How do you use AI without becoming less creative, more stupid, or deskilled? One strategy is to check whether your prompts are focused on an endpoint that you’re trying to get to, or on building the skills that will get you there — what I call “journey prompts” and “destination prompts”.

In creative work, for example, you might be looking for an idea, or aiming to produce a story or image. In journalism or learning, a ‘destination’ might be key facts, or an article or report.

But prompts that focus only on those destinations are less likely to help us learn, more likely to deskill us — and more likely to add errors to our work.

To avoid those pitfalls, it is better to focus on how we get to those destinations. What, in other words, are the journeys?

Focusing on the journey

In creative work, we arrive at ideas through strategies such as brainstorming, the SCAMPER technique, or a subject-specific approach such as the seven angles approach to data-driven stories.

We then have to identify what criteria makes a good idea, choose the best one(s), and develop it, refining and adapting in response to new information and challenges.

In journalism or education we find facts by identifying sources, acquiring information from those, verifying it and synthesising it with other information.

We then produce a story, article, or report through a series of steps: planning, drafting, reviewing and editing.

It is in those journeys that we show — and improve — our skill and professionalism. In other words, the journey matters more than the destination.

An example destination and journey prompt

A classic example of ‘destination’ prompting is a simple factual question like “Where was this image taken?“

A ‘journey’ version of the same prompt might be something like: “How would I find out where this image was taken?“

In the slideshow below you can see the difference in results between using the two types of prompt. The destination prompt results in a dead end (and may also be incorrect); the journey prompt provides multiple opportunities for learning and verification:

ChatGPT is not a search engine

Factual questions are the most common example of destination prompting, particularly when large language models (LLMs) such as ChatGPT and Gemini are used as a search engine. But there are three problems with doing this:

1. AI’s tendency to hallucinate false information. This is because generative AI generates sentences — it does not retrieve information in the way search engines do, even though it may provide a source alongside the information (sometimes also hallucinated).
2. The lack of explainability. Even the creators of AI models cannot explain exactly how any response is generated (and don’t ask ChatGPT to explain), and because the accuracy of training material varies enormously, assessing the credibility of any response requires researching outside the model.
3. The potential for passive engagement with responses. One research paper concluded that using AI as a search engine “makes learning more passive than in standard web search, where users actively discover and synthesize information sources themselves.”

So asking ChatGPT or Copilot for a fact like “Who is the head of English at Birmingham City University” doesn’t help us to develop any research skills, and may lead us to an incorrect answer.

A journey prompt equivalent would be to ask a question like “Suggest how I might find out who the head of English is at Birmingham City University. Don't give me the answer.“

This not only avoids deskilling, but also puts a human in the loop — a crucial step in any workflow involving AI.

Destination prompts are OK when it’s not your destination

This doesn’t mean that destination prompts should never be used. Translation and image generation, for example, are two areas where a destination prompt might be preferred to a journey prompt, where those are not skills we need to develop (indeed, they may be tasks that prevent us from spending more time developing the skills we do need to develop).

Other situations might involve a choice between two ‘journeys’. Using AI to summarise a large document, for example, denies us the experience of wading through it ourselves — but it may free up time for other processes: chasing interviewees, sending FOI requests, or analysing data, or pursuing an investigation that we would otherwise never be able to do.

Similarly, we might use AI for research if we intend to use the time freed up to practise and build verification skills — our actual destination.

A useful approach here is to design hybrid prompts that combine the desired ‘destination’ with the context of the bigger journey you are making. I’ll talk more about those in a separate post.

Journey prompts for problem-solving, article writing, image generation and other scenarios

Below is a table of common scenarios in journalism, along with example destination and journey prompts.

A key quality of journey prompts is that they should not only help to reduce deskilling — they should also help the user to think critically about their own processes.

Regardless of the response, designing the prompt itself then becomes a learning process.

Purpose	Destination Prompt	Journey Prompt
1. Geolocating an image	Where was this image taken?	What techniques and tools could I use to find out where this image was taken, and what are the pros and cons of each?
2. Checking an image	Is this image authentic or manipulated?	What techniques could I use to verify whether this image is authentic? What are their strengths and weaknesses?
3. Generating ideas	Give me 10 story ideas about climate change	What brainstorming techniques could help me generate compelling story ideas about climate change for [AUDIENCE]?
4. Writing an article	Write a 500-word article about the housing crisis	What structure could I employ when planning an article about the housing crisis aimed at [AUDIENCE]? Ask me questions to help me clarify my thinking
5. Analysing data	What are the main findings in this dataset?	What analytical approaches should I use to identify meaningful patterns in this dataset?
6. Solving a problem	How do I fix this error in my code?	What debugging steps should I follow to identify and resolve this error?
7. Research synthesis	Summarise the key arguments about AI regulation	What framework could I use to compare and synthesise different perspectives on AI regulation?
8. Generating an image	Create an illustration showing the impact of flooding on this town	What visual elements and composition principles would effectively illustrate the impact of flooding for this story, considering guidelines on the use of AI imagery in news [ATTACHED]?
9. Finding sources	Who are the experts on local government reform?	What strategies should I use to identify credible experts and diverse voices on local government reform?
10. Fact-checking a claim	Is this politician’s claim about crime statistics accurate?	What sources and methods could I use to verify this politician’s claim about crime statistics? What pitfalls should I be careful to avoid?

Designing a journey prompt should also involve thinking critically about the biases of the model itself, and the techniques needed (such as role prompting, RAG and negative prompting) to counterweight those.

Do what you enjoy doing

Identifying the differences between destinations and journeys can be the difference between work that feels fulfilling — and allowing AI tools to take that fulfilment away.

It can be the difference between output that feels stilted and superficial, or alive and original — and in education it can be the difference between a pass and a failure.

AI is notorious for its ability to offer shortcuts. It is synonymous with automation — but its most powerful applications lie outside of that. Role playing prompts, for example, allow you to continue doing the work that you enjoy while adding a mentor or ‘devil’s advocate’ to stimulate and challenge your work.

There is no reason for AI to replace the parts of your work that you enjoy — the creativity, the digging, the feeling of growth, building confidence and increasing mastery. Design your prompts so you can do more of that, not less.