Choosing which prompts to test is one of the most important parts of AI visibility testing.
It is also one of the easiest parts to overcomplicate.
You could generate 100 questions about your business. You could collect keywords, variations and long-tail searches. You could ask an AI tool to suggest dozens more.
But a long list does not necessarily tell you whether you are testing the situations that actually matter to your customers.
A better starting point is the customer journey.
Instead of beginning with:
What prompts should we monitor?
begin with:
What problems, questions and decisions do real customers encounter on their way towards choosing a business like ours?
Then turn a representative selection of those situations into AI visibility tests.
This section brings together the guides and experiments exploring how to do that.
Why Start With the Customer Journey?
Imagine an accountancy firm in Birmingham.
It could test:
- best accountant Birmingham;
- recommended accountant Birmingham;
- good accountant Birmingham;
- top Birmingham accountant.
Those are four prompts.
But they are largely four ways of expressing the same broad need.
Now consider:
- accountant for a Birmingham manufacturing company using Xero;
- accountant for a construction company operating CIS;
- accountant for a property investment business;
- accountant for a growing company needing monthly management accounts and outsourced finance support.
Again, there are four prompts.
But this time they represent substantially different customer situations.
Different services may matter.
Different expertise may matter.
Different evidence may be required.
And different accountancy firms may genuinely be the best fit.
That distinction sits at the heart of our approach:
Prompt variation is not the same as customer-situation coverage.
The purpose of a customer-journey approach is to identify the different situations in which somebody might genuinely need what your business provides.
Start Here: Build Your First Customer-Journey Prompt List
If you are starting from scratch, begin with:
How to Build Your First AI Visibility Prompt List From the Customer Journey
The process begins with the customers the business actually wants to understand.
Ask:
- Who are they?
- What problem brings them to the business?
- What are they trying to achieve?
- What questions arise while they investigate their options?
- What requirements could change which solution is suitable?
- What information might stop them choosing one provider over another?
- What would they need to know as they move closer to a decision?
You can then turn a representative selection of those needs into prompts.
The objective is not to describe every customer perfectly.
It is to build enough coverage to investigate whether AI connects your business with the kinds of problems it is genuinely equipped to solve.
Think in Customer Needs, Not Just Search Phrases
One underlying customer need can be expressed in many different ways.
Someone looking for an accountant might mention:
- their industry;
- location;
- business size;
- accounting software;
- payroll;
- tax;
- management reporting;
- budget;
- pricing preferences;
- or several of those requirements together.
Trying to discover every exact sentence somebody could use is unlikely to be practical.
Fortunately, it may not be necessary.
The more useful objective is to identify the underlying customer-need themes and choose sensible prompts representing them.
Read:
You Don’t Need to Discover Every AI Search Query: Test Customer-Need Themes Instead
The central question becomes:
How visible are we when potential customers describe situations our business is particularly well suited to solve?
That is much more commercially meaningful than asking whether the business appears somewhere across a huge collection of loosely related searches.
Turn Customer Situations Into an Actual Test Set
Once you have identified the customer needs worth exploring, you need to turn them into usable prompts.
Our more detailed guide is:
How to Build an AI Visibility Query Set Around Real Customer Situations
A useful prompt does not need to be long simply for the sake of being long.
What matters is whether the information changes the problem.
For example:
Recommend an accountant in Birmingham.
is very different from:
I run a manufacturing company in Birmingham, use Xero and need payroll and monthly management accounts. Which accountants should I consider?
The second prompt introduces requirements that may genuinely affect suitability.
That gives the AI visibility test something more useful to investigate:
Does the AI recognise which businesses are particularly relevant when the customer explains what they actually need?
The Type of Detail Can Matter More Than the Amount
Our experiments have also suggested an important distinction between adding more detail and adding decision-relevant detail.
In one CRM experiment, changing the customer from a generic small business to a recruitment agency materially changed the recommendations.
Adding that the agency specialised in temporary workers changed the requirements further. Features such as scheduling, worker availability, timesheets, compliance and pay-and-bill became more important.
But simply adding that the company had been trading for five years produced a much smaller change.
That led to a useful working principle:
The kind of specificity may matter more than the amount of specificity.
Read the experiment:
Does the Type of Detail in an AI Search Matter? We Tested It
This matters when constructing customer-journey prompts.
Do not make a prompt longer simply because AI allows you to type more.
