Web Design Client Questionnaire: Collect Content and Goals With AI Help

A good web design client questionnaire asks about goals, audience, content, competitors and practicalities in about twenty questions, then marks each answer complete, partial or missing. AI can turn the completed answers into a site brief, a sitemap and a content plan, and draft one chase email for what is still outstanding.

The questionnaire is the cheap part. The expensive failure is the client who answers "we will send the text later" and then does not. Everything below is built to surface that on day two, not week six, with a template you can copy and a set of prompts that process the answers without inventing any.

Follow me on Instagram@sagnikteaches

The twenty questions, grouped by what they feed

Group the questions so that each block feeds a specific part of the brief. Clients also tolerate a form better when the sections have plain names. Mark the starred ones as essential: if they are blank, work cannot start.

Connect on LinkedInSagnik Bhattacharya

Goals and success (5)

1. In one sentence, what does the business do and for whom?*
2. What is the main thing a visitor should do on the site: call, book, request a quote, buy, or something else?*
3. What is the second most important action?
4. How will you know the site has worked in six months? Give a number if you can.
5. What is wrong with the current site, or why do you need one?

Audience and positioning (4)

6. Who are your best customers? Describe two types.*
7. What questions do they ask before they buy?
8. What do you do that competitors do not, or do better?
9. Name three sites you like and say what you like about each.

Content you already have (5)

10. List the services or products you want on the site, with one line on each.*
11. Do you have a logo, brand colours and fonts? Where are the files?*
12. Do you have photos of your work, team and premises, and are they yours to use?
13. Do you have testimonials, reviews, case studies or accreditations you can show?
14. Do you have existing text (brochures, old site, proposals) we may reuse?

Practicalities (6)

15. Who is the single person who approves content and design?*
16. What is the launch date, and what is it tied to?
17. Which domain name and email address will be used, and who controls the registrar login?*
18. Which tools must the site connect to: booking, payments, CRM, email list, maps?
19. Who updates the site after launch: you, your team, or us?
20. Anything legal or regulatory the site must show (licences, disclaimers, opening hours)?

The questions are ordered so the first block can be answered in five minutes. Ask them in a form that saves progress, not a long email, so partial answers are not lost.

Subscribe on YouTube@sagnikteaches

The worked example: a four-person roofing company

Illustrative case: a roofing company with four staff has an old five-page site that takes few enquiries. The owner fills in the form on a Tuesday. Here is the raw output, trimmed:

Q1: Roof repairs and replacements for homeowners and small landlords.
Q2: Phone call for a quote.
Q3: Quote request form.
Q4: Maybe 10 enquiries a month instead of 3?
Q5: Looks old, doesn't work well on phones.
Q6: Homeowners with leaks, landlords with several flats.
Q7: How much, how long, are you insured, do you do emergencies?
Q8: We answer emergencies same day.
Q10: Repairs, replacements, flat roofs, gutters, chimney repointing.
Q11: Logo yes, colours unsure.
Q12: About 40 photos on my phone, some are of other firms' roofs.
Q13: 3 reviews, no accreditations to show.
Q14: Old site text, 800 words.
Q15: Me.
Q16: Before winter.
Q17: Domain with the old web guy, no idea of login.
Q18: Nothing special.
Q19: Me.
Q20: Insurance details.

That is typical: strong on the services, thin on proof, and with three red flags buried in it (photos that may not be his, a domain whose login the owner cannot reach, and a launch date with no content plan). A human reading it carefully will find these. A tired project manager reading twelve of these a week might not. This is where the AI step earns its place.

