Key Takeaways
- Spend the questions like a budget, not a wishlist. Twenty-five slots is enough to cover five distinct jobs properly and not enough to cover all of them plus everything you are curious about. Every slot you add somewhere comes out of somewhere else.
- Weight it toward the questions that name nobody. The questions where the buyer does not know you exist are the ones with something to learn. Your own business name is a check, not a test.
- Use your buyers' words, not your category's words. Nobody types the phrase on your services page. They type the symptom, the city, and the price question, in the order they think of them.
- Fix the set before you run it. A question set that changes between runs measures nothing. Lock the wording, then only change it deliberately and note the date you changed it.
- Keep two slots for questions you expect to fail. A set you always pass is a set that has stopped telling you anything. The uncomfortable question is where the next piece of work comes from.
You can ask an AI assistant anything about your business. You cannot ask it everything and still have a result you can repeat next month.
So the set gets fixed at 25, and picking those 25 is the part that decides whether the answer is useful or just interesting. Most people fill them with questions about themselves, which is the one thing the exercise cannot tell them much about.
Here is the split we use, why each type earns its slots, and how to word them so the results survive being run again.
How do you pick the 25 questions that matter most?
Cover five distinct jobs, weight the set toward questions where the buyer does not know you exist, and write every question the way a buyer would type it. Then lock the wording.
The five jobs are direct discovery, problem-first, comparison, qualifying, and branded. Each one answers something the other four cannot, and a set missing any of them has a blind spot you will not notice, because a blind spot looks exactly like a passing grade.
The weighting matters more than the exact wording. A set that spends most of its slots on your own name will tell you your own name works, which you already knew.
The five jobs a question set has to cover
Direct discovery is the baseline: the buyer wants what you sell, names the city, and has no idea who you are. This is the harshest and most honest test you can run, and it deserves the largest share.
Problem-first questions catch the buyer earlier, when they have a symptom and no category yet. Someone typing "my team keeps missing deadlines" has not decided to hire a coach, and whoever gets named in that answer is in the conversation before the shortlist exists.
Comparison and qualifying questions do the diagnostic work. Comparison shows you who gets named beside you and why. Qualifying, which covers price, timeline, and credentials, shows whether the facts you publish match the criteria being used to sort providers. Branded questions come last and are a check rather than a test.
The bar is one set of 25 slots, divided. Widening any segment narrows another, which is the decision the set is actually making. The split shown is the one AnswerHalo uses as a starting point; a niche with heavy price shopping earns more qualifying slots, and a crowded local market earns more direct discovery.
How to split 25 slots across the five
Start at 8 direct discovery, 6 problem-first, 5 comparison, 4 qualifying, and 2 branded. That is 25, and it puts more than half the set on questions where nobody has heard of you.
Adjust for your market rather than your curiosity. A niche where buyers shop hard on price earns extra qualifying slots. A crowded city market earns extra direct discovery. A field with two obvious approaches, like group programs against one to one, earns extra comparison.
Take the extra slots from branded first, then from comparison. Never take them from direct discovery, because that is the only type that measures the thing you are actually trying to fix.
| Type | Slots | What it tests | What it cannot tell you |
|---|---|---|---|
| Direct discovery | 8 | Whether you appear at all to a buyer who does not know you | Why you were left out |
| Problem-first | 6 | Whether you reach buyers before they know what to hire | Whether they would have picked you once they did |
| Comparison | 5 | Who gets named beside you, and on what grounds | How large the market beyond those names is |
| Qualifying | 4 | Whether your published facts match the criteria being applied | Whether the buyer agrees with those criteria |
| Branded | 2 | What an assistant says about you once it has your name | Anything at all about discovery |
Getting the wording right
Write each question the way a buyer types it, which is shorter, blunter, and less professional than anything on your website. "Is a business coach worth the money" beats "evaluating the return on executive coaching engagements" every time.
Three rules keep the wording honest. No industry terms your buyers do not use. No leading phrasing that plants your own positioning inside the question. And one variable per question, so that when an answer changes you know which word caused it.
Pull the raw material from where your buyers already speak. Your own inbox, the first two minutes of discovery calls, and the questions people ask in the communities your niche lives in. A starting list of self-directed ones is in questions to ask ChatGPT about your own business.
Structured markup is a separate lever from wording, and worth knowing about: schema.org's FAQPage type exists so a question and its answer stay attached to each other in a form a machine can read. That helps a page be quoted. It does not change which questions are worth asking.
Locking the set so the results mean something
Freeze the wording, then run the same 25 the same way each time. A question set that drifts between runs produces a before and an after that cannot be compared, which is worse than not measuring.
Record the date, the assistant, and the exact wording alongside the results. Google's guidance on AI features in Search is explicit that these systems change without notice, so your own record of what you asked and when is the only stable reference point you will have.
Change the set deliberately, no more than once or twice a year, and treat each change as a new baseline rather than a continuation. The method for running it is in how to check if AI recommends you, and the patterns that keep showing up across our own runs are in what 25-question test runs keep showing us.
If you would rather have the set drafted and run for you, the free check runs 25 real buyer questions from your niche through ChatGPT at $0 and emails a plain-English readout within the hour. It runs as soon as you ask for it.
Questions we hear the most
How do you pick the 25 questions that matter most for your business?
Cover five jobs: direct discovery, problem-first, comparison, qualifying, and branded. Weight the set toward questions where the buyer does not know you exist, write every one in your buyers' words, then lock the wording so runs stay comparable.
Why 25 questions and not 10 or 100?
Ten cannot cover five distinct question types with enough of each to see a pattern. A hundred takes long enough that nobody repeats it. Twenty-five is the point where the set is broad enough to trust and short enough to re-run.
Should I include my own business name?
Include it twice, no more. It confirms what an assistant says about you once someone already has your name, which is worth knowing, but it cannot tell you anything about whether you get discovered.
How specific should the location be?
As specific as your buyers actually type. For most local services that is the city, sometimes the neighborhood. A state or a region is usually too broad to produce the shortlist a buyer would act on.
Can I change the questions later?
Yes, deliberately and with a note of the date. Changing wording between runs makes the before and after incomparable, so treat an edit as starting a new baseline rather than continuing the old one.
What does the free 25-question check include?
It runs 25 real buyer questions from your niche through ChatGPT and emails a plain-English readout: your score, who gets named where you are not, and the first fix to make. It costs $0.
Does AnswerHalo write the 25 questions or do I?
We draft them from your niche and your buyers, and you get to see them. A set built for one niche is not reusable in another, so the questions are drafted per business rather than pulled from a template.
Can I run the free check today?
Yes. It runs as soon as you ask for it, and the readout lands within the hour between 7am and 5pm Central. Writing your own 25 questions with the split in this post takes about an hour and can happen today.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →