Get found on Google.
Get quoted by the AI.
Most buyers now get an answer before they get a list of links. We build pages that win both, scoped from a live pull of the results for your terms: who ranks now, how their pages are built, and whether you can realistically beat them.
We tell you before we start if a term is not winnable
What TheCold.Ai builds
Two gates, not one
Ranking and being quoted are different problems. A page can rank and never be cited, and a page can be cited without ranking first. Both have to be engineered.
Gate one: get into the retrieval pool
Before a language model can quote you it has to find you. That is relevance, entity coverage and a flat structure a crawler can read without executing anything. Most pages fail here for mechanical reasons rather than editorial ones.
Gate two: be the passage that gets extracted
Once you are in the pool, the model lifts a block. Answers have to sit in real block-level containers, sized to the retrieval window, so a whole self-contained answer can be taken out intact. A good answer buried mid-paragraph does not survive extraction.
SEO, AEO and GEO, without the jargon
Three terms get used interchangeably by people selling this, and they are not the same thing. The difference decides what work your money buys.
| Term | What it means | What it wins you |
|---|---|---|
| SEOSearch engine optimisation | Getting your page into the list of blue links, and as near the top of it as possible. Still the largest source of search traffic for most B2B businesses. | Position on the results page |
| AEOAnswer engine optimisation | Getting your page quoted in the answer that appears above those links, or inside ChatGPT and Perplexity. Depends far more on how the page is structured than on where it ranks. | Being the source that gets cited |
| GEOGenerative engine optimisation | The broader job of making a language model treat your business as a known entity, so it mentions you even where no single page is being quoted. Mostly won off your own website. | Being recalled at all |
Most agencies sell the first and hope the other two follow. They sometimes do, but not reliably: a page can hold first place and never be quoted, because the answer a model needed was written as a long paragraph rather than a block it could lift out whole.
What actually runs before a word gets written
Most SEO proposals are one person's opinion with a keyword tool screenshot stapled to it. This is the pipeline underneath ours: what we pull, what we read, and what you get out of each stage.
Nothing here is a guess about your market
Every brief starts as a live pull of the results people actually see for your terms, joined to third-party authority and keyword data, and joined again to your own Search Console once we have access. Three sources that disagree with each other, which is the point: where they disagree we publish both figures rather than picking the flattering one.
Demand and difficulty
Search volume, difficulty, cost per click and buyer intent for every candidate term, in the country you sell in. Ranked by what a term is worth rather than by how big it looks.
Can you win it
Domain rating, backlinks, referring domains and broken links, for your site and for the pages currently ranking. When the gap is unbridgeable we say so before you spend anything.
Read the live results
The current top ten pulled and parsed: how long each page is, how it is structured, which topics every one of them covers, and which none of them do. That last set is where a new page earns its place.
Build to be retrieved and quoted
Written in self-contained blocks sized to how an answer engine reads, with answers in real containers rather than buried in paragraphs, headings named for actual things, at least one data table, and structured data underneath.
Score it before it ships
Fifty-eight checks, forty-nine to pass, scored and handed over with the page. Anything a specialist would call filler goes back, and every factual claim is tagged and checked against a source before it publishes.
Corroborate, then watch
Companion pieces on platforms search engines already trust, so your site is not the only thing on the internet describing your business. Then your own Search Console data, read on a schedule, including where two of your pages are quietly competing.
The pipeline, stage by stage
| Stage | What runs | What you get |
|---|---|---|
| Keyword economics | Volume, difficulty, cost per click and buyer intent for every candidate term, from Ahrefs and Semrush data in the country database you sell into | A shortlist ranked by what each term is worth, including the terms we advise against and why |
| Authority check | Domain rating, backlink and referring-domain counts, broken links and the real competitor set, for your domain and for the ranking ones | A winnable or not-winnable verdict, with the comparison it rests on, before you commit |
| Live results pull | The current top ten from a live results API, with the ranking pages parsed for length, heading structure and topic coverage | A brief built from what is winning today, sized to the length that is actually competitive |
| Question and phrase capture | The questions the results page shows alongside your term, and the exact phrases the engine bolds in its own snippets | An FAQ that answers what people ask, in the words they use, rather than what we imagined they ask |
| Supporting-page gap analysis | The internal pages your competitors link to that you have no equivalent for, ranked by how many of them have one | A build order for the supporting pages, so the set is planned rather than accumulated |
| Your own search data | Your Search Console queries, positions, impressions and clicks, pulled on a schedule once you grant access | Where two of your own pages compete for one search, which to keep, and what to do with the other |
| Quality gate | Fifty-eight checks covering information gain, structure, extractability, schema and honesty, with forty-nine required to pass | The scorecard itself, so you can see what passed and what did not before it goes live |
| Scheduled publishing pipelines | Where a sector moves fast enough to need it: jobs that search the day's news, discard what is already covered, rewrite each item as an original piece with attribution and a link to the source, publish it and notify the search engines within minutes | A site that keeps publishing without anyone opening a laptop, built for you and running on infrastructure you own |
What a normal SEO retainer does, and what we do instead
Not a criticism of agencies that do the standard thing well. It is a different shape of work, and it suits a different buyer.
