Get Found in AI Answers: AEO for SMEs

The concept: write for the question, not the keyword

Stop optimizing content for a search engine that spits out ten blue links. Write for the AI answer that hands your customer a ready-made solution and names a handful of sources along the way. That is the core of AEO (Answer Engine Optimization): making sure ChatGPT, Google's AI answers, Perplexity, and similar tools use your content as evidence and cite you as the source. The mechanics differ from classic SEO. An AI answer favors content that answers a concrete question clearly, briefly, and with proof. One precise answer backed by a verifiable number beats ten vague paragraphs. Once you understand that, you stop writing for the keyword and start writing for your customer's actual question.

The essentials in 60 seconds

• AEO is not the new SEO. It is a second layer alongside it: you want to be cited, not just ranked.
• AI answers favor content that answers a concrete question clearly, briefly, and with proof.
• The first step costs nothing: collect the real questions your customers ask. Sales and support already know them.
• A citable answer: roughly 80 words max, one verifiable number, one traceable source, the direct statement right at the start.
• Structure helps the machine: a clear heading phrased as a question, a short answer underneath, then the reasoning.
• Measure monthly whether the AI names you for your target questions, find the gaps, and close them.
• Be patient: effects build up over weeks, because AI systems first need to crawl and index your content.

SEO, AEO, or both? A quick framework

Before you invest time, work out what your business actually needs. AEO does not replace SEO, it complements it. Both draw on the same good content but emphasize different things. The framework below helps you pick the right level.

• Level 1, foundation (almost everyone): good, honest answers to your customers' real questions, cleanly structured. This helps search and AI answers alike. Everyone should start here.
• Level 2, targeted citability (anyone selling something that needs explaining): you shape individual answers so the AI can pick them up easily, and you check back regularly. Worth it if customers research before buying.
• Level 3, monitoring system (anyone treating visibility as a channel): you continuously track, across several AI engines, which questions name you and where competitors lead, and you steer content accordingly. This is the setup we run ourselves.
• Rule of thumb: the more advice-heavy and explanation-dependent your offer, the more AEO pays off. For pure impulse purchases or purely local walk-in business, the lever is smaller.

Phase 1: Collect your customers' real questions

Goal: within an hour, build a list of the twenty to thirty questions your customers actually ask before buying. Not the questions you assume they should ask, the real ones. Sales and support are the best source, because they answer these questions every day.

• Ask sales and support: which three questions come up in every other conversation?
• Check your inbox and ticketing system: what wording do customers use, word for word?
• Type your topic into ChatGPT and Perplexity yourself and note which follow-up questions the AI suggests.
• Prioritize: which questions sit closest to the purchase decision? Those come first.
• Note roughly how often each question comes up per week. That becomes your baseline later.

Copy-paste prompt: extract questions from real conversations

If you have a collection of emails, tickets, or call notes, let the AI pull out the recurring questions and phrase them the way your customers would.

Here are anonymized excerpts from customer conversations and support requests: [paste text].
Extract the 20 most common questions customers ask BEFORE buying.
Phrase each question the way a customer would literally type it into ChatGPT or Google, so natural, full sentences.
Group them by buying stage (orientation, comparison, decision) and flag the five questions closest to the purchase decision.

Phase 2: Write one citable answer per question

A citable answer is built so an AI can pick it up easily. That means the core statement comes first, it is brief, and it is backed by evidence. Think of a press spokesperson delivering a clear, quotable core message instead of talking around the point.

• Use the question, word for word, as the heading (an H2, for example).
• Answer it directly and completely in the first one to two sentences, no preamble.
• Back the statement with a concrete, verifiable number or a clear example.
• Only after that comes the reasoning, context, and exceptions.
• Stay honest: label estimates as estimates and never invent sources.

Write a citable answer to this question.
Structure:
1) The question, word for word, as the heading.
2) A direct answer in 80 words maximum, the core statement in the first sentence.
3) Exactly one verifiable number or concrete example, clearly labeled as an estimate if it is one.
4) Two to three sentences of reasoning.
Tone: matter-of-fact, direct address, no hype, no dashes.
Do not invent sources or studies.
Question: [insert real question].
Background you may use: [insert your facts, numbers, specifics].

Phase 3: Make the answer technically discoverable

A good answer does not help if AI systems cannot read it cleanly. This is not about programming, it is about structure and a few machine readable signals. These three levers matter most, and you do not need technical staff for them.

• One question, one heading, one answer: avoid catch all pages where ten topics blend together. Clear separation makes each answer individually citable.
• FAQ markup (Schema.org): store the question and answer as structured data. Most website builders and CMS platforms have a feature or plugin for this.
• Make the date visible: AI answers often favor recent content. A visible last updated date helps.
• Internal linking: link related answers to each other so the AI recognizes the connection.

