Rise Up Media

A recent study analysed 1.2 million AI-generated answers and 18,012 verified citations to understand how models like ChatGPT select sources. 🔍 The clearest finding was structural: 44.2% of citations came from the first 30% of an article, while only 24.7% came from the final third. In practical terms, that means content buried halfway down your “ultimate guide” is far less likely to be surfaced by AI systems. The pattern continues at the paragraph level. AI most often cites information from the middle of paragraphs, not just the opening line. Around 53% of citations came from the middle sentences, compared to roughly 24.5% from the first and 22.5% from the last. The study also identified five traits that make content more likely to be cited: 📍 Clear, definitive language performs better than hedged statements 💡 Question-based headings often function as prompts, with the following paragraph acting as a direct answer. 👀 Highly cited content includes significantly more proper nouns, such as specific brands, tools, and people, rather than vague generalisations. 🤝 Balanced and moderately objective writing outperforms both dry encyclopaedic prose and highly opinionated takes. 📖 Shorter sentences and plain business language outperform complex academic phrasing. The practical implication is simple but uncomfortable. You need to front-load your definitions, entities, and core claims, have key concepts early, and answer real questions directly. If AI systems are becoming a primary discovery layer, then content strategy needs to evolve with them.

New research analyzing 1.2 million #ChatGPT citations reveals interesting insights for AI visibility. 💡 Key findings highlight a 'ski ramp' pattern, in which 44.2% of citations come from the first 30% of an article, showing the need to front-load critical information. That means moving away from the traditional narrative suspense and towards a "bottom-line up front" approach. AI-preferred content looks like: 🔍 Definitive language: AI favors direct statements and clear definitions. 🤝 Conversational structure: Content with questions in headings and immediate answers sees higher citation rates. 💸 Entity richness: Articles dense with proper nouns (brands, tools, people) are more likely to be cited. ⚖️ Balanced Sentiment: An "analyst voice" that explains how facts apply, rather than just stating them, is preferred. 🧐 Business-grade writing: Clear, concise language at a college reading level is favored over overly academic prose. This confirms that adapting your content structure and linguistic style is essential for staying ahead in the AI era.

Search is slowly starting to be rerouted through AI systems like ChatGPT, Perplexity, Gemini, and Google’s AI Overviews, and more users are getting full answers without ever clicking a blue link. 🔁 Here's what agencies are doing differently from classic SEO. 🎯 Listicles are becoming an AI playground. Agencies are prioritizing “best of” placements because AI models cite these formats disproportionately when generating answers. 🏉 Teams are building brand entities with clean site structure, structured data, and consistent mentions across the web so the model recognizes the company as a “thing,” not just a URL. 🧹 A growing number of agencies are tracking LLM mentions, which brands show up in ChatGPT, Gemini, or Perplexity for specific queries, and treating that as a performance metric alongside traffic and conversions. The agencies are testing, measuring, and re-educating clients on what success looks like in an AI-first SERP.

59.8% of “why” queries trigger an AI Overview. AI systems are looking for clean, intent-matched answers. ✨ 💡 About 76% of AI citations come from pages already in the top 10, with a median rank of #2, AI systems pull from the same index as search. 📉 There’s near-zero correlation between length and AI visibility. Most grounding happens around ~540 words, and returns flatten hard past 2,000, the answer needs to be early and direct. 🔍 Brand mentions matter, with mentions across authoritative pages correlating strongly with AI Overview visibility. Nearly 50% of citations come from “best of” lists. SEO is becoming search presence engineering: part rankings, part brand distribution, and part technical groundwork for machines.

Google just took AI assistant a step closer to personal operating system. 🤖 Gemini has rolled out a beta feature in the U.S. called Personal Intelligence, and the idea is simple: let the model see your Google world. You can connect apps like Gmail, Photos, YouTube, and Search, and Gemini can start answering questions using your own context. 👀 A few practical examples: 😎 If you need a tire size, it can scan your Photos instead of sending you back to Google. 🏖️ If you're planning a trip it can look at past bookings in Gmail and locations in Photos to suggest something that actually fits how you travel. You choose exactly which apps get access, you can disconnect them anytime, and Gemini doesn’t train on your personal data. It references the data to answer, while training happens on filtered prompts and responses, not your inbox or photo library. If AI is reasoning over private data, public web, brand content, visibility becomes “be the source the model recognizes when it’s stitching together answers inside someone’s personal context.”

