Website feedback: how to test your website with real people and turn more visitors into customers

The page every customer sees

Have you ever landed on a website and had no idea what the company actually does? You read the headline, scroll a little, watch something slide in from the side, and leave none the wiser. There is a good chance that some of your visitors feel the same way about yours.

It is rarely for lack of effort. Companies invest a lot in their websites, and for good reason. But a website is usually built from the company out. You know your product inside out, so you write about it in the words you use every day, and you try to say everything you know. The agency that builds it wants to impress you, so it adds the animations that make your team say “wow” in the review meeting.

Your visitors are sitting on the other side of the screen, looking at your page for the first time. They do not have your context, and they are not interested in your vocabulary. They are trying to answer two simple questions: what do you do, and how can it help me? And they will not try for long. In one well-known study, people judged how a web page looked in about 50 milliseconds, and that snap judgment held when they looked for longer. You do not get many chances with a potential customer. If your page does not make sense to them on the first visit, you may not get a second.

Your website is the one page almost every customer sees, and it is the one page you build without asking them. Website feedback is how you stop guessing: you put your pages in front of real people and hear, in their own words, what they understand, what they look for and what makes them leave.

When website feedback matters most

You can test your website at any time, and you should do it regularly. But there are three moments when it matters most.

The first is when you launch. As a new company, you are still learning how to explain what you do in a way your prospects connect with, and you only find out by watching some of them read your page.

The second is when you redesign or rebrand. A new website often throws away words that worked and replaces them with words that sound better in the boardroom, so test the new version before it goes live, while changing it is still cheap.

The third is when you reposition, and right now a lot of companies are rewriting their story around AI. That is often the right move for the product. But the new story has to make sense to customers who were never in your strategy meetings, and that is exactly where it tends to go wrong.

What HubSpot’s new headline shows

A fresh example: on 12 November 2026, Brian Halligan returned as CEO of HubSpot and called it a refounding of the company. The homepage already has a new headline: “HubSpot: The Customer Relationship Harness.”

From the inside, “harness” probably makes perfect sense. It is how the leadership sees HubSpot’s place in a world full of AI agents. But it is hard to imagine many of HubSpot’s customers reading that line and recognizing their own problem in it. Harnesses are everywhere in AI right now. Lovable is one, and Legora is one. But look at how they talk about themselves. Lovable says “Build something Lovable.” Legora, the legal AI company, says “Legal work, without limits.” Neither leads with the word harness. Both talk about what the customer gets done.

HubSpot’s best line is already on the page, right under the headline: “AI knows the world. HubSpot knows your world.” That is a sentence a customer could say. And that is the lesson for the rest of us. The vision is how you see your company. The headline is how your customers would say it. You find the second one by asking them.

Four mistakes that cost you customers

If some of the most successful companies in the world struggle to say clearly what they do, what chance do the rest of us have? Fortunately, it does not have to be hard. The same four mistakes come up again and again, and each one is easy to spot the moment you watch a real person read your page.

1. The words are yours, not theirs. Fancy words that mean nothing to a visitor. When we looked at Crisp’s homepage in November 2026, the headline read “Augment your customer experience.” Crisp makes a helpful chatbot and support platform for websites, but those words say nothing about chat or support, so a visitor has to read on to find out what Crisp makes.
2. The effects pull attention away from the message. Text that slides in on one side while an animation plays on the other looks impressive in a design review. For a visitor, it splits their attention, and many simply scroll past without reading either.
3. The information they came for is missing. First-time visitors want to know what the product does, how it works, who it is for and what it costs. Notion is a good example. Many people know it as the place for their notes, docs and wikis, yet its homepage leads with “Where teams and agents Think together,” and none of its section headings mention docs, notes or wikis.
4. There is no clear next step. We tend to treat a website as a static page that has to be perfect. In reality, it is one step in a funnel. A good page guides the visitor onward, to a sign-up, a demo or a price, and makes that step feel obvious.

None of these show up in your analytics. Say your dashboard shows that 70% of visitors leave your pricing page. It cannot tell you that they leave because they cannot work out which plan is for them. For that, you have to hear it from them.

