How to Improve Website Enquiry Forms Without Attracting Low-Quality Leads

Remove fields that do not change what you do next, explain the ones you keep, and add one or two banded qualifying questions instead of a long form. Then deal with spam, fake details and low-intent enquiries as three separate problems, because each has a different fix. Finally, judge any change by how many qualified leads it produces, not by how many forms are submitted.

Shortening a form and lengthening it are both blunt tools. A short form can flood sales with enquiries they cannot use; a long one can lose good buyers who were ready to talk. The useful question for each field is: what does this cost the visitor, and what does it tell us that we would act on?

This table is a starting point for a typical service business: an agency, consultancy, software studio or B2B supplier. The friction and value ratings are judgement, not measured data, so adjust them to how your sales process works. "Ask later" means collect it on the first call or in a follow-up email, once the visitor has a reason to share it.

"Low-quality leads" usually mixes three things. Sort a month of enquiries into these buckets before changing anything, because the fixes pull in different directions.

1. Bots. Gibberish names, link-stuffed messages, bursts of submissions at odd hours. Fix these out of sight of real visitors. A honeypot is a hidden field that people never see but simple bots fill in; reject any submission where it has a value. Validate everything again on the server, since bots can skip your page entirely. If spam persists, add a background check. Cloudflare Turnstile has three widget modes: Managed (the visitor is asked to tick a box only if a further check is needed), Non-interactive (a spinner, never a prompt) and Invisible (no widget at all; Cloudflare says you must then reference its Turnstile privacy addendum in your privacy policy). Google reCAPTCHA v3 returns a score for each request without interrupting the visitor, where 1.0 is very likely a good interaction and 0.0 very likely a bot; Google says that "by default, you can use a threshold of 0.5" and that you should verify the action name on your server.

2. Fake details from real people. Phone numbers like 0000000000, throwaway email addresses, "asdf" in the company field. These are often a symptom of the form, not the visitor: people invent details when a field is required and they do not see why. Baymard's checkout research found 39% of benchmarked sites required a phone number "without any sort of explanation", and 14% of 1,026 surveyed shoppers said they would never give an online store their phone number. Make such fields optional, explain why you ask, and the fake answers tend to become blanks, which are more honest and easier to handle.

3. Genuine but low-intent enquiries. Students, job seekers, people wanting a free opinion, or buyers whose budget or scope will never fit. CAPTCHAs do nothing here because these are real humans. Fix them with copy and routing: say above the form who you work with and what a typical engagement involves; offer a self-serve route (a guide, a pricing explainer, a careers page) for people who are not buying; use the service and budget answers to route enquiries automatically, so a poor fit gets a polite, useful reply instead of a sales call.

Small details decide whether a good buyer finishes the form. NN/g recommends clear, visible labels placed outside empty fields rather than placeholder text, which disappears as soon as someone types. For errors, NN/g advises waiting until the user has finished with a field and moved on before showing an error, placing the message next to the field. The exception is complex input such as a new password.

The GOV.UK Design System's question pages pattern says to "only ask users for information you really need" and to make it clear why you are asking each question. That second point is the cheapest improvement most enquiry forms can make: a short hint under the budget or phone field often does more than removing it. If your form needs several steps, see how to design multi-step forms.

Several WCAG 2.2 criteria apply directly to enquiry forms, and meeting them tends to help everyone, not only people using assistive technology. For a broader view, see accessibility in UI/UX design.

If the only number you track is submissions, any change that adds junk looks like a win. Google Analytics 4 has recommended events for the whole lead lifecycle, listed in the GA4 events reference: To connect those quality signals back to the channels that produced them, see measuring conversions across multiple channels.

This is a practical summary, not legal advice. Rules differ by country and change, so check current rules with the regulator or a lawyer for each market you serve. The common thread: say who you are, why you want the details and where your privacy policy is, and ask for marketing consent separately with an unticked box.

As a factual reference, not a result: the ZSpace Labs contact form asks for name, email and project details as required fields; company, website and phone are optional, and the website field accepts an address without https://. It asks what the visitor needs and a budget range from a short list that includes "Let's discuss". A one-line privacy notice with a link to the privacy policy sits under the submit button. A hidden honeypot field is checked on the server, and the generate_lead event is sent to GA4 only after the server confirms the enquiry was delivered, never on the button click.

Hypothetical example: a 12-person IT consultancy gets around 60 form submissions a month. A partner tags a month of them and finds roughly a third are spam, a third are job seekers or students, and a third are genuine buyers, several with fake phone numbers.

They make four changes. A honeypot and Turnstile in Managed mode for the spam. The phone field becomes optional, with "only if you'd like a call back". The careers page and a short guide are linked above the form for non-buyers. A banded "What are you looking for?" question routes partnership and job enquiries away from sales. Submissions fall, which is expected. Whether the change worked is judged by qualified enquiries per month and the reasons in disqualify_lead, compared over the next two or three months with an eye on seasonality.

If almost all your submissions are genuine but few become customers, the form may not be the problem: the traffic, the offer or the follow-up might be. Work through why a site gets traffic but no leads first. If you get very few enquiries of any kind, adding qualifying questions will make that worse; focus on reach and the offer, and see the wider lead generation guide. For high-volume, inbound-led businesses, scoring and routing after submission may matter more than the form; automated lead qualification covers that side. If the form is fine but enquiries still do not close, work through how to improve lead quality.

Measurement and bot controls: Google: GA4 recommended events reference; Google: reCAPTCHA v3 (vendor documentation); Cloudflare: Turnstile widget modes (vendor documentation).

Form usability: Baymard: checkout flow average form fields (ecommerce checkout research, published 26 June 2024); Baymard: explain the phone number field (checkout research); NN/g: placeholders in form fields are harmful; NN/g: guidelines for reporting errors in forms; GOV.UK Design System: question pages.

Accessibility: W3C WCAG 2.2 Understanding documents for 3.3.1, 1.3.5, 3.3.7 and 3.3.8.

Regional: GOV.UK: DUAA factsheet, PEC Regulations; ACMA: consent expectations for businesses using direct marketing; PDPC: Do Not Call Registry for organisations; Baker McKenzie: UAE key data laws (law-firm commentary).

Accessed 14 October 2026. Vendor features, analytics events and data protection rules change; check the current versions. Nothing here is legal advice.

Export the last two or three months of enquiries and tag each one as bot, fake details, low intent or genuine. Then walk the field table against your own form, fix whichever bucket is largest first, and set up generate_lead and qualify_lead so the next change can be judged on quality. If you want an outside view, a CRO audit can cover the form, the page around it and the tracking together.