Search engines and AI systems are increasingly treating brands as entities. What gets recommended depends on how clearly and consistently your brand shows up across the web. 🔍 At the basic level, an SEO brand guide forces alignment on your official name, how it’s shortened, what category you actually sit in, and how someone else should describe you in a sentence or two. Those signals shape how search systems understand who you are and when you’re relevant. It also sets real language rules, words you use, words you avoid, and how technical you go by default. That consistency makes your content easier to recognize and easier to trust. 🤝 Search personas are people who don’t yet know your brand, and they’re still trying to name their problem. If you understand what they search for and where they look for answers, you can show up earlier. Differentiation and structure distinguish what makes you different, captured and written in plain terms. And clear rules for how pages are built, linked, and laid out so everything feels coherent, whether it’s written by a human or an AI tool.

SEO revolves almost entirely around links, but another signal is quietly catching up. 🏃♀️ LLMs read context. They notice when your company shows up in industry reports, comparison guides, interviews, and explainers. Over time, that repetition defines which brands get pulled into AI-generated answers. Here’s how to become a reference material brand ⬇️ Create content people will genuinely cite, such as research with numbers, and that people will rely on. Build diagrams that explain something others struggle to teach. If your work helps someone else tell a story, your brand travels with it. #LLMs draw heavily on trusted publications in each niche, and this is where they learn from. Find those sources, then add something useful instead of begging for a link, like providing missing data or clarifying a concept the writer glossed over. When you release something meaningful, time the announcements, interviews, and articles to go out together and be the helpful expert. A short burst of mentions is loud, and repeated mentions over time build a durable reputation. Brand mentions are becoming an authority layer of their own, and you need to be discoverable wherever people ask questions, including #AI tools.

AI can help you move faster in SEO, but it can’t bypass the fundamentals. 🤷♂️ AI doesn’t understand why someone is searching, and it doesn't have real experience with a product, a market, or a trade-off. It's also not responsible for accuracy, claims, or outcomes. Search engines aren’t trying to “detect AI.” They’re still doing what they’ve always done: comparing pages and asking which one is more useful, clearer, and more trustworthy. If AI helps you research better, clarify thinking, and reflect reality more accurately, it’s a win. If it just helps you publish more pages without improving the thinking behind them, it’s just faster noise. AI just made the gap more obvious.

Most people use LLMs to write faster. 😎 That’s the least interesting thing they can do. Tools like ChatGPT or Claude can scale research you’d otherwise never have time to do correctly. Here are three places where this works in practice: 1️⃣ Customer feedback Instead of skimming a handful of survey responses, LLMs can help you make sense of thousands. By querying structured data, spotting patterns, and summarising what customers complain about, praise, or get confused by. 2️⃣ Expert knowledge Subject matter experts are busy and tired of repeating themselves. A well-designed LLM interviewer can capture their thinking asynchronously, one question at a time, then turn it into usable insight for positioning, content, and strategy. 3️⃣ Competitors Review site copies, job ads, and historical messaging from competitors. LLMs are good at synthesising this into a clear picture of how competitors position themselves, what they over-promise, and where they leave gaps.

For years, SEO was about ranking on Google. 🤝 Today, every AI model is quietly building its own version of the internet. 🔍 Each has a unique knowledge economy: ✨ OpenAI prioritizes credible web sources and verified entities. ✨ Perplexity favors transparency, structured citations, and freshness. ✨ Anthropic focuses on contextual trust, which is referenced by whom. These differences mean visibility is no longer universal. Your brand might be highly trusted by Google… but invisible to ChatGPT. The question now isn’t “How do I rank on Google?, it’s “Which AI model knows I exist?"