Building a Survey: Question Types, Documents and AI Drafting
Overview
Building a survey is the part of the job that decides everything afterwards. The questions you write determine what your results can tell you, and once the first response arrives those questions are fixed. This page covers how to choose the right question types, how to word them so people actually answer, how documents and sign-off work, and how to use Seaty AI to get a sensible first draft.
At its simplest, the builder answers one question: what exactly are you going to ask, and how will people answer it?
Who uses this: Organisation administrators with the Surveys permission.
Key capabilities:
- Choose the type (survey, questionnaire or form), which changes how the item behaves
- Add questions of seven different types, including documents that require a sign-off
- Add optional helper text under any question except a document, and an explanation line under any answer choice
- Branch single choice answers so people only see the questions that apply to them
- Mark questions required or optional
- Present terms, policies or codes of conduct and record that people read or agreed to them
- Draft a complete starting point with Seaty AI, or load one of the built-in templates
- Reorder questions by dragging, until the first response arrives
For distribution see Sending and sharing a survey. For what the organiser can and cannot see afterwards, see Anonymity and data protection.
How It Works
At a glance: Choose the type, write or generate the questions, check the settings that lock, then open it for responses.
1. Choose what kind of thing you are making
The create dialog opens by asking you to accept the Terms of Service and the Data Processing Agreement, and to confirm that your organisation is the data controller for the responses you collect. You cannot go any further until you do, and the same check is applied again when the item is saved.
Next it asks What would you like to create? A survey gathers opinions about something people attended. A questionnaire gathers information more directly, like preferences and details that help you plan. A form collects required details from members and always records who responded.
This is not a label. It changes real behaviour, most importantly that forms always record who responded and cannot be made anonymous or confidential. It also changes the wording throughout, so a form tells respondents "You have already responded to this form", not "to this survey".
2. Get a first draft
You can start from scratch with no questions, load one of the built-in templates, or ask Seaty AI to draft the whole thing from a short brief. All three routes produce an ordinary editable item. Nothing generated is fixed until responses arrive.
3. Write and refine the questions
Each question has its text, an optional line of helper text, an answer type, and a required setting. Choice questions have their own list of options. Document questions instead carry a body of text and a sign-off mode. You can drag questions into the order you want. On a single choice question you can also give an answer a pathway, so it ends the form or skips ahead rather than always continuing to the next question.
4. Check the settings that will lock
Type, anonymity mode and every question freeze permanently the moment the first response is recorded. Read them once more before you open the item. Everything else, including the title, the introduction, the thank-you message and the opening and closing dates, stays editable forever.
5. Decide whether it belongs to one event
Setup asks Is this for a specific event? Leaving it as Not tied to an event keeps it reusable across everything you run, which is what most things want. Pinning it to one event files it under that event and starts the Send screen on that event's attendees. It does not change how responses are recorded. Unlike the type and the questions, this stays editable forever.
See Surveys, Forms and Questionnaires for what pinning changes.
6. Open it and share it
An item only collects responses once it is open, and it cannot be opened while it has no questions. From there you share the link or QR code, or send it by email.
Think of it this way:
- The type decides how the item behaves and what it is called on screen
- The questions decide what you can learn, and freeze on the first response
- The settings decide who can respond, and mostly stay editable
Choosing the Type
You choose the type first, in the create dialog, and you can still change it under Type and anonymity in Setup right up until the first response arrives.
| Type | Best for | Anonymity |
|---|---|---|
| Survey | Post-show audience feedback, member satisfaction, opinion gathering | Your choice of all three modes |
| Questionnaire | Planning input, preferences, availability, research-style information gathering | Your choice of all three modes |
| Form | Cast lists, membership records, volunteer sign-up, consent capture | Always identified, this cannot be changed |
Why do forms always record who responded? A form exists to collect operational details from named people. A cast availability form or a membership record is worthless if you cannot tell whose it is, so Seaty removes the choice rather than letting you build something unusable. The anonymity picker is hidden for forms and the setting is enforced when the form is saved.
Forms also open on the Responses table rather than the summary charts, because for a form the individual rows are the point.