Ask instead:
What customer circumstances genuinely change what a suitable solution looks like?
Those are often the details worth testing.
A Customer Journey Can Contain Several Information Needs
A detailed AI search may look like one question to the customer while containing several separate issues the AI system needs to investigate.
Consider:
I’m looking for an accountant in Birmingham for a manufacturing business that uses Xero. Which accountancy firms should I consider, and why?
Potentially relevant subjects include:
- Birmingham accountancy firms;
- manufacturing experience;
- Xero expertise;
- accounting requirements particular to manufacturers;
- evidence of relevant experience;
- pricing;
- management reporting;
- stock and costing;
- payroll;
- and other related needs.
This is where query fan-out becomes particularly interesting.
Read:
What Query Fan-Out Means for AI Visibility Testing
Query fan-out provides another way to think about the different questions sitting underneath a larger customer problem.
For AI visibility testing, that can help us move beyond:
Which single prompt should we rank for?
towards:
Which connected customer questions and information needs should we understand our visibility across?
The prompt is the thing we use to perform the test.
The customer problem is what we are ultimately trying to understand.
Query Fan-Out Can Also Help You Discover New Tests
We explored this practically using Otterly.ai’s free query fan-out tool.
Starting with a Birmingham manufacturing-accountant question, the tool suggested related areas that could be investigated.
We then tested a selection of them.
Different businesses appeared depending on whether the question focused on subjects such as:
- general suitability;
- pricing;
- reviews;
- case studies;
- manufacturing costing;
- inventory;
- software implementation;
- or outsourced financial direction.
Read:
Can Query Fan-Out Help You Choose Better AI Visibility Tests? We Tested Otterly.ai
The important point is not that every suggested fan-out should automatically become a permanent monitored prompt.
It should not.
A fan-out tool gives you possible directions to investigate.
Human judgement still needs to decide:
- Is this genuinely relevant to our customers?
- Does it represent a different need?
- Is it commercially important?
- Does it tell us something our existing prompts do not?
- Is it worth monitoring over time?
Used that way, query fan-out can help broaden your understanding of the customer journey without turning prompt research into an exercise in collecting as many queries as possible.
Realistic Customer Scenarios Can Reveal Different Competitors
One reason customer-journey testing is so useful is that increasingly specific needs can change the apparent competitive landscape.
A customer looking for:
dentist Leeds
may create one recommendation problem.
Someone looking for a private Leeds dentist because they have severe dental anxiety after a bad experience, have avoided treatment for years, want a non-judgemental approach, may require conscious sedation and would like clear published fees creates a much more specific one.
We tested exactly that kind of situation.
Would AI Mode Find the Dental Practice With the Strongest Evidence? A Five-Run Test
The same principle has appeared in tests involving accountants, commercial cleaners and CRM software.
The customer is not simply looking for a category of business.
They are looking for a business that can solve their particular version of the problem.
That is what customer-journey testing attempts to capture.
Not Every Stage of the Journey Has the Same Commercial Value
A customer can ask questions at very different points in their journey.
For example:
What is a white-label help desk?
and:
What should an MSP look for when choosing a white-label help desk provider?
could both be relevant to the same company.
But they do not necessarily represent the same commercial situation.
The first person may simply be learning about the service.
The second may be getting much closer to evaluating providers.
That does not mean early-stage questions are unimportant.
They can help people discover a subject, understand a problem and develop requirements.
But when measuring AI visibility, it is useful to know where in the customer journey a prompt sits and why appearing for it matters.
This also connects directly with the RAMP Framework.
RAMP begins with Relevance:
Was this a customer search where appearing actually mattered?
A high visibility percentage across low-value prompts is not automatically more useful than strong visibility across a smaller number of customer situations closely connected to the services a business wants to sell.
Your Prompt Set Should Cover the Journey, Not Every Wording
Suppose you have identified an important customer need.
There might be dozens of ways to phrase it.
You do not necessarily need dozens of permanent monitoring prompts.
Instead, think about coverage.
Does your prompt set represent:
- different customer types;
- different problems;
- meaningful requirements;
- different levels of buying intent;
- important objections or constraints;
- comparisons;
- and the situations in which your specialist strengths should matter?
A smaller portfolio with genuinely different customer situations can be more informative than a much larger portfolio filled with slight wording variations.
The aim is representative coverage rather than volume for its own sake.
