AI SEO / Answer Engine Optimisation
Your next client will ask an AI before they ask you.
When someone types "best [your category] in [your city]" into ChatGPT, three or four companies get named. That shortlist is now the top of your funnel — and it is decided by signals most agencies are not even measuring. This is the part of the market we specialise in.
- ChatGPT · Perplexity · Gemini · AI Overviews
- Citation tracking included
- Practised on this very site
The short version
How AI SEO differs from SEO
Traditional SEO competes for a position in a list of links. A person scans ten results and picks one. Being fourth still gets clicks.
AI SEO competes to be inside the answer itself. A language model reads a question, synthesises a response from sources it trusts, and names two or three companies. There is no fourth place. You are cited, or you are invisible.
The two disciplines overlap — pages that rank well are more likely to be retrieved and cited — but the optimisation targets are genuinely different:
- Retrieval, not ranking. Answer engines pull passages, not pages. Content has to be structured so a single self-contained chunk answers a question without the surrounding page.
- Entities, not keywords. Models reason about things — your company, your services, your locations — and how consistently they are described across the web. Contradictory facts make you a risky citation.
- Corroboration, not backlinks. A model is more likely to name you if independent sources agree you exist and do what you claim. Mentions, directories and reviews carry weight even without a link.
- Extractability, not engagement. Clear definitions, direct answers, tables and clean HTML beat a 2,000-word narrative that buries the answer in paragraph fourteen.
- Freshness is weighted harder. Retrieval-augmented systems favour content with visible, recent, verifiable dates.
Anyone who tells you AI SEO is "just good SEO" has not run the tests. The overlap is real; the difference is where the opportunity is.
What's included
The AI SEO programme
Every item here is something we can measure before and after. Where a tactic is still emerging and unproven, we say so rather than bill you for a hunch.
AI visibility baseline
We run your real buying questions through ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews and record who gets named. That is your starting line — most companies have never seen it.
Prompt & question mapping
Buyers do not type keywords into an AI, they type sentences. We build a library of the actual prompts that precede a purchase in your category, and target those.
Entity optimisation
Your business described identically everywhere that matters — site, schema, Wikidata, Crunchbase, LinkedIn, directories, review platforms. Consistency is what makes a model confident enough to cite you.
Structured data at depth
Organization, Service, FAQPage, HowTo, Product, Article and BreadcrumbList JSON-LD, correctly nested and validated. Machine-readable facts are the cheapest trust signal available.
Extractable content architecture
Question-shaped headings, answer-first paragraphs, definition blocks, comparison tables and summaries, so a retrieval system can lift a complete answer from one chunk.
llms.txt & AI crawler policy
A published llms.txt describing your site for language models, plus a deliberate robots policy for GPTBot, ClaudeBot, PerplexityBot and CCBot. A commercial decision made on purpose rather than by default.
Corroboration & mentions
Getting your business accurately described on the third-party sources models actually retrieve from — directories, review sites, comparison pages, publications.
Rendering & crawl access
Most AI crawlers do not execute JavaScript. If your content only exists after hydration, you are invisible to them. We test what each crawler genuinely sees, and fix it.
Citation tracking & reporting
Monthly re-runs of your prompt library across every major engine, plus AI referral traffic isolated in analytics. Share of voice, tracked over time.
How we work
Measure, fix, re-measure.
This field is young and full of confident nonsense. We treat every tactic as a hypothesis and hold ourselves to the citation numbers.
- 01
Baseline
Weeks 1–2
Build your prompt library, run it across every major answer engine, and document who is cited today and why. You get the raw transcripts, not a summary.
You receive
- Prompt library
- Cross-engine citation baseline
- Competitor citation analysis
- 02
Make the site machine-legible
Weeks 2–6
Structured data, entity consistency, crawler access, llms.txt, and restructuring key pages so answers are extractable in a single chunk.
You receive
- JSON-LD implementation
- llms.txt
- Crawler access fixes
- Content restructuring
- 03
Build corroboration
Ongoing
Publish the definitive answers to your category questions, and get your business consistently described on the sources models retrieve from.
