What is GEO (generative engine optimisation)?

GEO (generative engine optimisation) is the discipline of getting your site read, referenced, and cited by the 5 AI search surfaces buyers now use: ChatGPT, Perplexity, Google AI Overviews, Gemini, and Bing Copilot. It measures whether AI assistants can find your content, understand what you do, and recommend you when a buyer asks them a question. SEO answers "will Google rank me". GEO answers "will AI recommend me". In 2026 buyers ask both, so both scores matter.

What is SEO (search engine optimisation)?

SEO (search engine optimisation) is how you rank in Google. Technical crawlability, on-page signals, content depth, authority, and site speed. Done right, Google sends organic traffic for the queries that match your business. SEO has driven the bulk of qualified traffic for UK buyers for twenty years and still does. The shift in 2026 is that AI Overviews intercept some of those clicks before they reach your site, which is why SEO now has to sit next to GEO.

What is AEO (answer engine optimisation)?

AEO (answer engine optimisation) is the discipline of getting your site cited inside answer surfaces: Google AI Overviews, featured snippets, People Also Ask, and voice assistants. Where SEO measures rank and GEO measures whether AI engines like ChatGPT or Perplexity recommend you, AEO measures whether your content gets pulled into the answer the user actually reads. The technical floor (structured passages, FAQPage and HowTo schema, agent-readiness) overlaps with GEO, which is why Pulse covers AEO inside the same audit. 2 of the 7 Pulse categories (AI Citability and Platform Optimisation) pick it up directly. There is no separate AEO audit; if you run Pulse AI or Pulse Complete, AEO is included.

How much does a Pulse audit cost?

Pricing is quoted per buyer scope, not from a fixed menu. The catalogue is 9 SKUs (3 disciplines: Pulse SEO, Pulse AI, Pulse Complete; 3 tiers: Pulse Essentials, Pulse Growth, Pulse X) plus an ongoing retainer option. Use the contact form or email info@valux.uk for a quote. Most engagements scope in a 30-minute call so the proposal matches the actual site rather than a menu.

Should we use OpenAI, Anthropic, Google, or run our own model?

Architecture decision, not preference. Anthropic Claude leads on long-context reasoning and tool use; OpenAI on multi-modal and agent ecosystems; Google Gemini on cost-per-token at scale and on tasks where Google Search grounding matters; open-weight models (Llama, Mistral) on data-residency or cost floors below what hosted providers offer. Most production builds end up routing across two or three providers depending on the request shape, with a model-router layer hiding the choice from product code.

How do we know the AI feature is working in production?

Three signals tracked from day one. Quality: the eval harness runs against a frozen golden dataset on every deploy, plus a sampled LLM-as-judge pass on real production traffic. Cost: spend per feature, per cohort, per query type, with alerts on per-query cost regression. Latency: P50 and P95 response time, with streaming response budgets where the user expects sub-second feedback. Vibe-checking AI features in production is the failure mode this engagement is designed to avoid.

What if the AI feature is too expensive at our scale?

The model router is the first lever: cheap default (smaller, faster model) for the 80% of queries that don't need the flagship model, expensive fallback only when the cheap model declines or returns low-confidence. Caching is the second: deterministic retrieval results cache on the embedding hash, deterministic completions cache on the prompt hash, with TTLs tuned to the freshness budget. The third lever, fine-tuning a smaller model on your own data, is reserved for cases where the cost-per-query is structurally dominant in the feature's economics.

UX Studio

Specialised engagement

AI feature integration.

A production AI feature that does what it promises and does not balloon your inference bill. Architecture chosen against the actual problem (retrieval-augmented generation for grounded answers, an agent loop for multi-step tools, a fine-tuned classifier for narrow tasks); evaluation harness running before any prompt ships, so quality regressions get caught at CI time rather than at customer-complaint time; cost and latency budgets agreed at scope and enforced at every model call.

Timeline: 6 to 12 weeks
Scope: Quote-driven

Book a scoping call Send a brief

01What ships at engagement end

Concrete deliverables. 9 line items.

Every item below is a real artefact, document, or component handed over at the end of the engagement. No placeholders, no abstract milestones.

01Architecture decision record (RAG vs agent vs classifier vs hybrid)
02Evaluation harness with golden-dataset + LLM-as-judge passes in CI
03Vector store and embedding pipeline (when RAG is the architecture)
04Tool layer with structured-output schema validation (when agent is the architecture)
05Prompt layer with version control, A/B testing, and rollout flags
06Model router: cheap default, expensive fallback, by-feature overrides
07Cost dashboard tracking spend per feature, per user cohort, per query type
08Latency budget enforced via streaming responses and parallelised tool calls
09Observability: trace every call, log every prompt + completion, alert on regression

02More in Specialised

Adjacent services in the same group.

See all ten services

Service questions