Measurement first
We never start with tactics. Every engagement begins with a clear picture of current AI visibility.
Version 1.1 · Proprietary
How we work
Every engagement runs through the same core method. Repeatable, measurable, and focused exclusively on how often — and how favourably — a brand is cited or recommended inside AI-generated answers.
Guiding principles
Attribution over activity
Every Structure, Engineer or Amplify action is tied to a hypothesis linked to specific prompts. We do not ship work that cannot later be tested.
Compounding is the product
AI engines build persistent entity and source models over time. One-off work decays; consistent, measured signals compound.
Models trust sources, not pages
Entity clarity, corroboration and consistent description matter more than classic on-page tricks.
Answer-first content
Content is structured so generative systems can extract, attribute and trust it.
Continuous proof
A fixed, version-controlled prompt set is re-run on cadence so progress is visible and accountable.
Pure GEO
We optimise for citation and recommendation inside generative engines — not for ten blue links.
Five diagnostic layers
When a brand is not selected, we diagnose against these layers. Work on a higher layer produces limited results if a lower one is still broken.
01
Retrieval probability
Is the content eligible to enter the candidate set at all — crawlable, accessible, topically relevant?
02
Extractability
Can a generative system cleanly parse, extract and attribute a usable passage?
03
Entity clarity
Is the brand described clearly, consistently and with the right relationships across the open web?
04
Corroboration
Are independent, trusted sources reinforcing the brand’s claims and entity?
05
System memory
Has the brand established lasting associations that survive model updates and time?
The six stages
Stage 01
Measure
We define a category-specific prompt set (typically 40–120 prompts) covering brand, competitor, category, problem/solution, comparison and local prompts. Each is run across priority engines. We record mention, citation, sentiment, competing brands, and supporting sources, then produce a baseline Visibility Score and Share of Voice.
Outputs
- Baseline report
- Priority prompt clusters
- Trusted-source list
Success. A quantified picture of current AI visibility and the prompts that matter most.
Stages 1 and 2 are diagnostic and always completed first. Stages 3–5 are delivery and run in parallel where sensible. Stage 6 runs continuously on retained work and feeds learning back into the earlier stages.
How stages map to products
| Product | Stages | Cadence |
|---|---|---|
| Geo Audit | Measure + Map, with high-level recommendations for stages 3–5 | ex VAT · 2–4 weeks |
| Geo Monitor | Measure + Prove & Iterate | ex VAT · monthly |
| Geo Programme | Full cycle, continuous monthly iteration | Typical mid-market £4,500 – £6,500 · ex VAT |