GEO: How to become visible in AI search

What is GEO and why is it becoming so important now?

Around 50% of Google searches now include AI-generated summaries, and this share is expected to rise above 75% by 2028 (McKinsey, 2025). Generative Engine Optimization, or GEO, encompasses all measures that ensure AI models find, understand, and cite your content.

The difference from traditional SEO? SEO optimizes for the Google index and the ten blue links. GEO optimizes for being cited as a source in answers from ChatGPT, Perplexity, Google AI Overviews, and other AI systems. Both complement each other, but the mechanisms differ significantly.

The economic significance is enormous. According to McKinsey, over 750 billion US dollars in revenue will be generated through AI-powered search by 2028. Yet only 16% of brands systematically track their AI search performance (McKinsey CMO Survey, September 2025). The remaining 84% are flying blind. Those who act now secure a lead that latecomers will struggle to close.

How dramatically is AI search changing user behavior?

The shift is already measurable and dramatic. A Seer Interactive study of 3,119 informational search queries from June 2024 to September 2025 shows: organic click-through rate dropped by 61%, from 1.76% to 0.61%, when AI Overviews are displayed (Seer Interactive, 2025). Paid ads lose 68% CTR. Even for queries without AI Overviews, organic CTR fell by 41%.

At the same time, 58.5% of all Google searches in the US end completely without a single click (SparkToro / Similarweb, 2024). On mobile devices, the zero-click rate is 75%. In Google's experimental AI Mode, it rises to 93%. Users get their answers directly in the search engine without needing to visit a website.

But there is a positive side that many overlook. Those cited in AI answers benefit enormously: traffic from AI searches converts at 14.2% compared to 2.8% for classic Google traffic, roughly 5x better (Exposure Ninja, 2025). Claude traffic even converts at 16.8%. Overall reach decreases, but the quality and purchase intent of visitors increases significantly.

The scale of transformation becomes even clearer: in June 2025, AI platforms generated 1.13 billion referral visits, a 357% increase year-over-year (Similarweb, 2025). And 50% of consumers now actively seek AI-powered search results. 44% describe them as their primary information source, ahead of classic Google search at 31% (McKinsey AI Discovery Survey, August 2025).

Which GEO strategies are proven to work?

The most important foundational research on GEO came from a team at Princeton University, Georgia Tech, and the Allen Institute for AI. Their study, published at the prestigious KDD Conference 2024, systematically examined which optimization methods increase visibility in generative search engines.

The results are clear:

Particularly noteworthy: the combination of readability and statistics outperforms any single measure by 5.5%. Content that reads fluently while providing concrete numbers with sources performs best. Keyword stuffing, on the other hand, performs 10% worse than baseline. What once worked in traditional SEO actively hurts in GEO.

Schema markup also plays a central role. BrightEdge confirms that structured data increases AI citation rates by 44% (BrightEdge, 2025). SearchVIU demonstrated that ChatGPT, Claude, Perplexity, and Gemini all consider schema markup when selecting sources.

How does E-E-A-T strengthen AI visibility?

AI systems rate content as reliable when they detect clear signals of experience, expertise, authoritativeness, and trustworthiness. Since the December 2025 Google Core Update, E-E-A-T criteria apply not only to sensitive topics like health and finance, but to all competitive search queries.

Specifically, this means for GEO:

• Author profiles with verifiable references: Every article needs a named author with Person schema. AI systems link author names with expertise signals through LinkedIn, specialist publications, and company affiliation. Anonymous or generic attributions ("Admin", "Team") are ignored.
• Back every claim with data: For pages ranked 5th or lower, citing authoritative sources increases visibility by up to 115%. The format matters: number, source as link, year. Unsupported claims are less likely to be picked up by AI systems.
• Show experience, not generic advice: Phrases like "In our project for client X, we..." or "From our experience with 50+ web projects, we see..." are stronger signals than generic recommendations. AI systems recognize these patterns and weight them higher because they cannot be generated from general training data.
• Build brand authority: Brand mentions on the web correlate at 0.664 with appearances in AI Overviews. Backlinks only at 0.218 (Profound/Semrush, 2025). This is a paradigm shift. Brand awareness far outperforms traditional link building.

How should content be structured for AI citation?

