What is answer engine optimization (AEO)?

An answer engine does not rank ten documents and hand the choice to a human. It retrieves a handful of passages, synthesizes them into one response, and cites a small number of sources. Typically three to eight sources make it into a given answer, which means the practical competition set is far narrower than a traditional top-ten result page.

That changes what optimization means. Retrieval favors passages that answer the question in the first sentence, use consistent entity naming, and carry supporting specifics the model can restate without hedging. A 2,000 word essay that buries the answer in paragraph nine is less citable than a 400 word section that opens with a definition.

• Lead every section with a one-sentence direct answer.
• Use consistent entity names, never pronouns, for your brand and product.
• Attach concrete numbers, ranges, or named methods the model can quote.
• Keep the supporting evidence on the same page as the claim.

The measurement unit in search is the keyword. In AEO it is the prompt. You define a battery of buyer prompts your ICP actually types, run it on a schedule across every engine, and record whether your brand appears, where in the answer it appears, and how it is characterized.

Share of voice is the headline number: prompts where you were mentioned, divided by prompts run, per engine. Prominence weights a first-sentence mention above a footnote citation. Sentiment catches the case where you are mentioned as the expensive option, which looks like a win in a raw count and behaves like a loss in pipeline.

• Prompt battery: 50 to 300 stable prompts covering problem, category, comparison, and brand intent.
• Share of voice: mentions ÷ runs, per engine, per week.
• Prominence: position and citation type within the generated answer.
• Sentiment: how the answer frames you against the alternatives.

In a link-based world, a buyer saw ten options and picked. In an answer-based world, the buyer sees three, and the other seven never enter consideration. Being absent from the generated answer is functionally being absent from the market for that prompt.

The practical consequence for revenue teams is that top-of-funnel is shrinking while intent quality is rising. Traffic counts fall, and the visits that remain arrive further along the buying process. Judging AI search by sessions alone will make a channel that is working look like a channel that is dying.

The sequencing matters more than the tactics. Teams that start with content production before baselining have no way to prove which change moved anything. Teams that baseline first can attribute each content change to a measurable share-of-voice delta within two to four weeks.

• Week 1: build the prompt battery and record a baseline across engines.
• Weeks 2 to 4: fix extraction on your highest-intent pages, comparison pages first.
• Weeks 4 to 8: publish the missing answer pages your battery reveals you are absent from.
• Ongoing: join mention windows to CRM first-touch and report AI-influenced pipeline in dollars.