Search used to end in a list of ten blue links. Now it often ends in an answer. Ask ChatGPT which domain registrar to pick, or ask Google how to write a collection page, and you get a few paragraphs with a handful of sources underneath.
For brands, that changes the job of content. A page still has to rank, but it also has to be the page an AI tool chooses to quote. I spend a large part of my week editing content for exactly this, so here is what I have learned about writing for AI search, also called generative engine optimization, or GEO.
How AI search picks what to cite
AI search tools do not read the web the way a person does. They break a page into passages, judge whether each passage answers a question, and pull the clearest ones into a reply. Most of them also lean on signals that search engines have used for years: whether the site is trusted, whether other sites link to it, and whether the page ranks for the topic.
That gives us three practical rules:
- A passage has to make sense on its own. If a paragraph only works after you have read the three paragraphs above it, it is hard to quote.
- The answer has to be near the question. A heading that asks "How long does a domain transfer take?" followed by two paragraphs of history before the answer gives the tool nothing to lift.
- The page has to deserve trust. Clear authorship, accurate facts and a site that already ranks all make citation more likely.
None of this is a trick. It is the same thing a good editor asks for: say the useful thing early and say it clearly.
Write answers, not introductions
The habit I cut most often from AI drafts, and from human drafts, is the slow opener. "In today's fast-paced digital landscape" tells the reader nothing and gives an AI tool nothing to cite.
Compare these two openings for a section titled "How much does a domain cost?":
Domain pricing can vary depending on many factors, and it is important to understand them before making a decision.
Most .com domains cost between $10 and $20 a year to register. The renewal price is what to watch: some registrars charge double after the first year.
The second one can be lifted straight into an answer. It also serves the human reader better, which is the point. Good GEO and good writing pull in the same direction.
Structure that helps machines and people
Structure is where most content can improve quickly. These are the changes I make on almost every page I edit:
- Headings that match real questions. Write headings the way customers phrase their searches. "Is a .co domain worth it?" beats "Domain extension considerations".
- One idea per section. When a section tries to cover three things, the passage an AI tool lifts will be muddled.
- Short definition first, detail second. Start a section with a one or two sentence answer, then expand.
- Lists and tables for comparisons. If you are comparing products, plans or options, a table is easier to cite than a paragraph.
- An FAQ where it earns its place. A short set of questions at the end of a page, with direct answers, gives both Google and AI tools clean passages. Mark it up with FAQ schema so search engines can read it.
Facts, sources and the trust problem
AI tools are cautious about repeating claims they cannot support. Content that is vague, padded or wrong is less likely to be cited, and when it is cited, it can damage trust in your brand.
This is where I lean on my science background. Every statistic gets a source. Every product claim is checked against the product. If a claim cannot be confirmed, it comes out. That sounds strict, but it makes the page stronger for readers and for AI search.
A few habits that help:
- Name your sources in the text, not only in a link.
- Add a real author with real experience to each article, and an author page that explains who they are.
- Date your content and update it when facts change.
- Avoid stating opinions as facts. If something is your view, say so.
Your product belongs in the answer
GEO is not only about being cited. It is about being cited in a way that helps your business. That means your product should appear where it genuinely solves the reader's problem.
If you sell a domain registrar, the useful answer to "how do I transfer a domain?" includes the steps on your platform, with a screenshot. If you sell skincare, the collection page that answers "which moisturiser for oily skin?" names the products and explains why. Product-led content gives AI tools a reason to mention you, not just your category.
A simple way to start
You do not need to rewrite your whole site. Start with the pages that already bring in traffic or sales:
- List the questions. Write down ten questions your customers ask before they buy.
- Test them. Ask each one in ChatGPT, Gemini and Google. Note which sources appear and whether you are one of them.
- Edit the pages that should answer them. Move the answer to the top of the section, tighten the headings, check every fact and add an FAQ where it fits.
- Repeat monthly. Run the same questions again and track what changes.
That process is most of what I do for clients editing AI-generated content: take a draft that sounds fine and turn it into something accurate, on-brand and easy to cite.
The short version
AI search rewards the same things readers always wanted: clear answers, honest facts and a page that respects their time. Write for the person asking the question, structure the page so each answer can stand on its own, and make sure your product shows up where it helps.
If you want a second pair of eyes on your content, send me a brief.