Generative Engine Optimisation
Generative engine optimisation, or GEO, is entity-level work. It makes the facts about an organisation consistent and corroborated everywhere they appear, so that a model can name it with confidence. It is the slowest of the three disciplines because it depends on sources you do not own.
Being the source a model reaches for
When an assistant is asked to recommend a provider, it is not ranking pages. It is drawing on what it holds about organisations as entities: who they are, what they do, where they operate, and whether the sources describing them agree. Contradiction produces hedging, and hedging produces omission.
GEO is therefore mostly an exercise in agreement. The same company name, the same address, the same service descriptions and the same claims across your own site, directories, professional profiles, industry listings and any third-party coverage. Where those disagree, the model has no confident answer to give.
This is the slowest of the three disciplines, and the one where honesty about limits matters most. OMAV states plainly what is influenceable, what is measurable, and what nobody can promise, before any budget is committed.
What GEO covers
Six workstreams, from audit to dated measurement.
Entity audit
Every place your organisation is described — your own site, directories, professional profiles, industry listings, review platforms and press coverage — collected and compared.
Contradiction list
Each disagreement recorded explicitly: name variants, wrong addresses, stale service descriptions, incorrect founding details and superseded claims, with an owner assigned to each correction.
Organisation structured data
Schema describing the organisation, its people, its services and its locations, consistent with the corrected facts and with each other.
Corroboration plan
A route to being described accurately by sources a model already trusts, through material worth referencing rather than through purchased placement, which carries risk and decays.
Answer baseline
A dated record of how the assistants your buyers use answer your priority questions: which providers are named, in what order, and whether the facts stated about you are correct.
Re-measurement
The same questions re-run on a stated cadence and compared against the baseline, so that change is evidenced rather than asserted.
How a GEO engagement runs
Baseline
Priority questions defined with you, then asked across the relevant assistants and recorded with dates, competitors named and factual errors captured.
Audit the entity
Every source describing your organisation collected, compared, and every contradiction listed with an owner.
Correct and describe
Facts corrected at source where you control it, requested where you do not, and structured data aligned to the corrected position.
Build corroboration
Material produced that trusted sources have reason to reference, and relationships pursued where a listing or profile carries weight.
Re-measure
The baseline questions re-run on the agreed cadence, with movement, stasis and remaining factual errors reported plainly.
GEO next to the disciplines beneath it
Why GEO comes last and what it depends on.
| Dimension | AIO | AEO | GEO |
|---|---|---|---|
| Unit of work | The template and response | The passage | The entity and the source |
| Question answered | Can a machine read it? | Can a machine quote it? | Can a machine name you? |
| You control | Nearly all of it | Nearly all of it | Some of it |
| Verified by | Server logs | Extraction testing | Answer and citation tracking |
| Time to effect | 4 to 10 weeks | Weeks to 2 months | 3 to 9 months |
| Depends on | Nothing above it | AIO | AIO and AEO |
What nobody can promise
No one controls whether an AI system cites you. Retrieval and attribution are decided inside systems that are not documented publicly and change without notice. Any agency offering guaranteed inclusion is either misunderstanding the mechanism or misrepresenting it, and the distinction does not help you.
GEO also cannot manufacture standing. Where an organisation is genuinely new, small or absent from the sources its market pays attention to, consistency work makes it describable but not yet notable. The remedy is doing things worth describing, which is slower than any optimisation programme and is the actual constraint more often than buyers expect.
- Contradiction produces hedging, and hedging produces omission. Agreement is the core of the work.
- Citation cannot be guaranteed by anyone. Visibility can be measured against a dated baseline.
- GEO depends on AIO and AEO being in place; run out of order it wastes money.
- Where an organisation is genuinely not notable yet, consistency work has a ceiling.
Questions buyers ask about this
Can you guarantee we appear in ChatGPT answers?
No, and nobody can. Retrieval and attribution are decided inside systems that are not publicly documented and change without notice. What we can do is make your facts consistent, machine-readable and corroborated, and measure how assistants describe you over time against a dated baseline. Anyone guaranteeing inclusion is selling something they do not control.
How is GEO different from SEO?
SEO works at the level of the page and aims at ranking and clicks. GEO works at the level of the entity and aims at being the source a model reaches for when it composes an answer. They overlap in that both reward being crawlable and credible, and they diverge in unit of work, lever and measurement.
How do you measure progress?
A dated baseline of priority questions across the assistants your buyers actually use, recording which providers are named, in what order, and whether the facts stated about you are correct. The same questions are re-run on an agreed cadence and compared. Movement is reported with the dates attached; so is the absence of movement.
What are the most common entity problems?
Name variants across profiles, an old address that persists in directories, service descriptions that describe what the company did three years ago, and inconsistent founding or registration details. Individually each looks trivial. Collectively they are why a model hedges instead of naming you.
How long does GEO take?
Three to nine months for meaningful change, and longer where corroboration is the constraint. Corrections you own can be made in weeks. Sources you do not own move on their own schedule, and pretending otherwise sets an expectation that will not be met.
Marked up as FAQPage structured data, matching the visible text exactly.
Still not sure this is the right service?
Answer four questions and we will tell you which one fits — or that none of them do.
Tell us what you are trying to fix.
A short conversation about the objective, the constraints and the timing. If we are not the right fit, we will say so.