ChatGPT
yourorgFT
Surface we shape
AI now answers what search used to return.
ChatGPT, Perplexity, Gemini, Microsoft Copilot and Google’s AI Overviews don’t rank pages — they synthesise answers from a source layer most marketing teams aren’t built to manage. Generative engine optimisation is the discipline of being the source those systems cite. Specific to the brand, the buyer, and the channels. Technical, structural and accountable to evidence, and relatively new as a discipline, which means most organisations don’t yet have a partner for it.
How it shows up
The mechanics.
Three visible outcomes of the work — what AI actually surfaces, where the citation web is thin, and what improvement looks like over time.
01 — Answer sweep
What AI actually says, side by side.
The same buyer question, asked to ChatGPT, Perplexity, Gemini, Microsoft Copilot and Google's AI Overviews. Different answers, different citations, different gaps — the surface every prospect now lands on first.
Prompt What does [your organisation] do?
Perplexity
ReutersForbes
Copilot
yourorg
Gemini
Trade press
Google AI
yourorgSector
02 — Citation graph
Where the source layer is thin.
A map of the third-party sources large language models treat as authoritative for your category — and where the brand isn't yet present. The diagnostic that drives every correction and every citation-network workstream below.
Q. What is [your organisation] known for?
03 — Compounding
Improvement that holds across model updates.
Share-of-answer measured by quarter, across all five major surfaces. Source-layer work compounds — every training cycle re-ingests the same authoritative web — so corrections carry forward and gains accumulate.
Share-of-answer · 5 quarters Each block = quarter's added share. Stacking = compounding.
Source-layer corrections compound — every model training cycle re-ingests the same authoritative web.
What we do
The work.
The discipline, broken down. Each lane is an executable workstream — measured, evidenced, recurring.
Prompt set & baseline
We build the questions your buyers actually ask AI into a tracked prompt set, then run it across ChatGPT, Perplexity, Gemini, Microsoft Copilot and Google's AI Overviews, in the geographies that matter. The result is a repeatable, comparable baseline of whether the brand appears, how it's described, and to whom.
Composite score 47/100
- Technical foundations 82
- Schema integrity 62
- Citation web 41
- GEO coverage 47
- AEO surface 31
Visibility tracking
The metrics that matter when there are no rankings: share-of-answer (how often the brand appears), average position when it does, and sentiment (how the models frame it). Tracked over time across every engine in the set, so movement is observable rather than anecdotal.
24 prompts
× 5 models
× 3 regions
= 360 tests
Competitive benchmarking
The same metrics against the competitors stakeholders genuinely weigh. Where rivals hold share-of-answer that should be the brand's, and where they're cited and the brand isn't. The comparative read that sets the priorities.
- Branded queries +12
- Category queries +28
- Buyer questions +37
- Comparison queries +46
- now gap target
Source & citation analysis
Which sources the models actually cite for your category — industry press, reference sources, review and community sites — and where competitors appear in them and the brand does not. The citation web mapped into a concrete coverage and source-building opportunity list.
Source-layer corrections & citation building
The execution a monitoring tool can't do. Structured data, knowledge-graph and reference-source fixes at the point the mis-description forms, plus presence built in the digital sources LLMs treat as credible: discreet, evidenced, and held to each source's editorial standard.
- Industry analyses +6
- Sector journals +4
- Reference databases +4
- Trade archives +4
Continuous re-measurement & reporting
Standing re-runs of the prompt set on a defined cadence, reported on a dashboard leadership can read. Corrections compound, new gaps surface, regressions are caught. The measurement loop is the engagement.
- Q1 Baseline measurement
- Q2 First re-test · improvement visible
- Q3 Gap corrections compound
- Q4 Share-of-answer milestone
Common questions
What to expect.
Straight answers on timeline, ethics, confidentiality and how engagements run.
How is this different from traditional SEO?
Traditional SEO optimises pages to rank in search results. Generative engine work shapes which sources AI assistants treat as authoritative and cite when synthesising answers. Different signals, different content structures, different success criteria. The two are complementary (most engagements combine both) but they're not the same discipline.
How do you measure visibility when models don't expose ranking data?
We run priority queries through ChatGPT, Perplexity, Gemini, Microsoft Copilot and Google AI Overviews on a standing cadence, capture the responses and citations, and track them over time. The measurement is observational — share-of-answer, accuracy of description, sentiment, and source coverage. It's the same evidence base a buyer evaluating your firm would see when they go and look.
Which AI platforms do you cover?
ChatGPT, Perplexity, Gemini, Microsoft Copilot and Google's AI Overviews and AI Mode at the surface layer, with Claude added where the brief calls for it. Each has different retrieval mechanics and different sources of authority. We weight effort to the platforms whose audiences and queries matter to the brief.
How quickly do changes show up in AI answers?
Source-layer corrections — knowledge graph entries, structured data, authoritative reference sources — start to influence model outputs within weeks. Citation network development is a longer arc that compounds as authority signals accumulate over time. We set milestones at both intervals so progress is visible early.
What about when AI models update their training data — does this work compound or reset?
It compounds. Models update by re-ingesting the web, which means the structural source-layer work (reference databases, press archives, structured data, citation networks) carries forward into every training cycle. The work is durable because the underlying signals are real.
Insights
All insights →
AI Search Visibility reading.
ChatGPT Ads: what the early data actually shows
— ChatGPT Ads are now open to everyone. Early advertiser reports point to high click costs, no conversion tracking and thin reporting. Here is a measured read, and what it means for B2B.
What is Generative Engine Optimisation (GEO)?
— A plain-English guide to GEO. How to make a brand visible, accurate and cited inside AI answer engines like ChatGPT, Gemini and Google's AI Overviews.
What Is Answer Engine Optimisation (AEO)? A Clear Guide
— What is answer engine optimisation? A plain-English guide to AEO, how it differs from GEO and SEO, where the term came from, and what it covers. Read on.