Generate a valid FAQPage schema in JSON-LD format according to Schema.org for the following question and answer, so I can add it to my page's source code.
Use my answer text exactly, do not shorten it and do not add anything.
Question: [question].
Answer: [your finished answer].
Explain in two sentences where on the page I should insert the code.

Phase 4: Publish and give the AI time

Unlike an ad, AEO does not work instantly. AI systems first have to crawl your page, index it, and fold it into their answers. That usually takes weeks, not hours. Make sure the crawlers can actually reach your content, then be patient.

• Check that your robots.txt does not block AI crawlers (for example GPTBot, Google Extended, PerplexityBot) if you want to be cited.
• Submit new pages for indexing in Google Search Console.
• Link new answer pages internally and, where it fits, from your newsletter or LinkedIn, so they get found faster.
• Set a reminder for four weeks out to run your first measurement. Checking earlier is rarely worth it.

Phase 5: Measure whether the AI names you

Without measurement, AEO stays a gut feeling. The simplest measurement costs nothing: ask the AI your own target questions and see who gets cited. Do it systematically and the same way each time, so you can spot changes.

• Take your top ten target questions and ask them monthly in ChatGPT, Perplexity, and Google's AI.
• For each question, note: are you named? Are you linked? Who else gets cited?
• Keep a simple table: question, date, named yes or no, competitor named.
• Where you are not named, look at the cited source and ask: what makes their answer more citable than yours?
• Sharpen the answer and measure again the following month.

I am checking my visibility in AI answers. Answer the following question the way you would for a user, and at the end explicitly list which sources or providers you named or drew on in your answer.
Question: [your target question].
Then tell me in one sentence what a source would need to offer for you to cite it preferentially on this question.

A worked example: a tax firm with a specialty

A tax advisory firm focused on doctors and healthcare professionals currently is not named by the AI for the question 'What tax advantages do self employed doctors have?' Instead, the AI cites two large portals. Here is the path to a citable answer, with example figures used as estimates:

• Starting point: for the target question, ChatGPT names two portals, the firm does not appear.
• Step 1: the firm gathers the 25 most common questions self employed doctors ask, drawn from client conversations.
• Step 2: each question gets a citable answer: the question as the heading, a direct answer in 80 words, one concrete number (labeled as an estimate), then the reasoning.
• Step 3: each answer gets FAQ markup and a visible date, and the pages are linked internally.
• Step 4: after roughly six weeks, the first measurement: for 6 of the 25 questions, the firm is now named in the AI answer.
• Result (estimate, measure it in your own pilot): zero mentions before, mentions on roughly 12 of 25 questions after three months, and noticeably more qualified first consultations from this target group.

How to implement it (two paths, no developer needed)

You do not need developers to get started. Choose based on what you already work with today.

• Path A (no tech, one afternoon): collect questions, draft an answer for each using the answer prompt, review editorially, and publish as an FAQ page or blog post in your existing CMS. A plugin often handles the FAQ markup. Ideal for proving that it works.
• Path B (ongoing, with some routine): turn collecting, writing, publishing, and measuring into a fixed monthly rhythm. A simple table tracks which questions are still open and where you are already being named. More discipline, but visibility builds continuously.
• Recommendation: run Path A first to prove it on five questions, then establish the Path B rhythm.

For advanced teams: the monitoring loop we run ourselves

Once the basics are in place, an ongoing measurement and steering system is what separates a one off effort from real, growing visibility. This is exactly the system we run for ourselves. If you have technical staff or the willingness to keep a routine, start here. The following techniques work together.

• Target question tracking across multiple engines: we maintain a fixed list of target questions (prompts) and run them automatically against several AI engines (ChatGPT, Perplexity, Google AI) on a regular basis. For each question we see whether we are named, at what share of voice, and who else gets cited.
• Finding answer gaps: where we are not named, we analyze the source that actually got cited and work out what makes its answer more citable. That gap becomes a content task.
• Writing a citable answer and remeasuring: we publish the missing answer, wait out the crawl cycle, and measure the same question again. The loop closes: measure, find the gap, write, remeasure.
• Building an entity (who you are): AI answers work with entities, meaning clearly identifiable actors. Consistent details across your website, legal notice, LinkedIn, and industry directories help the AI recognize you as a distinct, trustworthy source.
• Structured data (schema): we mark up content with FAQPage, Organization, and Article schema so machines can correctly attribute question, answer, and author.
• Providing an llms.txt file: an llms.txt file at your site root summarizes for AI systems which of your content is most important and most reliable. The standard is still young and not yet used by every engine, but it costs little and does no harm.
• Multi engine citation tracking: we log over time which engine cites us for which questions. That way we spot trends instead of drawing conclusions from a single measurement.