Question Types
There are seven answer types, chosen from a grid of tiles when you add or edit a question.
| Type | What it shows the respondent | Use it when |
|---|---|---|
| Single choice | Radio buttons, pick one option | The options are genuinely exclusive: age band, preferred night, yes or no |
| Multiple choice | Tick boxes, pick any that apply | People can honestly belong in several answers: what they enjoyed, how they heard about you |
| Rating | A row of numbers with fixed endpoints | You want a comparable score for one thing: the venue, the interval, the sound |
| Recommendation | A row from 0 to 10 | You want one headline loyalty number for the whole experience |
| Free text | An open comment box | You do not yet know what the answers will be, or you want people's own words |
| Confirmation | A single tick box | One simple declaration: "I hold a current DBS check" |
| Document | A titled card with a document to open | You need people to read, acknowledge or agree to something |
Single choice and multiple choice
Every choice question needs at least two options. You cannot remove the last two. If you switch a question to a choice type and it has no options yet, two empty ones are created for you to fill in. Switching between single and multiple choice keeps the options you have already written.
Each option can carry its own short explanation line, shown underneath it. This is where you put the detail that would otherwise bloat the option itself. Instead of an option reading "Concessionary rate (over 65, under 18, students, or in receipt of certain benefits)", the option reads "Concessionary rate" and the explanation carries the rest.
Why keep options short? On a phone the options are the thing people scan. Long options force wrapping and people stop reading them properly, which quietly biases your results towards whichever answer is shortest.
Watch for two common mistakes. Single choice questions that are not really exclusive ("what did you enjoy most?" when people enjoyed three things) push people into an arbitrary answer. And a choice question with no honest home for a respondent produces either a skipped question or a wrong one, so include an escape option where one is warranted.
On a single choice question, each answer can also carry a pathway that ends the form or skips ahead to a later question, so people only see the questions that apply to them. See Question pathways below.
Rating
A rating question shows a single row of numbered buttons from 1 up to the maximum you choose, anchored Poor at one end and Excellent at the other. On a narrow screen the row scrolls sideways rather than wrapping, so it always reads as one scale.
You choose the maximum from four presets: 1 to 3, 1 to 5, 1 to 7 or 1 to 10. The default is 1 to 5.
Which scale should you use? 1 to 5 is the safe default and is what most people are used to. 1 to 3 works when you only care about "poor, fine, great" and want a fast survey. 1 to 10 gives you finer resolution but people use it inconsistently, and it looks visually heavy on a phone. Whatever you pick, use the same scale for every rating question in the same item, or your averages will not be comparable to each other.
The fixed anchors are why rating questions should always be about quality of something. Asking "how likely are you to return?" on a scale labelled Poor to Excellent reads oddly. That is what the Recommendation type is for.
Recommendation
The Recommendation type is the standard Net Promoter question. It is fixed at 0 to 10, anchored Not at all likely and Extremely likely, and has nothing to configure.
Results treat it as an NPS score: 9 and 10 count as promoters, 7 and 8 as passives, 0 to 6 as detractors, and the score is promoters minus detractors as a percentage of everyone who answered. Passives add nothing to the top of that sum but they are still counted in the total, so a run of 7s and 8s pulls the score down. That is why the scale is not adjustable. The bands only mean anything on the standard 0 to 10 scale.
Use one, at most. It is a single headline number for the whole experience, not a per-item rating. If you want per-item scores, use rating questions.
Free text
An open comment box. This is where the genuinely useful material comes from, and also where the effort of answering is highest.
Practical guidance:
- One or two free text questions per item is usually right. Three or more visibly increases the number of people who give up part way.
- Put them near the end, after the quick questions. People who have already invested a few taps are more likely to write something.
- Make them optional unless you genuinely cannot use the response without them.
- Ask something specific. "Any other comments?" produces blank boxes. "What one thing would have made the evening better?" produces answers.
In the results summary, free text answers can be searched and sorted by most recent or longest, and each one opens the person's full submission. The summary shows at most the 500 most recent answers per question, with a notice above them saying so when there are more.
Confirmation
A single tick box for one declaration. It has no options and no scale.
There is one behavioural detail worth knowing: an unticked confirmation records nothing at all. In the results, the "not confirmed" figure is worked out by subtracting the people who ticked it from everyone who responded. That is normal and correct, but it means a confirmation is a poor way to capture a genuine three-way answer. If "no" is a meaningful answer you want to count, use a single choice question with explicit Yes and No options instead.
If a confirmation genuinely matters (a declaration, a consent, an agreement), consider whether it should be a document question instead. A confirmation records that a box was ticked. A document records exactly what wording was on screen when it was ticked.
