Back to blog
    AI SEO8 min read2026-05-17

    How to Optimize Your Website for AI Search (ChatGPT, Perplexity, Gemini)

    More and more customers ask ChatGPT and Perplexity instead of Googling. Here are 7 concrete steps to get your website cited by AI search engines in 2026.

    How to Optimize Your Website for AI Search (ChatGPT, Perplexity, Gemini)

    Over the last 12 months something quietly happened that will reshape SEO in the next few years: millions of people stopped Googling and started asking ChatGPT instead.

    When someone asks ChatGPT "what is the best marketing agency in Osijek?", it does not return 10 blue links like Google. It returns one answer with a handful of cited sources. Whether your business is one of those sources or not decides whether you get the client. That is the new fight. And right now, Croatian businesses are barely paying attention to it.

    What is AI Search Optimization (GEO, AEO, LLM-O)?

    Three terms you will see in the literature:

    • GEO (Generative Engine Optimization): optimizing content so it becomes a source for generative AI systems (ChatGPT, Gemini, Claude).
    • AEO (Answer Engine Optimization): the broader category, includes Google AI Overviews, Bing Copilot, Perplexity, You.com.
    • LLM-O (Large Language Model Optimization): newest term, focused on structuring content that LLMs can reliably extract and cite.

    They overlap. The principle is the same: your content has to be structured, factual, and easy to cite for AI models.

    Why now is the right time for Croatian businesses

    ChatGPT is already in the top 10 most-visited websites in Croatia. Perplexity and Gemini are growing. Croatian businesses that prepare their sites for AI search now will have a 1-to-2-year head start before this becomes the standard.

    Right now there is almost zero competition in Croatian for GEO and AEO terms. That means less effort to rank, first-mover citation status for Croatian business questions, and a market not yet saturated with citation battles.

    7 steps to optimize for AI search

    1. Let AI crawlers in

    First, check your robots.txt. AI search engines have their own bots that need to be able to fetch your content:

    User-agent: GPTBot
    Allow: /
    
    User-agent: ClaudeBot
    Allow: /
    
    User-agent: PerplexityBot
    Allow: /
    
    User-agent: Google-Extended
    Allow: /
    
    User-agent: Applebot-Extended
    Allow: /
    

    If you use Cloudflare, check Security → Bots → Block AI bots. If it is enabled, everything above is overridden at the edge, even if your robots.txt allows the bots. This toggle is on by default on new Cloudflare zones. Turn it off unless you genuinely want to block AI search.

    2. Add an llms.txt

    llms.txt is a new standard for AI search engines. A Markdown file at the root of your domain describing who you are, what you offer and where the important pages live. The equivalent of robots.txt, but for content.

    See our example: opus-studio.xyz/llms.txt. It includes a short intro, a list of services with links, a portfolio with client URLs, all main pages (EN and HR) and contact details. LLMs indexing your site use this file as a condensed brief. It saves them time and significantly raises the chance you are cited accurately, with fewer hallucinations about your business.

    3. Schema markup (LocalBusiness, Service, Article)

    AI models are trained on structured data. Schema.org markup makes it easier for them to understand who you are and what you offer:

    • LocalBusiness: address, phone, hours, coordinates, reviews
    • Service + Offer: exact prices and service categories (see our pricing page, every package has Offer schema with a EUR price)
    • Organization: founder, sameAs links to social profiles
    • Article / BlogPosting: for blog content
    • BreadcrumbList: page hierarchy

    When someone asks ChatGPT "how much does a website cost in Osijek", the AI will preferentially cite a page with explicit prices in Schema markup over one that just says "contact us for a quote". Structured data turns your page into a data source, not just a document.

    4. Write content as questions and answers

    LLMs cite short, clear answers to specific questions. Structure your key pages with FAQ sections where the question and answer sit alongside structured data.

    Bad: "Our agile methods enable fast delivery at high quality." Good: "How long does a landing page take to build? A standard landing page is finished in 1 to 2 days."

    Numbers, dates, concrete answers. No marketing fluff. AI models cannot cite what they do not understand, and they do not understand general claims with no substance.

    5. Pre-render the page (SSG or SSR)

    If your site is a pure React or Vue SPA with no server-side rendering, AI bots see an empty HTML shell and cannot cite anything. More and more AI models do not execute JavaScript when collecting data. Pages that depend on client-side hydration are effectively invisible to them.

    We use vite-react-ssg which pre-renders every route to a static HTML file at build time. ChatGPT, Perplexity and Google-Extended fetch the full content on the first request, without waiting for React to hydrate. The same applies to Astro, Next.js (App Router with SSG), SvelteKit and other modern frameworks.

    6. Authority signals (About, sameAs, reviews)

    LLMs favour sources with a clear author and organization. Make sure you have:

    • An About page with concrete information about the founder and team
    • Consistent sameAs links in Organization schema (LinkedIn, Instagram, Google Business)
    • LocalBusiness schema with address, phone and geo coordinates
    • Reviews with real names (Google Business Profile, Clutch)
    • Mentions on other authoritative domains (guest posts, interviews, podcast appearances)

    The more signals that you are a real, local business, the more likely an AI is to cite you in answers about local questions.

    7. Cite concrete numbers and dates

    AI models love verifiable facts. Instead of "we build fast websites" write "our sites load in under 1.5 seconds on Slow 4G (measured with Lighthouse)". Instead of "experienced team" write "in business since 2023, 15+ projects delivered".

    That is the foundation of GEO: concreteness is citation value. AI will preferentially cite a sentence with a number over a sentence with an adjective.

    How to check if you are being cited

    The simplest test: open ChatGPT, Perplexity and Google AI Overviews. Ask:

    • "What are the best web agencies in Osijek?"
    • "How much does a website cost in Croatia in 2026?"
    • "Who is [your business name]?"

    If you are not mentioned, you know where you stand. If you are mentioned by name, great. If a link to your content is cited, even better. Perplexity shows citations as numbered references. ChatGPT (web search mode) lists sources at the bottom of the answer. Google AI Overviews shows citations as a carousel below the answer.

    Run this test monthly for 5-10 key questions in your niche. It is the AI equivalent of a GSC report.

    What NOT to do

    • Keyword stuffing. AI models detect it and downweight your authority. Write for humans, not for the bot.
    • Hidden text and cloaking. Same penalty, plus risk from classic search engines.
    • Generic corporate language. "Our expertise and synergy" is death for citability.
    • Hallucination-bait content. Do not write "the best agency in the world" if you cannot back it up. AI will either ignore you or (worse) cite you incorrectly.
    • Blocking AI bots without a reason. If you do not want to be cited, block them. If you want to be visible in AI answers, let them in.

    What now

    AI search optimization is not a replacement for classic SEO, it is its continuation. Everything good SEO does (load speed, schema, content, authority) also works for AI. The difference is concreteness: AI wants facts, not marketing.

    If you need help with the implementation, check our SEO and AI Search package or get in touch at opus-studio.xyz/contact. Our Pro tier at 100 EUR/mo includes GEO and AEO optimization, schema markup for AI models, and monthly citation monitoring across ChatGPT, Perplexity and Google AI Overviews.