The self improving loop (how we run AEO for ourselves)

Target question tracking, gap analysis, and remeasurement together form a system that learns from its own results. The loop: we measure weekly which questions name us and who else gets cited. From that we work out which answer is missing next, write it to be citable, publish it, and measure again after the crawl cycle. That turns AEO from a one off project into a channel that gets a little more visible every month. That is the real reason a clean setup pays off: not the single cited answer, but the system behind it.

Tools and costs (estimates, as of 2026, check current pricing)

Content first, tools second. Getting started takes very little, the monitoring system adds a bit more. Prices are estimates and change, check the current rate.

• For writing: a good AI writing tool (for example Claude or ChatGPT) plus human editing, roughly 20 to 30 euros per user per month.
• For manual measurement: free, by asking your target questions in the AI engines yourself and logging the results in a table.
• For FAQ markup: included as a feature or plugin in most CMS platforms and website builders, often at no extra cost.
• For systematic citation tracking: specialized AI visibility tools offer automated multi engine tracking, starting around 100 euros a month depending on scope. Worth it only from Level 3 onward.
• Recommendation: start with free manual measurement. A tool only pays off once the number of target questions outgrows manual work.

A worked calculation

Example figures, replace with your own real numbers in a pilot. AEO does not calculate as cleanly as a time saving, but the lever becomes tangible.

• You publish 20 citable answers over two months.
• After three months, you are named (estimate) in AI answers for roughly 10 of those questions.
• Assume each named question brings an average of 2 qualified inquiries per month, that comes to roughly 20 inquiries a month.
• Compare that against your effort: roughly 20 answers at about 30 minutes of editing each, plus the monthly measurement.
• A single cited answer can bring more qualified inquiries than ten blog posts with no visibility. Hard to put a number on in advance, but easy to measure through incoming inquiries.

The most common mistakes

• Writing for search engines instead of people. Clear answers to real questions win, keyword stuffing does not.
• Burying the answer. The core statement belongs in the first sentence, not at the end of a long paragraph.
• Stating numbers without backing. AI answers favor what is verifiable, and honest estimates beat invented precision.
• Cramming ten topics onto one page. One question, one answer, one page is more citable.
• Writing once and never measuring. Without monthly measurement you do not know what is working or where the gaps are.
• Expecting instant results. AEO needs the crawl cycle, so weeks, not days.

Common questions

• Isn't AEO just SEO under a new name? No. SEO targets ranking positions, AEO targets being cited in the AI answer. Both draw on good content but emphasize different things.
• Does AEO hurt my regular Google ranking? No, the opposite. Well structured, clearly answered questions help both worlds.
• Do I lose visitors if the AI shows my answer directly? Sometimes someone reads only the AI answer. But being named as a source builds trust, and anyone making a real decision clicks through. Advice heavy offers benefit especially.
• How do I know if it is working? By asking your target questions in the AI engines every month and logging whether you are named. That is the most honest measurement.
• Do I need an expensive tool for this? Not to start. Collecting, writing, and measuring can all be done by hand. A tool only pays off once the number of target questions outgrows manual work.

GDPR and the EU AI Act in one paragraph

AEO works with public content on your website, so the data privacy situation is manageable. Still, pay attention: if you use customer conversations or tickets as a source for your question collection, anonymize them before feeding them into an AI tool, and use tools with a data processing agreement that do not train on your data, ideally hosted in the EU. The EU AI Act's transparency obligations mainly concern AI providers, not your content, but if you already label things cleanly and honestly, you are on the safe side here too.

When to skip this

If you do not yet have real, good content, start there. AEO makes existing good content visible, it does not replace it. The lever is also small for pure impulse purchases or a business that lives almost entirely on local walk in customers. AEO pays off where customers research before buying and ask questions, and where you can give honest, verifiable answers to those questions.

Your 30 day start

• Week 1: work with sales and support to collect the 20 most common customer questions and flag the five closest to a purchase decision.
• Week 2: write a citable answer for each of those five questions (using the answer prompt), review editorially, and publish with FAQ markup.
• Week 3: answer and publish the next ten questions, check robots.txt and indexing.
• Week 4: prepare your first measurement (set up a target question table) and take the first measurement in four weeks, then refine monthly.

Where this comes from

This playbook is not built on theory, it is built on a system we run ourselves. We measure weekly, across several AI engines, which questions name us and who else gets cited, find answer gaps, write them to be citable, and measure again. What is written here is what we built, measured, and refined on our own visibility. We do not just talk about AI visibility, we build it for ourselves.