You receive
- Answer-first content
- Directory & profile alignment
- Third-party mentions
- 04
Track & iterate
Monthly
Re-run the prompt library, report share of citations against named competitors, and reallocate effort toward whatever actually moved. We publish what did not work too.
You receive
- Monthly citation report
- AI referral analytics
- Revised backlog
Engines and tooling we test against
The landscape changes monthly. This list is reviewed every quarter.
- Answer engines
-
- ChatGPT Search
- Perplexity
- Google AI Overviews
- Gemini
- Claude
- Copilot
- Visibility tracking
-
- Profound
- Peec AI
- Otterly.ai
- Custom prompt harness
- Structured data
-
- Schema.org
- Rich Results Test
- Schema Markup Validator
- Entity
-
- Wikidata
- Google Knowledge Panel
- Crunchbase
- Industry directories
- Crawl verification
-
- GPTBot / ClaudeBot / PerplexityBot log analysis
- Server-side render checks
Questions
Fair scepticism, answered
Is AI SEO real, or is it repackaged SEO?
It is partly overlapping and partly genuinely new. Pages that rank well are more likely to be retrieved, so classic SEO still helps. What is new is optimising for retrieval and citation rather than ranking: chunk-level extractability, entity consistency across the web, structured data depth, and crawler access for models that do not run JavaScript. We measure citations directly, so you can judge whether the extra work paid for itself.
Does traffic from AI answers actually convert?
Volume is lower than classic organic and intent is typically higher, because the person arrives already shortlisted rather than still comparing. The larger effect is often indirect: a buyer asks an AI, sees your name, then searches for your brand. We isolate AI referrers in analytics and watch branded search alongside them so you can see both effects.
What is llms.txt and does it do anything yet?
It is a proposed plain-text file at your domain root that gives language models a curated map of your most important content. Adoption by model providers is not universal and it is not a confirmed ranking factor. It costs very little to publish, it forces your content structure to be explicit, and it positions you correctly if adoption grows. We are upfront that it is a low-cost bet, not a proven lever.
Should we block AI crawlers to protect our content?
That is a commercial decision, not a technical one, and it cuts both ways. Blocking protects content from training use but also removes you from the answers your buyers see. For most service businesses, visibility is worth more than the content itself. For publishers whose content is the product, the calculation is different. We lay out both sides for your specific case rather than apply a house policy.
How do you prove it worked?
A fixed prompt library, re-run monthly across every major engine, showing how often you are cited and where you sit against named competitors. Plus AI referral traffic isolated in GA4 and branded search volume tracked in Search Console. The baseline is recorded before any work starts, so the comparison is honest.
Can you do this without the SEO retainer?
Yes, but you will get less from it. Retrieval leans on many of the same signals that produce rankings, so an AI SEO programme on a technically broken site underperforms. If the classic foundations need work first, we will tell you before you spend on this.
Investment
Simple pricing. No surprises.
One price, agreed before we start. It does not change unless you ask for something extra, and you will know the cost before we do it.
Basic
Everything a small business needs to look credible and be found.
$300 one-off
- A complete, mobile-first website
- Written for your customers, not for us
- Set up so Google can find you
- Contact form that reaches your inbox
- Loads fast on a phone
- Yours to keep — no monthly lock-in
Advanced
For businesses with more to say and more to sell.
$500 one-off
- Everything in Basic
- A page for each service you offer
- Photo gallery of your work
- Local search setup for every area you serve
- Customer reviews built into the page
- Booking or quote request flow
- A month of changes after launch
Enterprise
For larger projects, online shops and anything custom.
Let's talk quoted after a call
- Everything in Advanced
- Online shop or customer login
- Custom features built for your business
- Ongoing SEO and marketing support
- Priority response
- A named person who knows your account
Prices are for the work described. If your project needs something outside it, we will tell you what it costs before starting — never after.
Ask an AI who it recommends in your category.
If your name is not in the answer, send us the screenshot. We will tell you what stands between you and that shortlist.