» What is actually changing
Search is shifting. Not overnight, but noticeably. Someone who once typed „accountant for freelancers Dublin“ into Google got a list and clicked through it. Someone asking the same question today of ChatGPT, Perplexity or Google’s AI overview gets a finished paragraph – with two or three names and a few source links off to the side.
For you as a solo professional that means two things. First, there is a new layer between you and your client. The AI reads your website and tells somebody else about it. Second, the competition narrows. In a list of ten results you might still be visible at number six. In an AI answer with three names you are either in or you are not.
That sounds more alarming than it is. The criteria are neither secret nor magical. They are largely what has always marked out good websites – only stricter.
» SEO, AEO, GEO: the terms sorted out
Three abbreviations, one goal: being found. The difference is where you appear.
- SEO (Search Engine Optimization) means ranking high in the classic list of results. Someone searches, sees your title, clicks. That is the foundation and it stays the foundation.
- AEO (Answer Engine Optimization) means your content is served as the answer. Not as a link to click, but as text that answers the question directly – in AI overviews, featured snippets, voice assistants.
- GEO (Generative Engine Optimization) means you appear inside an AI-generated text. As a cited source or – the real jackpot – as a named recommendation: „For small service businesses, X is often recommended.“
And now the sober part: the boundaries between AEO and GEO are blurry, and the industry throws the terms around freely. There are no separate disciplines with their own toolboxes. There is one way of working that serves all three. So do not be impressed by anyone selling you „GEO packages“ as though it were something entirely new.
» How AI search picks its sources
Nobody outside the providers knows the exact mechanisms, and anyone claiming otherwise is guessing. What can be observed, though, is that certain kinds of pages get cited noticeably often. Five patterns repeat.
Clear structure. AI systems break your page into sections and judge them individually. A section that makes sense even torn out of context is usable material. A paragraph that builds on three previous paragraphs is not.
Unambiguous answers to real questions. „What does a website cost for a solo professional?“ is a real question. A page that answers it with a number and an explanation is citable. A page that says „that depends entirely on your individual requirements“ is not – there is no information in it.
Verifiable facts. Specific figures, years, named sources, clear statements about service, audience and location. Vague marketing language cannot be quoted, because it claims nothing anyone could pass on.
Mentions elsewhere. What appears on several independent sites reads as confirmed. Directories, interviews, guest articles, partner pages, reviews – these are evidence that you really exist and what you stand for. This part you cannot handle on your own website alone.
Technical readability. And here it gets uncomfortably specific: the crawlers of the AI providers generally do not run JavaScript. They fetch the HTML and read what is in it. If your text is only assembled in the browser from components – as is common with many modern systems and page builders – the AI sees an effectively empty page. Not judged badly. Not seen at all.
» Seven things you can actually do
None of this needs a degree. Most of it is writing craft; a smaller part is technical.
- 1. Answer questions literally. Take the questions clients ask you in a first call and turn them into headings. In their exact wording. People type whole sentences to an AI, not keywords – your text should resemble those sentences.
- 2. Put the answer first. First sentence of a section: the answer. Then explanation, example, caveat. This is the single most effective lever there is – and the one almost nobody applies consistently.
- 3. Add structured data. A little invisible code in the source (Article or BlogPosting for articles, FAQPage for questions, LocalBusiness for your company). It tells machines unambiguously: this is the author, this is the date, this is the question, this is the answer.
- 4. Keep a clear heading hierarchy. One H1 per page, logically ordered H2s below it. Not headings that merely sound nice („where it all began“), but ones that name the content.
- 5. Use summary boxes. Three or four key statements at the top of a longer piece. A service for your readers, and a ready-made distillation of your article for machines.
- 6. Name numbers and sources. „Fast“ is nothing. „Under one second loading time“ is something. If you use a statistic, say where it comes from. Backed statements get picked up more often than asserted ones.
- 7. Deliver content statically and fast. The text has to be in the HTML the server sends – no detour through scripts. How much speed decides visibility applies to crawlers exactly as it does to people.
One small addition: llms.txt. A plain text file in the root of your website summarising in bullet points who you are, what you offer and where the important content sits. A young standard, not yet universally supported – but it is written in twenty minutes and cannot hurt.
» What does not work
Just as important as the list above: the things that only cost you time.
Keyword stuffing. Cramming the same term in thirty times stopped working for Google long ago. Language models understand meaning, not frequency – a text full of repetition simply reads to them as bad writing.
AI text with no substance of your own. If you have a language model write thirty articles made of generally known information, what you get is exactly what the AI already knows. There is no reason to name you as a source. What makes you citable is what only you have: your numbers, your experience, your examples, your opinion.
Tricks. Hidden text, fake reviews, invented statistics, pages full of unfounded claims. It gets found out, and if it does, the damage to your reputation costs more than a mention was ever worth.
» The honest reality check
There is no guarantee. Nobody can assure you that ChatGPT will recommend you, and there is no position one to book. AI answers are generated fresh for every query and come out differently depending on wording, timing and model. Two people can ask the same question and be given different names.
Measuring is awkward too. There is no dashboard telling you how often you were cited. What you can do: regularly ask your own typical client questions in ChatGPT and Perplexity and see who gets named. That is unscientific, but it is more honest than any promise.
What is fairly certainly true: someone who writes clearly, specifically and in an answering style, on a technically clean site, gets cited more often than someone who does not. That is not a guarantee, it is a probability – and probabilities are what visibility has always run on.
» An example: this website
So that this does not stay theoretical: the page you are reading is built to exactly these rules. It is static HTML – every piece of text sits in the source, no script has to assemble anything. Every article has a summary box at the top, headings that name the content, and an FAQ block at the end backed by FAQPage markup. Author and date are in the structured data. An llms.txt sits in the root.
None of that is spectacular, and none of it was expensive. It is simply the way I have chosen to build – and the foundation of every website I build for clients. What I recommend, I do myself.
» What this means for you
Start small. Take the three questions your clients ask most often and answer them on your website – the question as the heading, the answer in the first sentence, the explanation after that. Check whether your text is in the source without JavaScript. Add an FAQ block with structured data. That is a few hours of work, not a repositioning of your business.
And then what has always applied to visibility applies here too: keep going, write honestly, get specific. The machines have changed. What makes content good has not, at its core.
Common questions
What is AEO?
AEO stands for Answer Engine Optimization. It means writing content so that it can be served directly as the answer to a specific question – in AI overviews, voice assistants or AI search. The most important lever: the answer sits in the first sentence of a section, not three paragraphs in.
What is the difference between AEO and GEO?
AEO aims at your content being served as a finished answer. GEO aims at you appearing inside an AI-generated text – as a cited source or a named recommendation. In practice the two overlap heavily: writing clearly, specifically and verifiably works on both at once.
How do I get found by ChatGPT?
By making your website technically readable and answering real questions unambiguously. AI crawlers generally do not run JavaScript, so your text has to sit in the HTML itself. Add clear headings, FAQ blocks with structured data, specific details about service, audience and price, and mentions on other sites. No guarantees, but the odds rise considerably.
Can I buy position one in AI answers?
No. There is no bookable position and no agency that can guarantee a mention. Answers are generated fresh per query and vary by wording, timing and model. The only legitimate route is indirect: cleanly structured, honest, fast-loading content that works better as a source than your competitors’ does.