Exploration and Monitoring Are Different
There is also an important distinction between using prompts to explore a customer journey and using prompts for ongoing monitoring.
During exploration, variation is useful.
You might change:
- customer type;
- industry;
- budget;
- location;
- business model;
- desired features;
- previous experience;
- urgency;
- or another requirement.
You are trying to learn:
What changes the answer?
Once you identify a commercially important prompt that you want to monitor over time, consistency becomes more important.
If the wording and requirements continually change, it becomes much harder to know whether changes in the results reflect changing AI visibility or simply a different question.
So:
Explore widely.
Then monitor important prompts consistently.
Both activities matter, but they answer different questions.
Customer-Journey Research Is an Ongoing Process
Your first prompt list is not the finished product.
You might begin with 15 prompts and discover that:
- some are too broad;
- several overlap;
- some represent customers unlikely to buy;
- one customer segment matters much more than expected;
- an unexpected competitor repeatedly appears;
- a particular requirement changes the recommendations substantially;
- query fan-out reveals a useful new subject;
- or an apparently promising prompt produces very little useful information.
That is not a failure of the original research.
It is part of the research.
The process should be:
Map the customer journey → choose representative prompts → test them → learn from the results → refine the prompt set.
As your understanding of the customer and the market improves, the quality of the testing should improve with it.
What We Learned From OtterlyAI’s Prompt Framework
We also compared this approach with the prompt-selection framework provided by OtterlyAI during our trial of its monitoring software.
Read:
One useful lesson is that prompt research should not be treated as something completed before monitoring begins.
A broad initial set can help you explore the market.
Monitoring then teaches you which prompts deserve to remain, which should be replaced and which new areas might be worth investigating.
That is much more realistic than expecting any tool—or any initial brainstorming exercise—to produce a perfect permanent prompt portfolio on day one.
A Simple Customer-Journey Process
If you want to apply this approach to your own business, start here:
1. Choose one important customer group
Do not try to map everybody at once.
Pick a group you genuinely want to attract.
2. Identify the problem that brings them to the market
What are they trying to solve, change or achieve?
3. Map the questions and decisions they encounter
What do they need to understand before they can choose a solution?
4. Identify the circumstances that change suitability
Industry, budget, location, software, urgency, previous experience and specialist requirements can all matter—but only when they genuinely change the problem.
5. Create representative prompts
Turn those customer situations into natural questions someone could reasonably ask an AI system.
6. Test them
See which businesses appear, how the recommendations change and what evidence AI seems to use.
7. Refine the set
Keep useful core prompts. Remove weak or duplicative ones. Develop promising themes and add new customer needs as you discover them.
Then repeat the process.
The Customer Journey Comes Before the Dashboard
AI visibility software can tell you how often a business appears across the prompts you give it.
But software cannot make an irrelevant prompt commercially important.
That means there is a question that comes before any visibility score:
Are we measuring the customer situations we actually care about?
The customer journey helps answer that question.
Once those situations have been identified, AI visibility testing can investigate:
- whether your business appears;
- how consistently it appears;
- how competitors perform;
- whether the business is recommended or merely mentioned;
- how accurately it is described;
- and how those patterns change over time.
Automated monitoring can eventually make that much easier to manage at scale.
But deciding what is worth monitoring remains a human judgement.
Where to Go Next
If you are new to this approach, I recommend this sequence:
- How to Build Your First AI Visibility Prompt List From the Customer Journey — map an important customer group and turn its journey into an initial prompt list.
- You Don’t Need to Discover Every AI Search Query — understand why representative customer needs matter more than finding every possible wording.
- How to Build an AI Visibility Query Set Around Real Customer Situations — develop a more structured testing portfolio.
- Does the Type of Detail in an AI Search Matter? — see why decision-relevant details may matter more than simply making prompts longer.
- What Query Fan-Out Means for AI Visibility Testing — explore the different information needs that may sit behind one customer question.
- Can Query Fan-Out Help You Choose Better AI Visibility Tests? — see how we used fan-out suggestions as starting points for further testing.
From there, move into the main AI Visibility Testing section to learn how to repeat, record and measure the prompts you have chosen.
The principle running through all of these guides is straightforward:
Do not begin with the biggest possible list of AI prompts. Begin with the customer problems that matter, build representative tests around them, and allow what you learn to improve the questions you monitor.