Question Pathways
By default a survey is a straight line: everyone sees every question, in order. Pathways let a single choice answer change what comes next, so people only see the questions that actually apply to them.
Each answer on a single choice question can do one of three things:
- Continue to the next question. The default. There is nothing to set.
- End the form here. Choosing this answer finishes the survey. Use it for a clean opt-out near the top: someone who answers that they are not renewing, not taking part, or did not attend does not need the rest of the questions.
- Skip to a later question. Jump straight to a question further down, skipping the ones in between. Use it for a follow-up that only some answers need.
You set this in the question editor. Once an answer has some text, an "After this answer" control appears beneath it, reading "Continues to the next question" until you change it. Only single choice answers can branch. Multiple choice, ratings and every other type always continue.
Pathways only ever move forwards
A skip can only point at a question that comes later in the list, never an earlier one, and a survey can never loop back on itself. This is deliberate. It keeps the survey readable from top to bottom, and it makes it impossible to build a loop a respondent could get trapped in.
Because of that, the question order and the pathways are tied together. If you drag a question above an answer that skips to it, or delete a question something skips to, that pathway no longer makes sense. Seaty tells you before it happens: reordering that would break a pathway asks you to confirm first, and deleting a question that other answers lead to warns you that those pathways will be turned off. Nothing is changed silently behind your back.
What the respondent sees
When a survey branches, the respondent does not see all of it at once. They see the questions up to the first answer that decides a pathway, and the rest appear as they answer. An answer that ends the form finishes the survey there; an answer that skips ahead means the questions in between never appear at all. Submit only becomes available once their answers have carried them to the end of their particular path.
Results only count a branched question against the people who actually reached it, not everyone who responded, so a follow-up that half your respondents never saw is not treated as though the other half skipped it. On an anonymous or confidential survey, a branched question that fewer than five people reached is hidden for the same small-numbers reason the whole survey is hidden below five responses. See Reading results.
Note: Use pathways where they genuinely help, not everywhere. Most good surveys are a straight line. A pathway earns its place when a whole run of questions does not apply to some answers, or when an early answer should simply end the form. Seaty AI and several of the built-in templates already set up a pathway where the survey they are drafting calls for one.
Helper Text and Limits
Any question except a document can carry helper text, a short line of guidance shown underneath it. Helper text and document bodies are rich text, so you can use headings, bold, italic, lists and colour.
Use helper text to remove ambiguity, not to add words. Good helper text says "Think about the whole visit, not just the performance" or "Leave blank if you are not sure". Bad helper text restates the question.
| Field | Limit |
|---|---|
| Helper text | 500 characters |
| Explanation under an option | 160 characters |
| Document body | 20,000 characters |
| Introduction | 2,000 characters |
| Thank-you message | 2,000 characters |
Why is the option explanation so short? It sits under a single answer choice on a phone. Anything longer than a sentence turns a tidy list of choices into a wall of text, and the choices stop being scannable.
Required and Optional
Every question except a document has a single setting: "Yes, an answer is required" or "No, this question is optional". Required questions show a red asterisk to the respondent, and submission is refused until they are answered. The respondent sees a count of how many questions still need an answer, the page scrolls to the first one, and every outstanding question is flagged in place.
Be sparing. Every required question is a point at which somebody can decide not to bother. A good rule:
- Required for the small number of answers without which the response is useless to you, and for anything on a form you must have on file.
- Optional for everything else, especially free text and anything about the respondent themselves.
Why not require everything? A required question that somebody cannot honestly answer does not produce a better answer. It produces either an abandoned response or a made-up one, and the second is worse than a blank.
Document questions have no required toggle. Whether they are required is derived from the sign-off mode, described next.
Document Questions
A document question shows the respondent a titled document (terms and conditions, a code of conduct, a photography policy, a safeguarding statement) and optionally requires them to sign off on it. This is the strongest record the feature can produce, and it is worth understanding properly before you rely on it.
The respondent sees a card with the title and a View document button. Nothing is readable until they open it.
