Practical copy-and-paste endings for clearer prompts, more confident AI use, and better everyday output.
Add one line to the end of your normal prompt and the answer gets sharper, more honest or easier to use.
I’ve trained more than 15,000 people to use AI at work. These are the 23 endings I actually use and teach, each with a real prompt I used it on.
Find the ending that matches your problem, copy it, and paste it at the end of your prompt.
Use one or two at a time. Stacking too many makes the request confusing.
Not sure which one you need? Jump to the chooser at the bottom.
These endings make the AI check what you actually want before it writes anything.
Use it when you’ve made a quick request with little context.
“Before you write the final response, ask me up to five clarifying questions to ensure you understand my exact needs.”
Why it works: It replaces guesswork with a short conversation about your actual goal.
How I used it:
“I want to redesign my LinkedIn banner. I want to keep my photograph and the wording, but the current design doesn’t feel strong enough. Before you write the final response, ask me up to five clarifying questions to ensure you understand my exact needs.”
What happened:
The AI asked me what I wanted people to understand immediately from the banner, which elements I wanted to keep, what I disliked about the current design and the visual direction I wanted. It then used those answers to suggest a new design.
Use it when your request could be read more than one way.
“If any part of my request is ambiguous, stop and ask me to clarify before proceeding.”
Why it works: It cuts the risk of a confident answer to a question you didn’t ask.
How I used it:
“Go through my training bookings for 2025 and tell me which type of work made me the most money. If any part of my request is ambiguous, stop and ask me to clarify before proceeding.”
What happened:
The AI stopped because “type of work” could mean AI vs social media, keynote vs training, corporate vs charity, or full-day vs short sessions. It asked me what I meant before analysing the bookings.
Use it when you’re asking for a report, proposal, article or anything long.
“Give me a brief outline of what you plan to write, and wait for my approval before writing the full draft.”
Why it works: You fix the direction in 30 seconds instead of rewriting 1,000 words.
How I used it:
“I want to create a brochure explaining my Story Lens method to potential corporate clients. Give me a brief outline of what you plan to write, and wait for my approval before writing the full draft.”
What happened:
The AI gave me the proposed structure first. I could change the order, remove sections I didn’t need and spot what was missing before it spent time writing the full brochure.
Use it when the answer feels generic.
“What additional context or information do you need from me to make this answer 10 times better?”
Why it works: It makes the AI name the gaps in your brief, which is usually why the answer was generic.
How I used it:
“I want to create a YouTube strategy that helps me grow an audience and generate enquiries for my training and speaking work. What additional context or information do you need from me to make this answer 10 times better?”
What happened:
Instead of giving me a generic YouTube strategy, the AI asked about my target audience, what I wanted viewers to do after watching, the type of videos I wanted to make, how often I could publish and how I would measure success.
Use it when you have a half-formed idea you want to shape.
“Act as an expert consultant. Interview me one question at a time to help me refine this idea.”
Why it works: One question at a time means you think properly about each answer, rather than skimming a wall of questions.
How I used it:
“I have an idea for a framework called Story Lens that helps people make sense of significant change and decide what to do next. Act as an expert consultant. Interview me one question at a time to help me refine this idea.”
What happened:
Instead of developing the framework for me, the AI asked me one question at a time about the problem it solved, who it was for, how the stages worked and what someone could actually do differently after using it. My answers shaped the framework.
Use it when there’s more than one sensible way to tackle the task.
“List three different approaches we could take, and ask me which one you should pursue.”
Why it works: You make the strategic choice, and the AI does the work after.
How I used it:
“I want to reconnect with companies I’ve worked with before and generate new training and speaking opportunities. List three different approaches we could take, and ask me which one you should pursue.”
What happened:
The AI suggested three routes: direct email outreach, LinkedIn reconnection, and a content-led approach. I chose the route I wanted before it built the actual plan.
These endings stop the AI simply agreeing with you. They make it challenge your idea, show its assumptions, look at the problem from more than one angle and even critique your prompt.
Use it when you’re too close to an idea to see its weaknesses.
“Adopt the persona of a sceptical critic and point out the flaws in this idea.”
Why it works: AI tends to agree with you by default, and this asks it for challenge instead.
How I used it:
“I’m thinking of positioning Story Lens around helping people make ‘better judgements’ about change. Adopt the persona of a sceptical critic and point out the flaws in this idea.”
What happened:
The AI challenged “better judgement” as too abstract. A potential client couldn’t easily see what they were buying or what would be different afterwards. That pushed me towards defining a more tangible outcome for Story Lens.
Use it when you’re about to present a plan or recommendation.
“Play devil’s advocate and give me one strong reason why this solution might fail.”
