The GAO method: five stages, one question at a time.
Generative Engine Optimization (GEO) aims to make information about brands and products easier to understand, verify and cite in generative search and AI assistants. Generative Authority Optimization (GAO) is Creaitivo's method for applying GEO: choose questions, record a baseline, prepare a brief, publish and repeat the measurement.
SEO worked to win a click. GAO works to earn the recommendation. A mention means the name appears. A citation points to a source. A source check determines whether it supports the claim. Keeping these distinct helps identify information that is missing, ambiguous, or unsupported. Repeating the measurement under comparable conditions shows what changed in the sample, not what caused the change.
The cycle restarts
The same question: selected in Discovery → turned into a baseline in Measure → translated into a brief → compared again in Re-Measure. One question, one cycle, comparable evidence.
01 · Question selection
Discovery: where the cycle begins
Multiple AI systems · the starting point
The set is built with the client to cover the intents, market, language and constraints that matter in a decision. The output is a list of questions where the brand is absent, ambiguous, or replaced by a competitor in the collected observations. Each can become a case to verify through the full cycle.
Output: a prioritized list of decision questions where the brand has an observed gap. Each one enters the Measure → Brief → Re-Measure cycle.
02 · Documented measurement
Measure: the baseline
Consistent criteria · dated snapshot
Discovery may observe up to six AI systems. The baseline instead uses a fixed set of three AI assistants, ChatGPT, Gemini, and Claude, with documented conditions and search modes. To compare answers, I record presence, position, tone and competitors, distinguishing a cited source from one that supports the claim. The result is a dated snapshot of the sample, not an absolute value.
This baseline is the starting point. Brief, publishing, and Re-Measure test whether observed frequencies and patterns change consistently.
03 · Brief for the team
Brief: define the intervention
Structured analysis · intent + entities + sources
The same question from Discovery is analyzed by combining intent, entities, sources and observed signals for that topic. The output is a structured editorial brief with the instructions the team needs to create clearer, verifiable content that addresses the question.
Output: a ready-to-send brief. Not a suggestion list. A complete document the team executes immediately, built around the exact gap found in the earlier stages.
04 · Editorial execution
Publish: the distribution
Owned site · Industry media · PR · Third-party sources
The editorial team develops the source content and distributes it through appropriate channels: the owned site, relevant industry media, PR when there is a genuine news angle, authoritative directories, and pertinent third-party sources. The goal is to make the brand's position on the topic clear, consistent and verifiable.
Consistent information across your site and independent sources reduces ambiguity. The next comparison checks what changes, without promising citations.
05 · +90 days
Re-Measure: the comparison
+90 days · change against the baseline
I repeat the same question, keeping AI systems, search modes, conditions and assessment criteria as comparable as possible. I compare frequencies and answers with the baseline. Sources and answers can change without any intervention: re-measurement shows what changed in the sample, not why.
When frequencies and patterns change consistently, you have a signal worth investigating. When they do not, the cycle restarts with a different intervention.
Want to know whether this method fits your case?
Let's discuss your goalThe method is public. The tool is called SeenElse.
The five stages above are the method. SeenElse is not AI writing software: it is the proprietary tool I built on this method to work out what an AI, or an agent deciding on a customer's behalf, needs in order to trust a brand.
Instead of guessing what to publish, it treats brand authority as an engineering problem, with five functions working inside those stages. It is not a product for sale: I run it during the engagement, alongside your team.
The output is not "more content". It is verifiable, machine-readable information designed to make a company legible, credible and selectable. SeenElse measures and verifies: it does not guarantee citations.
Two services use the GAO method.
What problem do you want to solve first?
Tell me what you want to achieve and what is getting in the way. I'll reply personally within 48 hours to suggest where to start and what information we need.
Topic: Brand visibility
- 01A first reading of your case
- 02The recommended path and why
- 03A first step and the information needed
No automated replies: I read every request myself.