The four sign-off modes
| Mode | What the respondent must do | Appropriate when |
|---|---|---|
| Information only | Nothing. They may open it or not | Helpful background: directions, a programme note, what happens next |
| Must be read | Open it and scroll to the end. Reaching the end is the sign-off | Something they must have seen, but which they are not being asked to accept: a fire and evacuation notice, a rehearsal schedule |
| Must be acknowledged | Open it, scroll to the end, then tick "I have read this" | Something they must confirm receipt of without agreeing to be bound: a policy update, a privacy notice |
| Must be agreed to | Open it, scroll to the end, then tick "I agree to this" | Something creating an obligation: a code of conduct, membership terms, hire terms, photography consent |
When you switch a question to Document, the mode defaults to Must be agreed to.
A document is required if and only if its mode is not Information only. That is why there is no required toggle. If a required document has not been signed, submission is refused in exactly the same way as any other unanswered required question: a count of how many questions still need an answer, the page scrolled to the first one, and that card flagged.
Why does the tick box stay disabled until they scroll to the end? So that "I agree" means something. Anybody can tick a box next to an unopened document. Requiring the document to have been scrolled through makes the record defensible, which is the entire reason the feature exists. A document short enough not to need scrolling counts as read as soon as it is opened.
Respondents can use Download a copy to save what they saw. The saved copy carries your organisation name, the title, the full text and a short document reference. Once the document has been signed off it also carries a sign-off line saying it was read, acknowledged or agreed to and when, naming the respondent where their name is known. Somebody who downloads a copy before signing off gets no sign-off line.
Choosing between the modes
Pick the lightest mode that does the job. Every step up adds friction, and friction on a document sits right in the middle of the response.
- If nothing turns on whether they read it, use Information only.
- If you need to be able to say "this was put in front of them", use Must be read.
- If you need them to have positively confirmed it, but there is no obligation being created, use Must be acknowledged.
- Only use Must be agreed to where you are actually asking for agreement. Overusing it devalues it, and makes a long form feel like a contract signing.
How versions are frozen
This is the part that makes documents trustworthy.
Every distinct combination of title, document text and sign-off mode is stored as its own version. When somebody signs off, their answer is tied to the exact version they were shown.
The consequence matters: editing a document later cannot rewrite what somebody has already agreed to. Your edit creates a new version. Everybody who signed the old wording stays attached to the old wording, and the results screen shows each version separately with its own opened and agreed counts, alongside the frozen text of that version.
Email sends are stamped in the same way. Each send records which version of each document went out, to how many people and when, so you can evidence which wording was circulated on which date.
Respondents keep their access too. Somebody who has already responded through an emailed link sees a section listing the documents from that item, with the note that they can review or download a copy of exactly what they agreed to at any time.
Why go to this trouble? A signature attached to text that can be silently changed afterwards is not evidence of anything. Freezing the version is what lets a heritage venue or a theatre company show, months later, precisely what a volunteer was shown on the day they agreed to it.
Note that in practice you will rarely edit a document anyway, because documents freeze along with every other question as soon as the first response arrives.
What Locks Once Responses Arrive
The single most important operational rule in the feature. From the moment the first response is recorded, part of the item becomes permanent.
| Still editable | Frozen permanently |
|---|---|
| Title | Type (survey, questionnaire or form) |
| Introduction text | Anonymity mode |
| Thank-you message | Question text and helper text |
| Open and close status | Answer types and rating scales |
| Opening and closing dates | Answer options and their explanations |
| Allow more than one response per person | Whether a question is required |
| Whether a Seaty account is required | Document text, title and sign-off mode |
| The order the questions appear in | |
| How answers branch (the pathways) |
The interface shows you the lock rather than quietly discarding your edits. The drag handles disappear, and opening a question shows it read-only, with the answer type and sign-off displayed as plain values.
Why freeze the questions? Because results are only meaningful if everybody answered the same thing. If question 3 changed from "rate the venue" to "rate the interval" halfway through, the average underneath it would be a number with no meaning, and nobody looking at it later would know. The same logic applies to anonymity: you promised respondents something, and that promise cannot be retrospectively changed once people have relied on it.
What if you need a change anyway? Build a new item. There is no way to unlock a question, and no partial edit. This is why it is worth reading the whole thing through once, on a phone, before you open it.
Drafting with Seaty AI
Seaty AI can produce a complete first draft: a title, an introduction, a thank-you message, and a full set of questions with their types, options and helper text. Where the survey it is drafting calls for one, it also sets up a pathway, for example ending a membership form early for someone who answers that they are not renewing. It creates a real item that then opens in the builder for you to edit like any other.
