Answer engine optimisation: what it is and how to do it

An answer engine gives you an answer instead of a list of places where the answer might be. Google's AI overviews, ChatGPT with search, Perplexity, Claude and Gemini all do it. Answer engine optimisation is the work of being inside that answer.

The uncomfortable part: the traffic you used to get from being ranked fifth is largely gone in these formats. There is no fifth place inside a paragraph.

What changes and what does not

What does not change. Authority still matters. Being covered by other sites still matters. Clear writing still matters. The engines read the web, and the web they read is broadly the one search engines rank.

What changes. Three things, and they are significant.

1. The unit of success is a mention, not a position. You are named or you are not. There is no gradual slide down page one.
2. Your own site is often not the source. Answers lean on lists, directories and review sites more than on brand websites, because those read as neutral.
3. Results are unstable. Ask the same question twice and you can get different companies. This breaks every reporting habit built around rank tracking.

The four things worth doing

Be present in the sources that get read

Run your buying question through an assistant and read the citations. Those sources are your target list. Some will be directories you can join today, some are articles whose author you can email, some are review platforms where you need a profile.

This is the highest return work in AEO and almost nobody does it systematically, because it looks like admin rather than marketing.

Write facts that can be lifted

An answer engine is a summariser. Give it summarisable material.

On your key pages, state plainly: what you do, who you serve, where you operate, what it costs, how long it takes, and what is genuinely different about you. One sentence each, in text.

A quick test: copy your homepage text into a document and ask a colleague to answer five factual questions about your business using only that text. If they cannot, neither can a model.

Say the same thing everywhere

Your website, your Google Business Profile, your LinkedIn, every directory. Same description, same categories, same service list, same phone number. Contradictions are a reason to drop you.

This is dull and it works. Most businesses have three different descriptions of themselves in circulation and have never noticed.

Earn reviews and mention them factually

Rating and review count show up in answers constantly, especially for local and service businesses. Say the number on your own site too, in text, so it can be lifted.

What AEO is not

It is not a technical checklist you complete once. It is not schema markup alone, although schema helps. It is not a package you buy that guarantees placement, because no vendor controls an organic answer. And it is not something you can assess from a single question on a single model on a single day.

Measuring it properly

The method is the same one that works for GEO, and it is worth writing down because almost everyone skips it.

Pick the questions. Ten to twenty that a buyer would type, none of which contain your brand name. If they contain your brand name you are measuring whether the model knows you exist, not whether it recommends you.

Ask across models. ChatGPT, Claude, Gemini and Perplexity give meaningfully different answers. Checking one tells you about one.

Record two things per answer. Which businesses were named, and which sources were cited. The second list is your work queue.

Repeat weekly and compare against competitors. Your share of answers alone means nothing. Your share next to three rivals means everything.

A worked example from our own category: across twenty buyer questions in Australian SEO services, one agency is named in 23% of answers, a second in 15%, a third in 3%, and a fourth in 0%. The fourth has a perfectly good website. It is absent from the four directories the assistants kept citing. That is the entire explanation, and it is fixable in a fortnight.

Where to start on Monday

1. Write your twenty questions.
2. Ask them across the four assistants and record names and sources. Two hours.
3. List every source cited more than once. That is your target list.
4. Fix the facts on your own pages while you wait for listings to take effect.
5. Re run in a week.

Figo automates the asking and the recording and keeps the history, but nothing above requires a tool. It requires somebody to do it on a schedule, which is the actual scarce thing.