Answer engines (AI): the guide for a small business

What an answer engine is, and what changed

For twenty years, search meant one thing: you typed something, you got a list of links, you decided where to click. The engine showed you the shelf and stepped aside.

An answer engine does something else. It reads a few sources itself, synthesises them, and gives you a written answer with two or three citations inside it. It does not show you the shelf, it brings you the book. ChatGPT, Perplexity and Gemini work this way, and so does Google now with the summaries it places above its results.

The consequence for you is simple and unpleasant: if you are not one of those two or three sources, the customer gets their answer without ever learning that you exist. They did not pass you by, they never saw you. We have described this in detail in the invisible wall between you and your customers, and where else people now look for you is covered in beyond Google: where else they search for you.

What this work is called, and how it relates to SEO

The work you do to become one of those sources is called GEO, from Generative Engine Optimization. It is not a term some agency invented so it would have something to sell: it was established in an academic paper presented at the KDD conference in 2024.

The most common mistake is to assume GEO replaces SEO. It does not replace it, it stands on top of it. Answer engines do not read the web from scratch every time somebody asks a question. They retrieve sources from a search index and then synthesise. If you are not in the index, there is nothing to cite.

The comparison of the two, where they differ in goal, measurement and competition, is laid out with a table and in plain words in what the difference between SEO and GEO actually is.

Why it recommends some businesses and ignores others

The answer is not «because they pay». There is no paid slot inside an answer, at least not today. The choice is made on how easily a clear, supported statement can be pulled out of your text and quoted without risk.

Which means that a page full of generalities, however well written, gives the engine nothing to hold on to. The mechanism behind the choice is explained in why AI recommends some businesses and ignores others, and the experience and trust criteria Google asks for separately are covered in E-E-A-T: what Google actually wants.

What you actually do, starting tomorrow

This is where the work starts, and it is less technical than you fear. The three things that change the picture first are: answer the question within the first two sentences of each page, use specific numbers instead of «a significant improvement», and say where you know your claims from.

All seven steps, with what the research says about each and three misconceptions that cost money, are in what you actually do to get recommended by answer engines. If you want the business side first, what changes in practice for a company that does this properly, see GEO: the new digital advantage.

What the engine has to be able to read on your site

There is one point where the best writing saves nothing: when the engine opens your page and finds it empty. It happens more often than you would think, because most crawlers do not run JavaScript. If your content only appears once the page is «built» in the browser, for them it does not exist.

How to check this yourself, and what to do if it is happening, is in JavaScript SEO: when the engine sees an empty page. And since many people block crawlers believing they are protecting themselves, it is worth knowing exactly whom you are blocking and what it costs: which crawler are you blocking, and what do you think you gain.

How you measure whether it is working

There is still no Search Console for answer engines. The practical method is manual and, for all its apparent naivety, it works: build a fixed set of ten to fifteen questions a prospective customer would ask, put them to the same engines every month, and record whether you are cited and in what words.

The methodology, with what is worth measuring and what means nothing, is in how to check whether AI recommends you. If you want to see the measurement on real data, we built the tool that did not exist shows how and why we ended up building it ourselves.

Credibility is not built inside your site alone

Many people overlook this. Many answer engines do not rely on your site alone, they also draw on independent sources. A flawless site with no mention anywhere else struggles to reach anyone who does not already know your name.

A listing with consistent details, reviews and presence in directories is infrastructure, not a detail. The simplest and most neglected part of it is Google Business Profile. And if you work locally, it is worth seeing what the engines say about your area today: ask them about your area.

Two examples with numbers, both ours

Because all of the above sounds easy while someone is writing it, two cases with before and after measurements. mswriter.eu from D to A shows what changes on a site that starts with none of what we described. And osyllektis.com, from 19 seconds to 4.4 covers the speed side, which is a precondition rather than decoration.

The neighbouring subjects, for when you need them

Four more articles are not on the main path, but you will want them at some point. Where the visitor goes when the article ends is about the structure of links inside your own site. Changing domain or moving a site without losing anything is the nightmare that preparation avoids. CDN: when you need one and when you do not answers a question that is usually asked the wrong way round. And when Google Ads is worth the money is here because organic presence and paid presence solve different problems.

What nobody should promise you

The mechanism by which an answer engine picks its sources is a closed box. The researchers who established the term say so themselves. So nobody can guarantee you a place in any answer, and anyone who does either does not know or knows and says it anyway.

What a serious person can promise is specific work with a measurable before and after. Nothing more, and nothing less.