Step one: ask AI to grade the answers, not summarise them

The first prompt should audit, not write. Paste the answers and the list of essential questions. Illustrative prompt:

You are helping a web design agency review a client questionnaire.
For each question, mark the answer COMPLETE, PARTIAL or MISSING, and give a one-line reason.
Then list: (1) blockers that stop work starting, (2) risks (rights to images, domain access, approvals), (3) follow-up questions to ask on a call.
Use only the answers below. Do not guess or fill gaps. If an answer says "later" or "unsure", mark it PARTIAL.
Essential questions: 1, 2, 6, 10, 11, 15, 17.
Answers: [paste]

Illustrative output for the roofing company:

Blockers: Q17 domain login unknown (partial); Q11 brand colours unsure (partial).
Risks: Q12 some photos may belong to other firms; Q13 only three reviews, no accreditations; Q4 target stated with a question mark.
Follow-ups: Who is the web guy and can the owner contact him? Which photos are of your own jobs? Can you ask five recent customers for a review?

Check it against the raw answers before you act on it. Grading against a fixed list of essentials is the simpler job for a model; the risk list is where it is more likely to overreach, so read each line and delete any you cannot trace to an answer.

Step two: turn the answers into a one-page brief

Once the blockers are cleared, ask for the brief. Keep it to a page so the client will actually read it. Illustrative prompt:

Write a one-page website brief from the questionnaire answers below. Sections: business, audience, primary and secondary goals, differentiators, content available, content needed, constraints, approver, launch date.
Use the client's own wording where possible. Where an answer is missing, write "Not provided" and do not guess.
Answers: [paste]

The line "write 'Not provided'" is the control. Without it, a model fills the gap with plausible detail (for example "fully insured and fully licensed" for a business that only said "insurance details"). A false claim of that kind, copied onto a roofing site, is a real problem. Send the brief to the client for sign-off, and say that the signed brief is the reference if scope is later disputed. The proposal that comes before it is covered in the web design proposal template tutorial.

Step three: a sitemap and a content plan from the same answers

The services list (Q10) and the audience questions (Q6 and Q7) drive the sitemap. Illustrative prompt:

From the brief, propose a sitemap for a small-business site of no more than 8 pages. For each page give: purpose, the question it answers, the main call to action, and the content needed.
Then produce a content table with columns: page, content item, owner (client or agency), status (have, need, rewrite), due date.
Base everything on the brief. Do not add services the client did not list.

Illustrative output (shortened):

Notice how much of the "Need" column belongs to the client. That is the real project risk. A sitemap shows it on one page. If you work from wireframes next, AI wireframes from a brief shows how the brief feeds the layout, and how to write website copy with AI covers the writing once content arrives.

Step four: the missing-content chase email

This is the step that pays back the questionnaire. Draft one email, itemised, dated, and with the consequence stated. Illustrative prompt:

Write a friendly but firm email to the client listing the content still needed, from the content table (status "need" or "partial", owner client).
Give each item a due date, one line on exactly what to send and in which format, and say what happens to the launch date if items arrive late.
Plain English, under 200 words, signed [your name].

Illustrative output, after the agency's edit:

Hi [first name],

Thanks for the questionnaire. To keep the launch before winter we need five things by Friday 14th:
1. Login for the domain (or the name of the person who has it)
2. Twenty photos of jobs you carried out, as originals, not screenshots
3. Five recent customer reviews, with permission to quote them
4. Your insurance details exactly as you want them shown
5. Two or three lines on your emergency call-out process
If these slip by a week, the launch moves by a week. Reply with anything you're stuck on.

[your name]

Three things make it work. Each item says exactly what to send. Each has a date. The last line states the effect of delay on the timeline, which turns a hope into an agreement. If the client still stalls, the change order template handles scope that grows after the brief is signed.

How to use the graded list on the kickoff call

The audit output is most useful as a call agenda. Instead of walking through the form again, which clients find tedious, open with the blockers and spend the call on the partial answers. A twenty-minute call can then settle what an email thread would take a week to resolve: who holds the domain login, which photos are the client's own, who approves the work, and when the first batch of content will arrive.

Open the call by saying what you have already understood, in two sentences taken from the brief, and ask the client to correct them. This does two jobs. It shows you read the answers, which builds trust early in a project, and it catches misunderstandings while they are still cheap. Finish by agreeing three dated actions on each side and promising to send them in writing within the hour.

