Your buyers stopped searching.
They started asking.

Generative engine optimization is the practice of measuring and improving how often, how prominently, and how accurately an AI assistant names your brand when a buyer asks it a question. It targets the synthesized answer rather than the ranked list of links. You measure it by sampling real buyer questions across AI engines and recording what comes back.

In practice, nothing. Generative engine optimization, answer engine optimization, and AI optimization all describe optimizing for AI-generated answers. The terms came from different vendors at roughly the same time. We use GEO because it is what our buyers and the analysts we work with say, but the work is identical.

Eight: ChatGPT, Google Gemini, Google AI Mode, Google AI Overviews, Perplexity, Claude, Grok, and Microsoft Copilot. Which ones we prioritize depends on your category. B2B software behaves very differently from a manufacturer selling through distributors, and the engine mix reflects that.

By sampling, not by rank tracking. ORCA puts a fixed set of buyer questions to each engine on a schedule, with repetitions. It records whether your brand was mentioned, where in the answer, with what sentiment, alongside which competitors, and citing which sources.

Generated answers vary between runs, so the result is a rate with a confidence interval. We do not draw a trend line until there are enough snapshots to draw an honest one.

You cannot edit the model. You can change the evidence it retrieves and the consensus it reads. That means three things: publish content that makes a specific claim clearly and verifiably, mark up your entity so the claim is machine-readable, and earn mentions on the domains models already cite in your category. ORCA identifies those domains from the citations in your own measurement data.

The honest framing: GEO is influence over inputs, not control over outputs. Anyone promising control is selling something else.

Truth corrections often move within one or two snapshot cycles, because the fix is a specific verifiable claim published on your own site. Topic gaps, where a competitor owns a concept outright, behave more like traditional SEO and usually take one to two quarters.

Measurement starts on day one. That is the point. You get a defensible baseline before any work is committed.

Yes, more than the discourse suggests. Every engine on this list either crawls the open web or reads a search index to ground its answers. Pages that are slow, uncrawlable, thin, or unlinked do not get retrieved, and content that is never retrieved cannot ground an answer. GEO adds a measurement layer and a claim layer on top of technical SEO and content. It does not replace them.

It is the share of a model's factual claims about your brand that match your own verified source evidence. ORCA computes it per concept, and separates the two ways a model gets you wrong. Contradiction is when it asserts something false. Omission is when it leaves out a capability that would have changed the recommendation.

A brand can post strong visibility and a poor truth alignment score. That combination is common, and it is usually the more expensive problem.

Pricing depends on personas, question-set size, engine coverage, and cadence. A focused single-persona campaign is a very different scope from an eight-engine, multi-market enterprise build.

The starting point is the same either way. We run a free AI visibility audit against your brand and the competitors you name, so the conversation starts with your data instead of our pitch.