Why it works: It forces the AI to identify the strongest weakness, rather than giving you a long list of unlikely problems.
How I used it:
“I’m planning to post on LinkedIn five times a week to generate more conversations with potential clients. Play devil’s advocate and give me one strong reason why this solution might fail.”
What happened:
The AI pointed out that posting more content wouldn’t necessarily create commercial conversations. If the right buyers weren’t seeing or engaging with it, I could publish five times a week without generating a single enquiry.
Use it when you’re planning a launch, a change or a big decision.
“Assume I have overlooked a critical risk in my plan. Ask me how I intend to deal with it.”
Why it works: It looks for blind spots, then makes you think the answer through yourself.
How I used it:
“I’m planning to launch a weekly YouTube series about practical uses of AI and publish one new video every week. Assume I have overlooked a critical risk in my plan. Ask me how I intend to deal with it.”
What happened:
The AI asked what I would do if producing each video took longer than expected and I couldn’t maintain the weekly schedule. It exposed a risk I hadn’t planned for and made me think about creating a repeatable production process.
Use it when the answer depends on things you haven’t told it.
“Explicitly state any assumptions you had to make to reach this conclusion.”
Why it works: You can check the assumptions are sensible before you rely on the answer.
How I used it:
“Based on my current LinkedIn strategy, tell me which type of content is most likely to attract the audience I want. Explicitly state any assumptions you had to make to reach this conclusion.”
What happened:
The AI explained that it had assumed my target audience wanted practical AI advice, that tutorials and FAQs would be more useful to them than general opinion posts, and that my goal was to start conversations rather than simply increase reach. I could then check whether those assumptions were actually correct.
Use it when the answer wasn’t what you wanted and you’re not sure why.
“Critique my original prompt and suggest a better way I could have asked this question.”
Why it works: It fixes today’s answer and teaches you to write better prompts.
How I used it:
“Give me some ideas for videos about generative AI. Critique my original prompt and suggest a better way I could have asked this question.”
What happened:
The AI pointed out that I hadn’t specified the audience, type of videos, desired outcome or what made a useful idea for me. It then rewrote the prompt with those details included, giving me a much stronger starting point for future video research.
Use it when a decision affects different people, teams or customers.
“Give me a deep-dive analysis covering at least five different perspectives.”
Why it works: It pushes the AI past the first obvious angle.
How I used it:
“A company is considering allowing employees to use generative AI across the business. Give me a deep-dive analysis covering at least five different perspectives.”
What happened:
The AI looked at the decision from the perspectives of employees, managers, customers, IT and legal. Each had different opportunities and concerns, giving me a much broader picture than a standard list of pros and cons.
Finding these useful?
The Prompt Endings Library has many more, sorted by topic, and I add new ones whenever I find one that works.
These endings make the AI show its working, so you know what to check.
Use it when the answer includes facts, figures or dates.
“If you are uncertain about any fact or figure, flag it clearly with a warning.”
Why it works: It separates what the AI is confident about from what it’s guessing.
How I used it:
“Give me five statistics I can use in a presentation about how businesses are adopting generative AI. If you are uncertain about any fact or figure, flag it clearly with a warning.”
What happened:
The AI flagged figures where it wasn’t confident about the source, date or exact percentage. Instead of treating every statistic as equally reliable, I knew which ones I needed to verify before putting them on a slide.
Use it when you’re about to rely on an answer for something important.
“Give your answer a confidence score out of 100, and explain why you didn’t score it higher.”
Why it works: The explanation is the useful part, because it names what’s missing. Treat the score as a clue, not proof.
How I used it:
“Review my LinkedIn content strategy and tell me whether it is likely to reach the right audience. Give your answer a confidence score out of 100, and explain why you didn’t score it higher.”
What happened:
The useful part wasn’t the score. The AI explained that it hadn’t seen enough performance data, didn’t know which posts were reaching potential buyers and couldn’t tell whether engagement was turning into conversations. That showed me what evidence was missing.
Use it when you need to check claims or quote them.
“List the main sources or types of information you relied on for this response.”
Why it works: It makes checking easier. Always open important sources yourself, as AI can get them wrong.
How I used it:
“Tell me how LinkedIn’s algorithm currently decides which posts to distribute. List the main sources or types of information you relied on for this response.”
What happened:
The AI separated information from LinkedIn’s own documentation from third-party research, industry commentary and general knowledge. I could see which claims had stronger evidence and which ones I needed to check before using them in my training.
Use it when the task involves calculations, pricing, budgets or step-by-step reasoning.
“Check your maths and logic carefully before giving the final answer.”
Why it works: It prompts a review pass. For anything high stakes, still check the numbers yourself.
