm a power user. But here's the thing: AI is evolving so quickly that we do ourselves a disservice by getting too attached to any particular model. Once you understand the quirks, you can start working with GPT-5 instead of wrestling against it or longing for the days of GPT-4. Acclimating yourself to all the chatbots, many of them free, will help you adapt and adopt the newest models in the future.

This ChatGPT prompt 'unlocks 100% of GPT-5's power' - and it's so easy to use

OpenAI's latest ChatGPT model, GPT-5, has come under intense scrutiny since its initial launch just a couple of weeks ago. The new model has upset many users with its new reply style, inability to answer effectively, and generally being less engaging than GPT-4o. Combined, this is not the success OpenAI envisioned. Over the last week or so, social media platforms like Reddit have been flooded with users claiming their prompts "fix" GPT-5. While the ones I've tried haven't had much success, I think I might've finally found the answer to all your ChatGPT problems. The prompt in question, which you'll find below, is from a Reddit user called michel_333 who claims they've created "the ultimate prompt to unlock 100% of ChatGPT-5's power." The AI community loves hyperbole, and at first, I had my reservations. However, I've been using this GPT-5 prompt for the last few hours, and I'm thoroughly impressed with how efficient it makes OpenAI's latest flagship model. Hear me out, this prompt rules. Like seriously, I've been using it this morning, and I'm really happy with the results I'm getting from GPT-5. Once you ask GPT-5 to respond in that manner, you'll notice your prompts will receive a direct answer, followed by a concise step-by-step with no fluff, and finally, a short action plan based on your query. ChatGPT will also let you steer the tone or depth of each response depending on what you're asking it. For example, you can ask for a "Quick hit" and you'll get a response in 3-5 bullet points. Other options include "Deep Dive", "Plan Mode", "Show Math", and "Editor Hat". After using this prompt, I've decided to make it my go-to ChatGPT experience by adding it to the Personalization tab in Settings. I highly recommend you give this a try and be sure to leave a comment below with your experience of this excellent prompt.