| Dimension | A typical SEO retainer | TheCold.Ai |
|---|---|---|
| What you buy | A monthly retainer, with the work decided as it goes | Named pages, each scoped before it is written |
| Where the brief comes from | Experience and a keyword tool | A live pull of the current results: competitor structure, word counts, the questions people actually ask |
| Whether the term is winnable | Rarely discussed before the invoice | Checked first. If the pages ranking have far more referring domains than you do, we say so and pick a different term |
| AI answers | Treated as a side effect of ranking | Engineered separately, as AEO and GEO, because a page can rank and never be quoted |
| Proof it works | Case studies for businesses you cannot check | This site. We publish our own search numbers, dated, including the bad ones |
| Commitment | Six or twelve months | Monthly rolling, no minimum term |
Four steps, and you see the brief before anything is written
Pull the live results
We take the current search results for your term: who ranks, how long their pages are, how they are structured, and the questions searchers actually ask alongside it.
A brief built from what is winning today
Check the term is winnable
We compare the referring domains behind the ranking pages against yours. If the gap is unbridgeable we tell you before you spend anything, and we propose the nearest term that is not.
An honest go or no-go
Build the page
Written to the structure the brief calls for, with answers in extractable blocks, a real data table, and structured data so both a crawler and a language model can parse what you do.
A page scored against a 58-point gate before it ships
Support it off-site
A page with nothing pointing at it stays invisible however good it is. We build the off-site corroboration that makes search engines and answer engines trust the claim.
Signals that hold the ranking up
The parts we will not move on
Four numbers that describe the process, and seven things we commit to in writing. None of them is a promise about a ranking, because nobody controls that. All of them are promises about what we do, which we do control.
- 58
- checks before a page ships, 49 to pass
- <200
- characters in the answer block above every page
- 0
- words written before you approve the brief
- 30·60·90
- day reviews, with the numbers attached
- You see the brief, and the live data behind it, before anything is written.
- Every term gets a winnable or not-winnable verdict first, and we tell you when the answer is no.
- Every factual claim is tagged and checked against a named source before the page publishes.
- No keyword padding. Entities go in once, where they count, never repeated to hit a density score.
- One service, one place, one page. We will not build you five town pages with the name swapped.
- Where two data sources disagree about your market, you get both figures and which one we acted on.
- The pages sit on your website and the code is yours. Nothing depends on a licence from us.
What is actually delivered, per page
Every item below is a thing you can look at, not an activity we can claim to have done.
Before anything is written
- A brief built from the live results for your term, showing who ranks now and how their pages are put together
- The questions searchers ask alongside your term, taken from the results page rather than imagined
- A winnable or not-winnable verdict, with the referring-domain comparison it is based on
- The word count and heading structure the current top pages actually use
The page itself
- A summary block engineered to be lifted whole into an AI answer
- Answers in extractable containers under headings named for real things, not "Overview"
- At least one genuine data table, because tables get quoted and bullet lists get paraphrased
- Structured data describing the business, the service and the questions, so a machine reads the same page a person does
- An honest section on who the service is wrong for
After it ships
- Submission for indexing, and a check that it actually got indexed
- A baseline of where you started, so the next report is a comparison and not a claim
- Off-site corroboration, so the page is not the only thing on the internet saying what you do
- A review at 30, 60 and 90 days with the numbers attached
What this costs, and what the rest of the market charges
We surveyed every UK page ranking for the SEO price question and wrote down what each one charges. The ranges differ by a factor of sixty, which says more about the question than any single figure could. That survey has its own page.
We price per page rather than per month, so you can buy one, watch what it does, and stop if it does nothing. What we will not do is name a retainer before looking at whether your terms can be won, because that number would be a guess dressed up as a quote.
Priced per page, not per month of hoping
We scope SEO on the call, because the honest price depends on how competitive your terms are and how much off-site work they need. What we will not do is quote you a retainer before we have looked at whether your terms can be won.
Our outbound pricing is public and itemised down to what delivery costs us, and the SEO pricing will be published the same way. See what we charge for outbound.
What people ask before buying SEO
The first four come straight from the questions the results page shows alongside this term. The rest are the ones people ask us on the call, including the two about tooling that most agencies would rather not answer.
Where to go next
Want to know if your terms are winnable?
Book 15 minutes. We will pull the live results for a term you care about, compare the referring domains behind the pages ranking against yours, and tell you whether it is worth chasing. You keep the research either way.
Monthly rolling · we say when a term is not winnable