How I used it:
“I delivered 47 training sessions last year. 18 were full-day sessions, 16 were half-days and 13 were one-hour sessions. What percentage of my total delivery time came from each type of session? Check your maths and logic carefully before giving the final answer.”
What happened:
The AI had to calculate the actual hours delivered for each format before working out the percentages, rather than simply comparing the number of sessions. It checked that the three percentages added up to 100% before giving me the answer.
Use it when your prompt has several requirements.
“Review your answer against my original instructions and confirm you’ve met every requirement.”
Why it works: It turns your instructions into a checklist before you see the answer.
How I used it:
“Create a 45-minute workout with a 5-minute warm-up, four strength exercises, three cardio exercises and a 5-minute cool-down. I only have dumbbells and a resistance band. Review your answer against my original instructions and confirm you’ve met every requirement.”
What happened:
The AI checked the workout against every requirement and spotted that it had included an exercise requiring a bench. It replaced it with one I could do with the equipment I actually had.
Use it when the task is complex or needs data the AI may not have.
“If you can’t fully complete this request, tell me exactly what is missing or impossible.”
Why it works: It encourages honesty about gaps, instead of an answer that only looks complete.
How I used it:
“Plan a three-day trip to Barcelona for me, including restaurants, attractions and the best times to visit each one. If you can’t fully complete this request, tell me exactly what is missing or impossible.”
What happened:
The AI created the itinerary but flagged that it couldn’t confirm live restaurant availability or guarantee current opening times without checking up-to-date information. I knew exactly what still needed verifying.
Use it when you’ll need to defend a recommendation to someone else.
“Make sure every claim you make is supported by a clear reason.”
Why it works: You can judge the reasoning, not just the conclusion.
How I used it:
“Review my LinkedIn profile and tell me what you would change to make it clearer to potential clients. Make sure every claim you make is supported by a clear reason.”
What happened:
Instead of simply telling me to change my headline, banner or About section, the AI had to explain why each change would make the profile clearer and what problem it was solving. I could then decide whether I agreed with the reasoning.
These endings control focus and format, so the answer is ready to use.
Use it when you’re busy and might not need the full explanation.
“Give me the short answer first, followed by the detailed explanation.”
Why it works: You can stop reading as soon as you have enough.
How I used it:
“I’m deciding whether to focus my YouTube channel on practical AI tutorials or commentary on AI news. Give me the short answer first, followed by the detailed explanation.”
What happened:
The AI gave me its recommendation in the first sentence, then explained the reasoning underneath. I could get the answer immediately and only read the detail if I needed it.
Use it when you’re summarising a long document, report or piece of research.
“Summarise the core argument in one sentence before giving the details.”
Why it works: You get the main point straight away, before the evidence.
How I used it:
“I’ve uploaded a 40-page report about how people are using generative AI at work. Summarise the core argument in one sentence before giving the details.”
What happened:
The AI gave me the report’s main argument in a single sentence first. I knew immediately whether the document was relevant before deciding whether to read the more detailed breakdown.
Use it when the question is broad but you only care about one angle.
“Focus only on [the cost / the customer / the risk] and ignore everything else.”
Why it works: It stops the answer wandering into areas you don’t need.
How I used it:
“I’ve uploaded the results from a customer survey. Focus only on what customers said about our customer service and ignore everything else.”
What happened:
The AI ignored responses about price, products and delivery. It focused only on customer service, identifying the main themes, recurring complaints and positive feedback.
Use it when it’s a task you’ll do again.
“Turn this into a fill-in-the-blank template I can reuse.”
Why it works: A one-off answer becomes something you use every week.
How I used it:
“I’ve written an email to follow up with someone after a meeting. Turn this into a fill-in-the-blank template I can reuse.”
What happened:
The AI kept the structure and tone of my original email but replaced the specific details with fields for the person’s name, what we discussed, the next step and the follow-up date. I now had a template I could use after future meetings.
| If your problem is… | Try endings |
|---|---|
| It answered the wrong question | 1, 2 |
| The answer is generic | 1, 4, 5 |
| You’re asking for something long | 3 |
| There’s more than one way to do it | 6 |
| It agrees with everything you say | 7, 8 |
| Your plan might have holes | 9, 12 |
| You don’t know what it assumed | 10 |
| You keep getting weak answers | 11 |
| You’re not sure it’s right | 13, 14, 15, 16 |
| It missed part of what you asked | 17, 18 |
| You need to defend the answer | 19 |
| It’s too long | 20, 21 |
| It wanders off the point | 22 |
| You’ll need to do it again | 23 |
Pick one ending from this page and use it on your next prompt today.
These are the 23 endings I use most. The Prompt Endings Library has many more, sorted by topic, and I add new ones whenever I find one that works.
Sign up below and you’ll get the full library, ready to copy and paste.