Adapting the twenty questions to the sector

The questionnaire is a base, not a script. Change four or five questions for the trade, and keep the rest. Three illustrative adaptations:

• A plumbing firm. Replace question 19 about tools with emergency handling: do you take after-hours calls, and what should the site say about call-out charges? Add a question on service areas stated by the areas covered, because a plumber's page lives or dies on whether it names the places served. Add one on the trade licences or gas-work certifications the site must show, with the client confirming the exact wording in writing.
• A pest-control firm. Ask which pests generate most enquiries and which are seasonal, because the sitemap should give each main pest its own page. Ask about treatment safety information for households with pets or children, which the client must supply rather than the agency write. Ask how the client handles recurring contracts, since that changes the main call to action.
• A cleaning company. Ask for the split between home and commercial work and whether the two should have separate pages. Ask how quotes are produced: a fixed price list, a form, or a visit. The answer decides whether the site can carry a price calculator or only a request form.

Whatever the sector, keep the five essentials: the one-sentence description, the main action, the audience, the service list and the single approver. If those five are strong, the rest can be chased.

Turning partial answers into dates on a calendar

Most projects slip because content arrives late, not because the design takes long. Use the content plan to work back from the launch date. Illustrative sums for the roofing project, with launch set for 1 December and a typical build of three weeks once content is in hand:

1. Build can start by 10 November at the latest, so the first content batch is due by 7 November.
2. Photos and reviews are the slowest to collect, so ask for them first, with a due date a week before the first batch.
3. Allow one week for client review of the first draft, and a second week if the approver is travelling.
4. If the first batch arrives a week late, either the launch moves or the first release ships without the reviews page.

Writing the arithmetic into the chase email is blunt but effective. Clients tend to accept a clear rule ("every week of delay moves the launch by a week") more readily than a vague complaint. Agree it before the work starts, ideally in the proposal; the proposal tutorial linked earlier gives wording for that clause.

Keeping the client's answers safe when AI reads them

Questionnaires contain business detail and sometimes personal data: the owner's name and phone number, staff names, customer reviews with names, even pricing. Before you paste answers into any AI tool, decide three things. First, which tool: a paid business plan with data-controls you have read, not a free consumer account. Second, what to remove: strip out personal phone numbers, customer names and anything the client marked confidential. Third, what to tell the client: add one line to the form, such as "We may use AI tools to organise your answers; we do not publish them or use them for anything but your project." Then do what the line says.

If a client objects, offer the manual route: the same audit done by a person, with a slightly longer turnaround. Having that option ready makes the question easy to answer rather than a sticking point.

A scoring rubric for answer quality

If you run many projects, score each questionnaire out of ten so you can spot patterns. Illustrative rubric: two points for essentials complete, two for a stated primary action with a number, two for evidence supplied (reviews, photos, accreditations), two for named approver and domain access, and two for dated content commitments. Under five, treat the project as at risk and add a content-gathering step to the proposal. Eight or over, start the build. Over a few months you will see which question types clients fail on, and you can reword those questions or move them to the call.

What to do when two people answer the form

Question 15 asks for a single approver because the commonest late-stage fight is two owners disagreeing about a headline. Sometimes the form still arrives with two sets of answers, or with a partner's edits layered over the first. Do not merge them silently. Ask the model to list every place the two versions disagree, then take that list to the client as a decision sheet: "You said the main action is a phone call; your partner said a quote form. Which comes first?" A five-item decision sheet, answered in one email, is far cheaper than a design round built on the wrong choice.

Record each decision with the date and the name of who made it, and attach it to the signed brief. When a disagreement resurfaces during review, you can point to the decision rather than reopen the argument.

Where AI goes wrong on questionnaire answers

• It smooths over gaps. A "not sure" becomes a confident sentence. Counter with the "Not provided" instruction and a human read.
• It over-rates the audience. It may turn "homeowners with leaks" into three polished personas. Keep personas to what the client said unless you have data.
• It repeats the client's weak claims. "Best roofers" is not a differentiator. Ask the model to flag any claim with no evidence supplied.
• It never questions rights. It will not notice that a photo may belong to another firm. You have to ask.
• It drifts from the client's words. When the brief is read back, the client may say "that isn't how I'd put it". Quote their own wording and fix the text before the sign-off.

