AI SEO Agency

“AI SEO” has become one of those phrases that means whatever the person selling it needs it to mean. Before you buy any of it, it is worth separating the two things wearing the same name, because only one of them is a service you should pay a monthly fee for.

Two things are called AI SEO. Only one is a service.

The first is using AI tools to produce SEO work faster — mostly content generation. That is a workflow, not a service, and buying it by the month rarely ends well. The second is making a brand visible in AI-generated answers as well as in blue-link results: rankings, citations, entity clarity and measurement, run as one workstream. That is the service. This page describes the second.

What’s included

We do not run separate SEO and GEO campaigns, because the underlying work overlaps almost entirely and splitting it produces two invoices and one result. A single workstream covers:

Rankings. The conventional discipline, unchanged and still the largest lever. Keyword and intent mapping, on-page work, information architecture, internal linking and technical health. Assistants overwhelmingly draw their answers from pages that already rank, so this is not legacy work you do alongside AI visibility — it is most of how you get AI visibility.

Citations. The prompts your buyers actually ask, tracked across ChatGPT, AI Overviews, Gemini, Perplexity, Claude and Copilot, with work aimed at the sources those engines cite. That frequently means off-site: review platforms, comparison articles, trade publications and forums that the models already trust.

Entity clarity. Making the models’ understanding of your brand coherent and correct — what you sell, where you operate, who you compete with, what you are known for. Contradictory or thin entity data is the most common reason a page-one site never appears in the answer above it.

Schema and extractability. Valid structured data, server-rendered HTML that assistant crawlers can read, and key facts stated plainly enough to be lifted. Assistant crawlers are considerably less patient with client-side JavaScript than Googlebot is.

Measurement. A repeatable prompt set, re-run on the same schedule every month, alongside conventional rankings and organic revenue reporting.

Where AI helps us, and where we ban it

We use AI heavily, and we are specific about where.

It is genuinely good at clustering — grouping thousands of keywords into topics, spotting intent patterns across a large query set, and finding gaps between your coverage and a competitor’s. It is good at quality assurance: checking schema validity, flagging internal linking gaps, comparing rendered against raw HTML, spotting broken hreflang across a large site. It is good at drafting technical tickets from findings we have already made. In all of those cases a person checks the output before it goes anywhere.

We do not publish AI-written copy. Not on client sites, not on ours. Every page we publish has a named human author who is accountable for it, and every claim in it comes from first-hand data — a client campaign, a measured result, a tool we ran — rather than from a summary of what other pages on the internet already say.

This is a commercial position as much as an ethical one. Content that only restates the consensus has nothing an assistant would prefer to cite over the sources it already trusts. The pages that get cited contain something the model cannot get elsewhere: original data, a named expert’s judgement, a measurement someone actually took. A generated article is, by construction, the average of everything already published. Being the average is the one position from which you cannot be cited.

Proof

Official London Theatre reached 85% AI search visibility and is the #1 recommended source for its core queries — the clearest evidence we have that the combined approach works.

Koyo Loans shows the ranking half of the same workstream: a top three position for “loans”, a keyword with 95% difficulty, in a sector dominated by banks and comparison sites, with a 73% increase in organic conversions and non-branded organic traffic eventually overtaking branded.

Amulet started from no measurable search presence at all and reached a 123%+ year-on-year increase in organic traffic with a 437% increase in search visibility.

How we report both

One monthly dashboard, two halves. The organic half reports non-branded organic sessions and revenue, ranking movement on the commercial keywords the campaign was built to shift, and the referring-domain gap against your named competitors. The AI half reports citation share across the fixed prompt set, broken down by engine, alongside which sources are being cited in your place and what changed since last month.

Both halves use the same prompt set and the same keyword list every month, so the comparison is real. We also report what did not move, and why we think it did not.

What it costs

SEO retainers start from £1,100 a month ex VAT. GEO is inside every one of them — there is no AI visibility upgrade, no separate GEO line item and no premium tier that unlocks citation tracking. Charging twice for one workstream would be hard to justify when the underlying work is the same work.

One-off audits and project work are scoped to the brief. If you want to see where you stand before committing to anything, run our free GEO visibility score and our free website SEO audit first, then talk to us about what the numbers mean.

Frequently asked questions

Is AI SEO just SEO with a new name?

Largely, yes — and anyone telling you it is a wholly new discipline is selling something. The foundations are identical: rank well, be crawlable, be trustworthy. What is genuinely new is a second surface to be visible on, a different measurement problem, and much greater weight on off-site sources and entity clarity. That is an extension of the work, not a replacement for it.

Do you use AI to write our content?

No. We use it for keyword clustering, gap analysis and technical QA. Published copy is written by a named human author working from first-hand data, and we will tell you who wrote each page. Generated content has nothing an assistant would prefer to cite over its existing sources.

Should I still invest in traditional SEO?

Yes, and it should be most of the budget. AI answers draw heavily on pages that already rank, so ranking well remains the single most reliable route into being cited. A brand that abandons conventional SEO to chase citations usually loses both.

Do I need a separate GEO agency?

You need one team doing both, because the work is the same work. Splitting it across two suppliers means two audits of the same site, two content plans that contradict each other, and no one accountable when the numbers diverge.

How do you measure AI visibility?

A fixed set of buying-intent prompts, re-run monthly across every major engine, reporting how often you are mentioned, how often you are cited with a link, and which sources appear instead of you. Because these systems are non-deterministic, we report share across repeated runs rather than treating a single answer as fact.

How long before we see results?

Technical and indexing fixes can register within days. Ranking movement on commercial terms typically shows over three to six months. Citation share tends to follow ranking with a lag, because the engines have to recrawl and re-evaluate the sources they draw on. Anyone promising AI visibility in weeks is describing a different, riskier method.

Does llms.txt help?

There is no evidence that it does. No major AI platform has confirmed it uses the file, and we have seen no measurable effect from adding one. We will not present it as a deliverable. Server-rendered HTML, valid schema and pages worth citing are what actually move the needle.

Which markets do you work in?

The UK primarily, with established campaigns in Ireland and Australia. AI visibility is market-specific — an assistant’s answer to the same question differs by country — so we run and report the prompt set separately for each territory in scope. You can see who you would be working with on our team page.