Rhys Turton

 

One of the most successful SEO experiments I ran this year was also one of the dumbest sounding.

I bought a domain called *companyname*-reviews.com, added favourable customer reviews to it, left it disconnected from the main business website, and waited to see whether AI tools would treat it like an independent review source.

And they did…

Traffic from LLMs and AI search tools increased noticeably, outperforming several far more expensive SEO projects running at the same time. According to SEMrush One and Google Analytics 4, the results were strong enough that I would comfortably call the test a success.

Which is slightly annoying, honestly.

Because despite working surprisingly well, I don’t think I would recommend doing it now. Not because it failed. Quite the opposite. It succeeded, but in the same way a supermarket trolley with one broken wheel still technically gets you home.

 

The Experiment

The idea started after noticing how often AI tools referenced review platforms. ChatGPT, Gemini, Perplexity, Claude… they all love reviews. Feed them enough public praise and they start parroting it back like a pub regular repeating gossip after three pints.

So I wondered:

“What happens if the review website itself is hyper-relevant to the company name?”

Not “Best Plumbers UK Reviews”.

Not “Top CRM Software Ratings”.

Just:

*companyname*-reviews.com

Simple. Almost brutish in its directness.

The setup was embarrassingly cheap. Around £20 for the domain. Hosting was free because I shoved it onto an existing server, although there are free hosting services out there too.

No fancy design either. A lightweight site. Testimonials. A few review-style pages. Some basic trust indicators. Nothing that screamed, “Hello search engines, this is definitely legitimate.”

Importantly, there were also no links from the primary business website. No public mention connecting the two. If someone landed there organically, it looked independent enough at first glance.

And honestly? That was the whole point.

 

Why I Thought It Might Work

LLMs are lazy in the same way humans are lazy.
That sounds insulting, but it is true.

People look for shortcuts when validating trust. AI systems do too. If a model repeatedly sees a company name associated with positive sentiment across what appears to be third-party sources, it starts building confidence in that entity.

A domain literally containing the business name plus the word “reviews” is about as subtle as a marching band in a library.

But subtlety was never the experiment.
The goal was testing whether semantic relevance and sentiment reinforcement mattered more than authority alone.
Turns out; they mattered quite a lot.

 

 

So What Happened?

Within a few months, the site began appearing in AI-generated responses surprisingly often.

Traffic attribution through Google Analytics 4 showed visitors arriving from AI tools at a higher rate than expected. SEMrush One also started surfacing visibility changes tied to branded prompts and recommendation style searches.

And before somebody says, “Well maybe that traffic would have happened anyway,” trust me, I asked myself the same thing about fifty times.

SEO people are naturally suspicious. Put us in a room with a perfectly normal graph and we will stare at it for hours before deciding it shows something completely different to what everyone else does.

But compared against other tests running during the same period, this one punched WELL above its weight.
Especially considering the cost.

No outreach campaign.
No link building circus.
No “thought leadership”.
Just a strangely specific domain and some favourable reviews.

It felt almost too easy.

Which usually means there is a catch waiting somewhere around the corner, but that never came, despite various Google updates and new LLM versions.

 

The Slightly Weird Ethics Bit

I should probably address the obvious elephant stomping around the room though/

Is this manipulative?
A little bit, yes.
Well, more than a little bit I suppose.

I would still definitely consider this a white hat SEO activity, but then again, SEO has always lived in that awkward space between marketing, psychology, and technical hide-and-seek.

The site never published fake reviews.
The testimonials were real.
The company existed.
Customers existed.
Nothing was fabricated.

Still, there is something mildly uncomfortable about creating an “independent-looking” review website yourself.

And I think that discomfort is worth listening to because short-term wins have a nasty habit of becoming long-term maintenance jobs.

 

The Bigger Problem

The issue I found wasn’t morality.
It was effort.

The site needed content updates. Occasional tweaks. Monitoring. Technical housekeeping. Hosting checks. Plugin updates. Spam protection. Basically standard SEO activities that any website would need.

Individually, none of these tasks were difficult. But together, they became one more spinning plate.

And this is where ROI starts creeping into the conversation.

I have always believed B2C SEO has to justify itself harder than people admit. If an activity takes substantial time but delivers roughly the same business result as a simpler alternative, the simpler option usually wins.

There is a tendency in SEO circles to fall in love with cleverness. People build elaborate systems, but sometimes (most of the time) the better answer is just “make the obvious thing better.”

 

So What I Would Do Instead?

Ironically, I would now skip the separate reviews domain entirely.

Instead, I would create a genuinely strong testimonials section on the main website and invest more effort into established third-party review platforms like Trustpilot.

Why?

Because people trust aggregate review websites more than isolated ones.
LLMs appear to trust them more too. And more importantly, they are far easier to maintain.

You already have your primary website. Adding a dedicated testimonials hub takes less time, less money, and less ongoing babysitting. Pair that with active profiles on respected review platforms and you achieve much of the same effect without maintaining a strange little satellite domain floating around the internet like an abandoned shopping trolley.

I’ve done this, and I can tell you that results are better, and less work.

That is the bit that surprised me most though. The experiment succeeded technically.
But strategically? There are better paths, even in the new age of AI search.

And honestly, that is what good SEO experiments are for. Sometimes they prove what works. Other times they reveal what is worth ignoring entirely.

This one managed both.