Making the form easy to complete

The wording of the form decides how many answers you get. A few practical changes:

• Ask for examples, not abstractions. "What do customers ask before they book?" gets better answers than "Describe your value proposition".
• Offer a file upload on the content questions so photos and documents arrive with the answers rather than in a later email.
• Show the starred essentials first, and tell the client roughly how long the form takes: twenty minutes if the files are to hand.
• Save progress, so a half-finished form is not lost.
• Reply the same day to say what happens next. A form that goes into silence teaches clients to be slow too.

A second example: a cleaning company that sends everything

Illustrative opposite case: a cleaning company answers all twenty questions in full, attaches a style guide and twelve testimonials. The risk here is different. The brief runs to three pages because the model kept every detail, and the sitemap ends up with twelve pages for a business with three services. The fix is in the prompt: set a page cap, ask for the three most important goals only, and ask the model to list what it left out so you can decide, not the model.

A third example: the client who answers in a phone call

Some clients will not fill in a form. Run the same twenty questions on a call, record it only with the client's agreement, and paste the transcript into the audit prompt. Ask the model to map the transcript to the twenty questions and mark any it could not find. The output is the same graded list. The extra check is that the model may attribute a remark to the wrong question, so read the mapping, not just the summary.

Checking the brief, sitemap and content plan before they go out

1. Trace every claim. Each sentence in the brief should map to a questionnaire answer. Highlight anything that does not.
2. Count the pages. Does the sitemap match the services the client listed and the scope you quoted? If not, expect a scope argument.
3. Read the "need" column. Is every client-owned item dated? Items without an owner or a date do not get done.
4. Check the legal and rights points. Images, reviews, licences and accreditations are the client's to confirm in writing.
5. Send for sign-off. The brief, sitemap and content plan go to the single approver named in question 15, with a reply-by date.

Keep the signed answers, brief and chase emails in the project folder. If you are building several client sites from one process, website creation automation shows how the same inputs can feed a template, and free AI tools for a freelancer or small agency covers which tools can do the processing without a new subscription. If you would rather set up the whole intake step with someone, that is what my AI implementation consultation covers, at the contact page.

Questions agencies ask about client questionnaires

How long should a web design client questionnaire be?

Twenty questions is a workable ceiling for a small-business site. Beyond that, completion drops and answers get thinner. Split it into goals, audience, content and practicalities, mark which questions are essential, and ask follow-ups on a call rather than adding more boxes.

Should I send the questionnaire before or after the proposal?

Send a short version of three or four questions before the proposal, so the scope reflects what the client needs. Send the full twenty after they sign, so the answers drive the brief, sitemap and content plan.

Can AI fill in the questionnaire for the client?

No. AI should read the answers the client gives and organise them, not invent them. If you ask it to guess missing answers it will produce fluent but fictional detail about the client's business, which then ends up in the site copy.

What if the client sends half-answers?

Mark each answer as complete, partial or missing, and send one chase email listing exactly what is outstanding. Set a date, and say what happens to the timeline if the content arrives late.

Further reads

• AI Accessibility Audit: Can It Replace a Manual WCAG Check? — Whether AI can replace a manual accessibility check.
• Figma Make for Web Designers: Client Prototypes and Where It Breaks — Client prototypes and where they break.
• AI Tools for Website Design: Accessibility Checks Before Handover — What to check before handover.
• Website Maintenance Plan: Packages a Web Agency Can Add With AI — Packages to offer after launch.
• Website Development by AI: What Agencies Can Ship and What Needs Review — What AI can ship on the build side, and what needs review.
• AI Scope of Work Generator for Agencies: SOWs With Change Control — A worked example from client call to signed SOW: counted deliverables, exclusions, change-control wording and a filled change request form.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: no vendor prices or product features are cited; the questionnaire, prompts and sample outputs are illustrative templates.