Which pages to test

When people say website feedback, they usually mean the landing page. That is the right place to start, but it is not the only page that loses you customers. A visitor who understands your homepage still has to choose a plan and get through the sign-up, and each step is a new chance to lose them. So test the pages that decide whether someone becomes a customer.

Landing page testing with people is not the same as an A/B test. An A/B test tells you which of two pages converts better. Testing with people tells you why, and what to change next.

Beyond these, look at your analytics for the pages where a lot of visitors arrive: a product page, a comparison page, a popular article. If many people start their visit there, it is a landing page too, whether you designed it as one or not.

And if you are still deciding what to say, test the words before you test the page. A messaging test, or message testing, shows people two or three headlines and asks which one makes them want to learn more, and why. It takes an afternoon to set up, and it can save you from building a whole page around the wrong message.

How to get website feedback

There are a lot of website feedback tools out there, and they do quite different jobs. Analytics and heatmaps show you what people do on a page. A feedback widget in the corner of your site lets visitors leave a comment or a rating, and some tools let your team or your clients pin comments to a page to report bugs. All of these are useful. But none of them tell you how a new visitor experiences your page, from the first second to the moment they decide to sign up or leave.

For that, the best method is also the oldest: watch someone use your page while they think aloud. They scroll, read and click as they normally would, and say out loud what they notice, what they understand and what confuses them. This is website usability testing in its simplest form. The first time you try it, it is a revelation: you watch people struggle with your page in ways you never expected, and you hear the exact moment they lose interest.

There are three ways to do it.

• In person. Sit someone in front of a laptop, open your page and ask them to talk you through it. It works well, and you can ask questions as you go. But it takes a lot of time, it is hard to record, and a week later it is hard to remember who said what, let alone show it to your team.
• Ask for a screen recording. Send people your link and ask them to record their screen with a tool like Loom while they talk. You get a recording you can keep and share. But everyone gets the same vague instruction, you cannot ask a follow-up, and you end up with a folder of videos to watch one by one.
• Run a short study with screen sharing. You send one link. People open it, share their screen while they go through your page and answer a few questions, by voice or in writing. An AI can ask a follow-up about what each person just said. You get the recordings, the transcripts and the answers in one place, and it scales from five people to fifty without any extra work.

The third way gives you the depth of the first with the reach of the second, which is why it is the one we recommend, and why we built Feedbaq around it.

Who to ask, and how many

Ask people who match your target customer and who have not seen the page before. Their first reading is what you want to capture, and it only happens once. Your team, your investors and your friends know too much to give you that.

Where do you find them? Your waitlist and newsletter are good places to start, and so are communities where your customers spend time and your own network on LinkedIn. Recent sign-ups are valuable too: they remember what almost stopped them. Customers who have used your product for a while can tell you whether your page describes what they actually value, which is often not what you think.

And you need fewer people than you might expect. Jakob Nielsen famously showed that five users find most of the problems in a design. After that, you mostly hear the same things again. So five people per round is plenty. Test more only when you want to compare groups, such as current users and people who have never heard of you, and run another round once you have fixed what the first one found.

Website feedback questions to ask

So which questions should you ask? The best website tests are short, and they put the page first. Do not explain your product before people see it, because their first reading is the data. Our landing page feedback template uses the questions below, plus an optional email field so you can follow up. Each one has a job.

1. 1

   “How familiar are you with [product]?”

   Offer four answers: never heard of it, heard of it, tried it, and use it today. It lets you read the rest by group, because someone who uses your product every day reads your page very differently from someone who has never heard of you.

   Multiple choice

2. 2

   “Please open this page and talk us through your first impression.”

   The heart of the test. People share their screen, scroll and click as they normally would, and say out loud what they read, what they notice and how it strikes them, both the words and the look. An AI follow-up can ask about something they said, such as the line that confused them.

   Screen share

3. 3

   “In your own words, what does [product] do, and who is it for?”

   The single most revealing question on the list. If people cannot say what you do after looking at your page, nothing else on it matters yet. And when they can, listen to the words they use. They are often better than yours.

   Best spoken

4. 4

   “Overall, what is your first impression of the page?”

