How Negativity Can Get Positive ChatGPT Results

You are likely one of the millions who have asked ChatGPT a question and ended up with an answer that felt bland and uninspired. You’re not alone. ChatGPT is powerful, no doubt, but there’s a reason why responses can feel like bland, robot-generated rote. AI still isn’t very creative unless you show it how to be.

While it’s tempting to blame the bot, the real issue often lies in its training, or rather, the absence of teaching it what not to say.

So, how do you fix it? The same way we build healthy relationships. We set boundaries. 

I was thinking about this the other day and it occurred to me that marketers have been doing this for years by using “negative keywords” to filter out irrelevant traffic in their Google Ads campaign. Now, we can use the same logic to guide AI on what language to avoid, therefore dramatically improving its responses.

Using “Negative Keywords” for AI

For context, negative keywords in ads keep irrelevant searches from triggering clicks. If you’re selling high-end running shoes, for example, you might use “cheap” as a negative keyword to avoid clicks from bargain hunters. With AI, the approach is similar: teaching it what language to avoid can lead to responses that sound more engaging and specific.

The “Not Like This, Like This” Approach for AI

AI responds well to specific guidance: "Not like this, like this." This method isn’t just about adding examples of good responses; it’s about flagging the language you don’t want, which helps the AI avoid generic phrases like “amazing” or “fantastic.” This step sharpens the model’s language, making responses feel more intentional, relevant, and less like a bot on repeat.

Simple Tips to Improve Your ChatGPT Prompts

Want to get more tailored responses? Try out this “not like this, like this” method with your own prompts. Here are three examples to level up your ChatGPT experience:

1. Avoiding Generic Language: “Explain the benefits of regular exercise without phrases like ‘exercise is important.’ Start with something unique.”

2. Requesting Fresh Vocabulary: “Describe New York City without using words like ‘bustling’ or ‘exciting.’ Focus on details that highlight its vibe.”

3. Balanced, Non-Sensational Tone: “Give a balanced view on AI’s impact on jobs. Avoid words like ‘revolutionary’ or ‘groundbreaking’ and keep it fact-based.”

Setting a few boundaries can make all the difference in AI responses. If your results don’t feel quite right, remember: a bit of negativity can bring a big positive shift.

There you have it…possibly the only time I’ve ever encouraged anyone to be more negative :)