Making the request
From the create dialog, choose Use Seaty AI, then either pick a recipe or choose Custom. You then answer four short questions:
- What is this about? Our organisation, or A specific event. Recipes that need an event ask you to choose one straight away.
- Who is it for? People who attended, Our members, Our volunteers, or The general public.
- How long should it be? Short (3 to 5 questions), Standard (6 to 9) or Comprehensive (10 to 14).
- Tone. Warm and friendly, Neutral, or Formal.
There is also an optional free text box of up to 1,000 characters, where you describe what you actually want to find out. This is the field that makes the difference. A brief that says "we want to know whether the new later start time worked for people travelling from out of town" produces a far more useful draft than the tiles alone.
Generation can take up to a minute, and you need to keep the window open.
The recipes
Fifteen ready-made briefs, filtered by the type you chose.
For surveys
| Recipe | What it drafts |
|---|---|
| Audience feedback | What the audience thought of a show, tailored to the event you choose |
| Member satisfaction | How members feel about how the organisation is run. Best used confidentially |
| Choosing the next show | A poll of members or audience on what to stage next season |
| Visitor experience | Feedback on a whole visit to a house, garden, museum or heritage site |
| Visitor profile for funders | Evidence for grant reporting: reach, first-time visitors and impact |
For questionnaires
| Recipe | What it drafts |
|---|---|
| Rehearsal availability | Cast and crew availability so a workable schedule can be planned |
| Skills and interests audit | What members can offer beyond performing |
| Committee and governance review | Structured feedback to the committee on how the organisation is being run |
| Volunteer experience | How volunteers feel about their role, training and support. Best used confidentially |
| Access and inclusion review | The practical barriers to visiting, especially for older or historic buildings |
For forms
| Recipe | What it drafts |
|---|---|
| Audition application form | For performers applying to audition, tailored to the event you choose |
| Membership application or renewal | New members and renewals, with consent and emergency details |
| Volunteer and crew sign-up | People offering to help with a production rather than perform in it |
| Venue hire or wedding enquiry | Private, wedding and corporate hire enquiries with the detail needed to quote |
| Volunteer application | Regular volunteers for ongoing roles such as guiding, gardening, retail or catering |
These recipes are separate from the twelve built-in templates offered under Start from a template, which are fixed ready-made items rather than AI briefs.
What Seaty AI is given
Information about your organisation and your events: the basics, descriptions, venue details, your member type names, the names of recent events or productions, your event dates, ticket categories and the shape of your schedule. It also receives a short profile line summarising whether you are a registered charity and whether memberships, Gift Aid and donations are switched on. Recipes only receive the parts they need.
No respondent, attendee or order data is ever sent. Schedules are summarised into a pattern rather than listed out, and anything marked admin-only is excluded, so internal dates cannot leak into a public form.
The upside of that context is that a membership form comes back offering your real membership types, and a heritage survey comes back knowing what kind of site you run.
Always review what it produces
Treat the output as a draft written by a capable assistant who has never met your organisation. It aims to produce a finished, respondent-facing item rather than a skeleton with gaps, which means where it did not know something it will have written a confident, plausible guess rather than leaving a blank.
Read the whole thing through and check:
- Every word is fit for a respondent to see. The system is instructed never to leave editor's notes or bracketed placeholders, and there is a safety net that strips stray notes from helper text, but it is not a guarantee.
- Names and specifics are right. Candidate show titles, character names, role names, referee and deadline policies: these are the things it will invent most confidently.
- The type and anonymity mode suit what you are doing. These freeze on the first response.
- The questions themselves. After the first response, none of them can be changed at all.
- Anything touching health, access or special-category information. The generated wording will say why the information is needed and that it will only be used for that purpose. That wording has to match what your organisation actually does, and only you know that.
Deleting a question you do not want is quicker than writing one from scratch, so the usual editing pass is mostly removals plus a handful of rewordings.
How Long Should a Survey Be?
There is no enforced limit, and the honest guidance is that shorter items get more, and better, responses.
A workable starting point:
- Post-event feedback: 4 to 6 questions, at most one free text. People are answering days after the event with no particular incentive.
- Member or volunteer surveys: 8 to 12 questions is reasonable. Your members have a stake in the answer and will give you more time.
- Forms: as long as it genuinely needs to be. Somebody completing a membership renewal or an audition application has already decided to do it, and abandoning halfway costs them more than finishing.