   A rating from excellent to very poor gives you a number to compare between rounds. On its own it tells you little, so let an AI ask why they chose it.

   Multiple choice

5. 5

   “Please select any statement that you agree with.”

   Offer the four mistakes as statements, plus one about trust: I disliked how the page looks, the messaging was hard to understand, the page missed information I needed, it was not clear what to do next, and I was not sure I could trust you. Counting the answers shows you which problem is biggest.

   Multiple choice

6. 6

   “Please open the page again and share your second impression.”

   Optional. After the questions, people look at the page with fresh eyes and often notice what they missed the first time, or confirm what bothered them.

   Screen share

For a pricing page, the same approach works with different questions: ask people to look around and say what they notice, which plan they would pick and why, and how the price feels for what they get. Do not ask what they would pay. Ask how they read the price you show. For a sign-up flow, give people a goal instead of steps, such as “sign up and get started,” and ask them to say what they expect before each click. Then ask how far they got and what was hardest.

Most of these questions are open, and most of the answers are spoken. That is on purpose. People say about three times as much when they talk as when they type, according to a 2024 study with 1,001 people, and a first impression is something people say far more easily than they write.

What to do with the answers

Start with question three. Put what people said your product does next to what you want them to say. If most of them got it right, your message works, and you can move on to the smaller problems. If they got it wrong, or could not say at all, that is the problem to fix first, and no new animation or extra section will make up for it.

Then collect their words. Every time someone describes your product, your problem or their own situation in a way that clicks, write it down. These are the raw material for your next headline, because they are the words your customers already use. This is exactly what HubSpot could do with one question to its power users: how would you explain what HubSpot does for you? The best of what they say belongs on the homepage.

Next, look for patterns. Count how many people picked each problem in question five, and watch the moments in the recordings where that problem shows. Two people who could not find the price is a pattern worth fixing. One person who disliked your font color probably is not.

Finally, share a few recordings with your team. A one-minute clip of a real prospect saying “I still don’t know what this does” ends more debates about the homepage than any slide deck. Then fix the biggest problem, and test again. A second round with five new people tells you whether the fix worked, and it often reveals the next problem down.

How to run it in Feedbaq

Feedbaq is built for exactly this kind of test. You start from the landing page, pricing page or onboarding flow template, with the questions ready to go, and add the link to your page. If the page does not exist yet, the messaging test compares your headlines first. People open your study, share their screen on desktop and talk you through your page while they scroll and click, then answer the rest of the questions by voice or in writing.

You can add an AI follow-up to the questions where you want to dig deeper, and it asks about what that person just said, the way you would if you were sitting next to them. Every spoken answer is transcribed, and Feedbaq labels each spoken and written answer with themes such as Usability Issue, Pricing, Objection and Criticism, so you see at a glance how many people struggled with the page or the price. You can filter the answers by any choice, so you read the people who already use your product apart from the ones who have never heard of you.

You can share the study as a link with your waitlist, your customers or your network. You can also add Feedbaq to your own website with one line of code and open the study from a button, so you hear from real visitors while they are on your page. And if you work with an AI agent, you can hand most of the work to it. Connect Claude, ChatGPT, Codex, Cursor or Grok to Feedbaq, and it can set up the study for you, then read every answer and pull out the themes, quotes and words to use in your next headline.

Test your website this week

You do not need a research team or a big budget to start. Here is how to run your first website test in a week.

1. Pick the page. Start with the page that matters most right now: usually your landing page, or the new version of it if you are about to launch one.
2. Start from the template. The landing page feedback template has the questions in order, with screen sharing and AI follow-ups where they help. Add the link to your page and try it yourself first.
3. Find five people. Pick people from your target audience who have not seen the page, and send them the link. Each test takes five to ten minutes, and they do it when it suits them.
4. Watch, read and collect their words. Compare what they say you do with what you want them to say, count the problems, and share the clearest recordings with your team.
5. Fix the biggest problem, then test again. Duplicate the study and send it to five new people, so you know the fix worked.

Building and previewing the study in Feedbaq is free, so you can try every question yourself before a single visitor sees it. Collecting answers from real people costs $49 a month with every feature included.

Questions people ask