How to keep your AI prompts from going off the rails
Good inputs, good outputs.
By Lindsey Bradshaw, PR freelancer at Lindsey Bradshaw Communications.
AI is only as good as the prompts you feed it. To get the most out of AI, you (the expert) need to write prompts that push the tech beyond its generic answers and into information you can use. Here are some tips to get you started.
- Prompt AI to challenge its own output.
Instead of asking AI for a single answer, ask it to critique its own response. Techniques like “chaining,” are where answers are built upon sequentially to improve coherence. There’s also “tree-of-thought” prompting, which is more a more structured approach to thinking. For example, try a prompt like these:
“Write a pitch for a new product launch on. Now, list three reasons why this pitch would NOT work and suggest ways to improve it.”
“Create a PR strategy for <product/announcement>. Now identify the top three weaknesses of this strategy and suggest how to fix them.”
By rigorously examining and re-evaluating its outputs, AI models ensure that the answers generated are not just fast, but also precise and dependable.
- Use “layered prompts” to build on previous outputs.
Using “layered prompts,” where prompts build upon previous AI-generated responses, enhances the depth and relevance of AI outputs. Layered prompting approaches problem-solving by breaking down tasks into manageable steps, just like a human would. This guides AI through a logical progression, leading to more nuanced and strategic content. For instance:
Start with: “List the top 5 challenges facing the retail media industry.”
Followed by “Create a byline angle based on challenge #2, including three supporting data points.”
And concluding with “Draft a pitch outline for a journalist at <publication> focusing on this angle.”
According to research, decomposing complex tasks into smaller prompts enables AI to generate more accurate and contextually relevant responses. By systematically building upon previous outputs, layered prompting effectively steers the AI toward producing comprehensive and targeted results (source).
- Add constraints to force creativity.
Imposing creative constraints like word count, format, tone or style, makes AI-generated content more inventive. By setting clear boundaries, AI is guided to produce more focused and imaginative outputs. For example:
“Write a press release in 150 words or less using only three-syllable words.”
“Create a headline under 10 words that includes the phrase ‘AI-powered.’”
“Write a LinkedIn post about AI in PR in exactly 3 sentences.”
Applying constraints (or boundaries) directs AI’s generative process and encourages it to explore unconventional information within set limits.
By including factors like word count, tone or format, you provide the AI with a structured framework that promotes the generation of creative and tailored content. For example, specifying the tone and word count ensures the generated content aligns with the narrative style and depth you seek.
- Prompt AI to mimic a real-world style or persona.
Instructing AI to imitate specific writing styles or personas can improve the quality and relevance of its outputs. By directing AI to adopt the tone of a recognizable brand or individual, the generated content becomes more polished and aligned with your standards.
For example, you could try:
“Write a press release in the tone of The New York Times.”
“Draft a media pitch in the style of Apple’s product launch announcements.”
AI models, such as GPT, are trained on diverse datasets with various writing styles and tones. When prompted to mimic a particular style, the AI leverages this training to produce content that mirrors the specified voice.
- Find the blind spots.
When you identify gaps in your messaging or strategy, you can prompt the system to analyze and critique its own outputs, or any material you provide it. This uncovers potential weaknesses or overlooked aspects .Examples include:
“Create a media pitch about our new AI tool. Now identify three objections a journalist might have and suggest how to address them.”
“Write a PR strategy for a new product. What are three potential PR risks that haven’t been addressed?”
By asking AI to self-assess, you tap AI’s analytical ability to predict challenges and refine strategies. This makes your comms more robust by proactively addressing potential criticisms or gaps.
- Refine and shorten your written content.
Use AI to refine and condense written content builds clarity and engagement. By instructing AI to summarize or rephrase text, you can change lengthy passages into concise, impactful messages. Examples include:
“Using as much original content as possible, summarize this 500-word press release in two paragraphs.”
“Rewrite this pitch to make it more direct and engaging. Use as much original writing as possible”
“Reduce this article to bullet points for a social media post.”
AI tools can analyze text and extract key points, providing summaries that retain info while eliminating redundancy. This process not only streamlines content but also makes sure that the key message is intact. AI can condense articles, reports or documents down to the key points instantly, enhancing readability and impact.