Things that measurably cost you responses:
- Required questions people cannot honestly answer
- Several free text boxes in a row
- Long documents in the middle of the flow rather than near the end
- Long answer options that wrap on a phone
- Anything that reads as though it were written for the committee rather than the respondent
Things that help:
- Opening with something easy: a rating or single choice, not an essay
- An introduction that says plainly how long it will take and what you will do with the answers
- Choosing the lightest anonymity mode that suits the subject. People answer honestly about the committee when they know the results screens will not show the committee who said what
- Sending it while the event is still fresh
- Following up the people who have not answered yet, rather than emailing everybody again
Why does the anonymity mode affect the response rate? Because it changes what people are willing to say. On a member satisfaction survey, an identified item and a confidential one will get materially different answers to the same question. Choose the mode for the subject, not out of habit. See Anonymity and data protection.
Common Questions
Questions and types
Can I change a question after people have started responding? No. Question text, helper text, type, options, required setting, document wording and the question order all freeze permanently once the first response is recorded. The title, introduction, thank-you message and the response settings stay editable.
Can I reorder questions later? Only until the first response arrives. After that the drag handles disappear.
What is the difference between Rating and Recommendation? Rating is a Poor to Excellent scale of a length you choose, for scoring one specific thing. Recommendation is the standard 0 to 10 likelihood question with fixed wording, scored as NPS. Use rating questions for individual aspects and at most one recommendation question for the overall experience.
How many options can a choice question have? At least two. There is no practical reason to go far beyond about six though, because long option lists are hard to read on a phone and produce thinly spread results.
Why can I not remove an option? You are down to two. A choice question with fewer than two options would have nothing to choose between.
Can I ask for a date, a phone number or a file? Not as dedicated types. The seven types above are what is available. Free text with clear helper text is the usual approach for things like a phone number.
Pathways
Can an answer skip questions or end the survey? Yes, on single choice questions. Each answer can continue to the next question, end the form, or skip ahead to a later one, set with the "After this answer" control under the answer. Only single choice answers can branch.
Can a pathway jump backwards or loop? No. A skip can only point at a later question, and a survey can never loop back on itself. This keeps it readable and stops respondents getting stuck.
What happens to a pathway if I reorder or delete questions? Seaty checks it for you. Moving a question above an answer that skips to it, or deleting a question something skips to, is confirmed with you first, and the affected pathway is turned back to "continue" rather than left pointing at nothing.
Do pathways freeze on the first response too? Yes. Like the questions themselves, how answers branch is fixed once the first response arrives.
Documents and sign-off
What is the difference between a confirmation and a document? A confirmation records that a tick box was ticked. A document records exactly what wording was on screen, frozen as a version, tied to that person's sign-off. If it matters what they agreed to, use a document.
Can somebody tick "I agree" without reading the document? They can skim it, but they cannot tick it without opening it and scrolling to the end. The tick box stays disabled until then.
What happens if I edit a document after people have signed it? Your edit becomes a new version. Everybody who already signed stays attached to the version they saw, and results show each version separately. In practice this rarely arises, because documents freeze with everything else on the first response.
Do respondents get a copy? They can download one at the time, and if they came through an emailed link they can return later and review or download exactly what they agreed to.
Do I need to mark a document as required? No. It is required automatically unless its mode is Information only.
Seaty AI
Does Seaty AI see our attendees or orders? No. It receives information about your organisation and your events only. No respondent, attendee or order data is sent.
Do I have to use what it generates? No. It creates an ordinary item that you can edit, add to, or delete questions from freely, right up until the first response.
Why did it invent a name or a detail we never gave it? It is instructed to write a finished, plausible version rather than leave placeholders. Where your organisation or event information did not cover something, it fills the gap confidently. Always read the draft through before opening it.
What is the difference between the AI recipes and the templates? The recipes are briefs that Seaty AI writes from, tailored to your organisation. The templates are twelve fixed, ready-made items that load exactly the same way every time.
Getting it live
Why can I not open my survey? An item cannot be opened while it has no questions. Add at least one.
Can I test it before sending it? Open it and use the share link yourself. From the send screen you can also send a test email to yourself to see exactly what recipients will receive. Bear in mind that a real test response counts as a response and will freeze the questions, so test on a copy if you are still drafting.
Where do responses go? See Reading results.