TL;DR, the core in 5 points
• AI Overviews cite the first 1–2 paragraphs, put the complete answer there, not the introduction.
• FAQPage Schema is the most-cited structure: 5–8 real customer questions per page is the minimum.
• Entity relations (organisation, location, products) explicitly marked with Schema.org significantly increase citation probability.
• Original data, even a small in-house benchmark, gives you primary source status that AI engines actively seek out.
• Measure via Bing Webmaster + manually in ChatGPT/Perplexity, not just Google Search Console.
Step 1: Answer-first writing
AI Overviews scrape the first 1–2 paragraphs for the answer. Put the complete answer there, not the introduction. What, why, how, in 60 words.
Test this yourself: copy your first paragraph and paste it as standalone text. Does it answer the question your page title asks? If you need 'read more' to answer the question, rewrite the opening. A practical format: start with the definition (1 sentence), give the reason (1 sentence), give the action (1–2 sentences). That produces a paragraph of 50–70 words that is directly citable. Google's own Quality Rater Guidelines name 'direct answers' as a positive signal, not just for AI Overviews but also for featured snippets.
Expected impact: 20–35% higher chance of AI Overviews citation with consistent answer-first writing. Time investment: 30–60 min per page for restructuring.
Step 2: FAQ blocks with Schema
Concrete Q&As with FAQPage Schema.org. It's the most-cited structure in AI Overviews. Pick 5–8 real customer questions per page.
Get customer questions from three sources: Google Search Console (what queries bring people to your page?), your sales conversations (what questions do you answer every week?), and Reddit/forums in your niche (what do people ask anonymously?). Write answers of 80–120 words per question, too short gives too little context, too long loses the conciseness AI engines value. Always use the exact question phrasing as H3. AI models match literally on question text.
Expected impact: FAQ schema raises rich snippet display for 60–70% of implemented pages. Time investment: 2–3 hours per page for FAQ setup + schema implementation.
Step 3: Make entity relations visible
AI models understand entities (brands, places, people). Mark up your organisation, products and location with Schema.org. Connect entities in text ('Organiq Grow from Tilburg helps SMEs...').
Entity SEO goes beyond schema: it's about how your brand is connected to other known entities in Google's and AI models' knowledge graph. Mention concrete partnerships, trade organisations, certifications and locations. Write 'marketing agency in Tilburg, North Brabant' rather than just 'marketing agency'. The more specific the entity connections, the smaller the chance an AI model confuses you with a competitor or a generic category.
Expected impact: stronger entity profile raises citation probability in ChatGPT and Perplexity by an estimated 15–25%. Time investment: 2–4 hours for schema audit + rewriting entity descriptions.
Step 4: Citable data + sources
Original numbers, small studies, your own benchmarks. AI Overviews prefer citing primary sources. One original data point = gold.
You don't need to be a scientist to produce citable data. Analyse your own client results (anonymised), run a mini-survey among 20–50 business owners in your niche, or combine public data in a new way. Publish this as a clearly labelled study ('Analysis of 47 SME websites, 2026'). Put the key finding in the first paragraph. AI engines actively look for sources with original data, it raises your authority score in the information graph.
Expected impact: pages with original data are cited 3–5× more often in AI Overviews than pages without. Time investment: 4–8 hours for mini-research + publication.
Step 5: llms.txt + clean HTML
Place an llms.txt in root with your most important pages + short description. Make sure content isn't hidden behind JavaScript (server-rendered).
llms.txt is an open standard (llmstxt.org) that tells AI crawlers which pages of your site are most relevant. Think of it as robots.txt but for LLM indexation. List your pillar pages, services and FAQs with a 1–2 sentence description per URL. Server-rendered content (SSR) is critical: JavaScript-rendered text is not picked up by many AI crawlers. Next.js with server-side rendering has a structural advantage over WordPress with heavy JS builders here.
Expected impact: llms.txt implementation increases indexation by AI crawlers by an estimated 30–40%. Time investment: 1–2 hours to create and publish llms.txt.
Step 6: Brand mentions off-site
AI models weigh external mentions. Independent mentions in trade media, podcasts, marketplaces, build 'model memory'.
Brand mentions work differently from backlinks: it's not primarily about the click-through but about the association. Get mentioned in a trade article, an industry podcast or a comparison site you don't control yourself. Guest contributions in relevant media count. Every platform that links your name to expertise on a specific topic reinforces the language model profile AI engines build about you.
Expected impact: 5–10 external mentions in relevant sources build visible AI model memory within 3–6 months. Time investment: 2–4 hours per guest contribution, once a month as a rhythm.
Step 7: Measuring via Bing Webmaster + Perplexity
Google Search Console still shows limited AI Overviews impressions. Combine with Bing Webmaster (shows Copilot mentions) and manually search your brand in ChatGPT/Perplexity.
Build a simple measurement ritual: weekly, search 5–10 terms you want to rank for in ChatGPT, Perplexity and Google. Note whether you're cited and which page. Bing Webmaster Tools shows impressions from Microsoft Copilot, valuable because Copilot is growing to 150M+ users. Perplexity always shows source citations, if you're not there, that is measurable proof your GEO work is not complete.
Expected impact: systematic measurement reveals which pages are already cited so you can reinforce them. Time investment: 30–45 min per week for manual check + logging.
Tools we recommend
Bing Webmaster Tools (free): shows Copilot impressions and AI-related referrals. Connect alongside Google Search Console.
Perplexity.ai (free basic version): manually test whether your brand and pages are cited for relevant queries.
Schema.org Validator / Google Rich Results Test (free): verify that your FAQPage and Organization schema are correctly implemented.
Surfer SEO (from €79/mo): analyses answer-first structure and semantic completeness versus competitors in the SERP.
llmstxt.org generator (free): quickly build a correct llms.txt from your sitemap.
ChatGPT Plus (€20/mo): run brand monitoring queries and test how AI models describe your brand.
What changed in 2026
AI Overviews are now active for 40–60% of all searches in most markets, from mainly English queries in 2024 to broad adoption across languages in 2026.
Perplexity grew to 100M+ active users: for B2B audiences it has become a primary research tool, comparable to Google for specific questions.
OpenAI SearchGPT is integrated into ChatGPT, live web browsing with source citations makes GEO now as relevant as classic SEO for reaching B2B decision-makers.
Google has rolled out 'AI Mode' in multiple markets: a fully AI-driven SERP without blue links for a portion of queries. GEO is no longer optional.
Want to appear in AI Overviews?
Request a GEO audit →