Use Claude to See What a Client Finds When They Google You

You cannot see your own website, and that matters more than you think

When someone gives your name to a prospective client, that person does not email you straight away. They spend around five minutes online first. They search your name, look at your LinkedIn, scan your website for evidence you have actually done the work, and then decide whether it is worth making contact.

You have no idea what they come away with. And you cannot find out by doing the journey yourself, for two reasons. You are too close to the business, so you already know what you do and what you expect to find. And your own computer is logged into things, full of cookies and past searches that quietly change what you see.

What you need is a browser that has never met you. This walkthrough uses the in-app browser inside the Claude desktop app to act as a prospective client and work through that entire journey, from the first search to the point of getting in touch.

What you need before you start

You will need the Claude desktop app, which is available for Windows and Mac, and a paid subscription. The in-app browser sits inside the Code tab. Do not let that put you off — you are not writing any code, that is simply where the browsing functionality lives.

It may ask you to select a folder on your computer when you open that tab. Create an empty folder anywhere and choose it. Once you are on that screen you can start.

The prompt

The prompt does the heavy lifting here. It tells Claude who to be, what decision it is making, what order to work through, and exactly how to report back. That structure is what turns a vague website review into something you can act on.

Here is the prompt used in the video:

You’re a prospective client. A colleague you trust has mentioned [YOUR COMPANY] as a possible fit for a piece of work. You’ve never heard of them before now. You’ve got about five minutes and three other options on your list. You’re not hostile — you’d like this to work out — but you’ll move on if you can’t get comfortable.

Work only from what you find. Don’t use anything you already know about this company, and don’t fill gaps with assumptions about what a company like this probably does.

Work through these five steps in order. Don’t skip ahead to the website:

1. Search — search for the company name. Before you click anything, note what the results page alone tells you, and what it doesn’t.
2. The people — find the company on LinkedIn and work out who’s actually behind it. Would you know who you’d be dealing with?
3. Website — homepage first, then follow wherever the page pushes you next. Move through it as a visitor would; don’t audit the sitemap.
4. Proof — try to establish they’ve genuinely done this work. Who for, how recently, and what actually happened.
5. Contact — get as close to making contact as you can without submitting anything. Note how much effort it takes and what you’re asked to hand over.

At each step, report back: what you were trying to find out, what you actually found, what you looked for and couldn’t find, and whether your confidence went up, down or stayed level — and the specific thing that moved it.

Then finish with: a handback paragraph describing what this company does and who it’s for, as you’d explain it to the colleague who recommended them; the five things most likely to make someone in your position hesitate, ranked by whether they’d stop you making contact rather than by how easy they’d be to fix; one thing that raised your confidence; and your decision — email them, keep looking, or move on.

Two rules: judge substance, not design — ignore colours, fonts and layout unless they stop you finding something. And if you can’t tell something, say “I couldn’t tell” rather than guessing.

What happens when you run it

Open the browser pane from the top right and expand it, and you can watch the whole thing happen. It behaves like a real person would. It goes to a search engine and searches the company name, with no logins and no search history behind it — a genuinely clean start.

On our run against Candle Digital, it picked up the company LinkedIn page, found one of the co-founders and searched for him specifically, then moved on to the website. It noticed the homepage pushes visitors towards what we do, went looking for proof in our past work, checked the about page for the team, then worked out how to get in touch and how much friction there was in doing it.

It reached the discovery call page and stopped short of actually booking, but assessed what someone would have to do to get there. The whole thing took about ten minutes.

How to read what comes back

Because of the way the prompt is written, the report comes back in a fixed shape. For each of the five steps you get what it found, what it looked for and could not find, and where its confidence sat at that point.

The confidence line is the most useful part. On our run it started neutral and dipped slightly at the search stage — probably because it got blocked by Google. It recovered once it found the founders, went up again on the website, and climbed further on proof, where one particular case study clearly swung it. By the contact stage it found what it needed and confidence was up again.

Then comes the handback: what this prospective client would say about you to the person who recommended you. It identifies who you are, how many of you there are, what you do, and pointedly what it could not find — pricing, in our case.

After that, the five things most likely to make someone hesitate. Ours flagged a small footprint, case studies being self-published rather than independently verified, and founder dependency. And finally a decision: would this person actually make contact? In our case yes, because there was no commitment involved and it was straightforward to reach a founder.

An honest caveat

Results will vary depending on location and when you run it. What you are seeing is close to what a stranger would see, not identical to it. That is worth being clear about.

But it is objective in a way you cannot be about your own business, and it beats trying to replicate the process yourself with your own logins and your own assumptions filling in every gap.

Worth ten minutes of your time

Our run surfaced three or four things worth acting on. That is a good return on ten minutes, particularly for something you cannot see any other way.

You do not have to stop at your own business either. The same approach works on competitors, which tells you what a prospective client is likely to make of the alternatives sitting next to you on their shortlist.