Living case study

We are our own
worst case study.

A domain rating of zero, one referring domain and no organic traffic. This page carries our own search numbers from the day we started, updated from the same tools we would use on your site, including the months nothing moves.

Updated from a live pull · every figure names the tool that reported it

Why we publish our own numbers

Every agency case study is a graph that goes up, chosen after the fact from the client it went up for. You cannot check it, you do not know how many other accounts it was picked from, and you never see the ones that went sideways. So this is ours instead: a domain we own, measured from before it had done anything, republished on a schedule whether the line moves or not. If the method does not work you will see that here first.
The numbers

Every figure we have recorded, on every date we recorded it

Read from the dated pulls themselves. Each row names the tool that reported it, because the two we use disagree about this domain.

thecold.ai · 2 measurements · latest 3 September 2026
MetricTool2 September 20263 September 2026
Domain RatingAhrefs00
Referring domainsAhrefs11
Total backlinksAhrefs1616
Organic keywords rankingSemrush22
Monthly organic trafficSemrush00
AI visibility scoreSemrush00
AI mentionsSemrush00
Pages cited by AI answersSemrush00

Nothing has moved yet, and the table says so. The pages built on 2 and 3 September have not reached the live site, so a change in these figures at this point would mean the tools were wrong rather than that the work had landed.

Why our own domain

Three reasons this is more useful than a client win

You can verify every figure yourself

Those figures come from Ahrefs and Semrush, on a domain you can type into either tool right now and read for yourself. A client case study can only ever be checked by the client.

You see the starting line

Most case studies begin at the moment the work started, which is usually after a site already had authority. This one starts at zero, so the effect of the work is not mixed with the effect of a domain that was already trusted.

There is nowhere to hide a bad month

One domain, published on a schedule. We cannot quietly pick a different account to talk about, because there is no other account.

The work, dated

What has actually been done to this domain

Every change is a merged pull request with the reasoning in it. This is the list, so the numbers further down have a cause and not just a date.

2 September 2026

Measured the starting line before touching anything

Pulled the domain's authority and every candidate term against the UK database. Recorded a Domain Rating of 0, one referring domain and two ranking keywords, then published those figures on the service page rather than waiting for them to improve.

2 September 2026

Built three pages against live results, not guesses

A service page, a Middlesbrough page and a buyer's guide, each sized and structured from the results already ranking for its term. Each targets one intent, so none of them competes with the other two.

3 September 2026

Made the house rules run against the exported pages

The copy checks only ever read one folder, which did not contain the page copy. They now run over the words in the built HTML, so a claim or a supplier name cannot reach the site through a file nobody was checking.

3 September 2026

Recorded every term measured, and the verdict on each

Forty-nine terms with volume, difficulty, the number of referring domains behind the pages already ranking, and whether we judged it winnable. Three of them were dropped as unwinnable on links, which is the part of the record most agencies do not keep.

3 September 2026

Asked the AI engines whether they had heard of us

They had not, in every query that returned an answer. Recorded as a baseline before the pages went live, so the next run can be compared with it.

AI citation

Whether answer engines quote us

Two ways of measuring the same thing: a third-party score, and a set of questions asked directly. Both currently read zero, which is what a new domain should expect.

  • For "best AI SEO agency UK", every source the engine quoted was a third-party roundup rather than an agency's own page. That answer is won by being listed elsewhere, not by writing a better service page.
  • For "how much does SEO cost in the UK", the two pages the engine cited are two of the nine we source our own cost figures from. We are reading the same pages the engine chooses to quote.
  • For "SEO agency Middlesbrough", the local pack lists six agencies, every one with a street address. That is an eligibility requirement, not a ranking factor, and we do not meet it yet.

Only some of this can be automated. Two engines answered anonymously, one asked for a sign-up after two questions, and one served a bot check, which we did not work around. The rest is asked by hand once a month and dated when it is.

What happens next

What we have committed to publishing

  • The same five figures every month, from the same two tools, whether they rise, fall or sit still.
  • The date each was measured, taken from the pull itself rather than typed alongside it.
  • Which tool reported each figure, because the two we use disagree about this domain.
  • The terms we judged unwinnable, and whether that judgement turned out to be right.

No forecast, and no target. A new domain with one referring domain has no business predicting when it will rank, and a page that guessed would be the same graph-that-goes-up we are arguing against.

Questions about publishing our own numbers

Straight answers. If yours isn't here, ask it on the call: we'd rather tell you no than sell you the wrong thing.

Because it is true, and because it is the only number on the page that cannot be argued with. An agency that hides its own starting position is asking you to trust a method it will not demonstrate on itself. The figure is checkable in any backlink tool in about ten seconds, so publishing anything else would be a lie with a very short shelf life.

The method these numbers are testing is described on our SEO and AI search page. If you are choosing between agencies, the ten questions worth asking includes the two we answer badly.

Ask us what these numbers mean for you

Book 15 minutes and we will pull the same figures for your domain and the terms you care about, and tell you which of them are winnable from where you are standing.

No minimum term · we say when a term is not worth chasing