AI systems preferentially extract answers from the first sentences of a section. 44.2% of all LLM citations come from the first 30% of a text (Growth Memo / Kevin Indig, 2026). That is why structure is critical.

Use questions as headings

60-70% of H2 headings should be phrased as questions. AI systems extract answers directly from question-answer formats. Google's "People Also Ask" boxes also favor this format. The remaining 30-40% stay as statement headings for variety.

Answer-first: lead with the answer

Every H2 section starts with a 40-60 word opening paragraph containing the core message along with a statistic and source. AI systems scan this opener and decide within milliseconds whether the section is citable. Burying the answer in paragraph 3 means not getting cited.

Keep paragraphs short and concise

Between headings, aim for 120-180 words. This range achieves 70% more ChatGPT citations than longer blocks (SE Ranking, 2025). Each paragraph should lead with the most important point. Maximum 2-3 sentences per paragraph. 79% of users scan content rather than reading it (Nielsen Norman Group).

FAQ sections as AI triggers

FAQ sections with schema markup increase AI citation probability by 28%. Each answer should be 40-60 words and function as a self-contained information unit, understandable without surrounding context.

Tables and structured comparisons

Comparison tables with proper HTML markup achieve 47% higher AI citation rates (Search Engine Land, 2025). AI systems prefer structured data over running text because it is easier to parse and verify.

How do you increase reach beyond your own website?

88-92% of AI citations come from off-site signals, not from on-page optimization alone. Most companies invest 90% in their own content and 10% in external presence. For GEO, the ratio should be closer to 40% owned and 60% earned media.

The strongest correlations with AI visibility according to an Ahrefs study of 75,000 brands (December 2025):

YouTube is by far the strongest lever. YouTube citations in AI Overviews grew by 414% (BrightEdge, Q1 2025). How-to videos grew by 651%, visual demos by 592%. YouTube is cited 200x more frequently than any other video platform.

Reddit citations also grew by 450%. Google has a $60 million API deal with Reddit, which positions the platform particularly strongly in AI answers. However, Reddit requires authentic engagement: at least 3-5 months of active community participation before promotional content is accepted.

User feedback and review platforms

Reviews on Google, industry portals, and platforms like G2 (22-23% of all review citations) measurably strengthen authority. Companies with presence on 4+ review platforms receive 2.6-3.5x more AI citations than those with only one platform.

Regular updates are mandatory

76.4% of the most-cited ChatGPT pages were updated within the last 30 days (Digitaloft, 2025). 85% of AI Overview citations come from content less than 2 years old. Content older than 3 months receives 3x fewer AI citations. At Perplexity, relevance decay begins after just 2-3 days.

GEO is not a one-time project but a continuous process. Plan for at least quarterly content updates with substantial changes (30%+ new content).

What results are companies achieving with GEO?

The first companies to consistently implement GEO strategies report impressive results:

These numbers show: GEO delivers fast, measurable results. Go Fish Digital achieved 43% more AI traffic in just 3 months. The key was the combination of content structure, source citations, and regular updates.

Technical prerequisites for AI visibility

AI crawlers like GPTBot, ClaudeBot, and PerplexityBot do not execute JavaScript. A Vercel analysis of over 500 million GPTBot requests found zero evidence of JavaScript execution. 46% of AI agent visits start in read-only mode, meaning text only, no CSS or JavaScript.

Therefore, the following technical requirements are essential:

• Server-Side Rendering (SSR) or Static Site Generation (SSG): Content must be delivered as complete HTML. Single-page applications with client-side rendering are invisible to AI crawlers.
• Keep robots.txt open: Explicitly allow GPTBot, ChatGPT-User, ClaudeBot, and PerplexityBot. GPTBot traffic grew by 305% year-over-year, PerplexityBot by 157,490% (Cloudflare, 2025).
• Load time under 200 ms TTFB: AI crawlers have a hard timeout of 3-5 seconds. Pages over 600 ms TTFB are frequently skipped. Slow pages are completely excluded from the AI citation pool.
• Schema markup in HTML source code: Embed BlogPosting, Person, FAQ, and BreadcrumbList directly in HTML, not injected via JavaScript. Pages with 3+ schema types have a 13% higher AI citation probability.
• Provide llms.txt: A markdown file in the website's root directory that helps AI systems understand the site's content and structure. Keep it under 10 KB, simple URLs with short descriptions.