For twenty years, winning in search meant landing in Google’s top 3. Today a growing share of searches ends without a single click: an AI gives the answer. And in that answer there aren’t ten blue links, there are two or three recommended brands. The question is no longer just «how do I rank my website», but «does the AI mention me when someone asks about what I do?»
According to Similarweb, zero-click searches went from 56% to 69% in a single year after Google rolled out AI Overviews. And according to Ahrefs, when an AI summary appears, the first organic result loses around 35% of its clicks. The game has changed, and brand is at the centre of it.
SEO, AEO and GEO: what each one is
- SEO (Search Engine Optimization): ranking pages in search engine results. It’s still the foundation.
- AEO (Answer Engine Optimization): preparing your content so answer engines (AI Overviews, AI Mode, Perplexity, voice assistants) extract it and use it as a direct answer.
- GEO (Generative Engine Optimization): getting generative models such as ChatGPT, Gemini, Claude or Copilot to mention, cite and recommend you when they generate an answer.
They aren’t separate disciplines: they share foundations. But AEO and GEO add something classic SEO could afford to ignore: your brand’s reputation outside your website.
How an LLM generates an answer
To influence what an AI says, you first need to understand where it gets what it says. Large language models (LLMs) answer by combining two sources:
1. What they learned during training
A model is trained on huge amounts of text: websites, books, forums, press, documentation. It doesn’t store pages, but patterns and associations: which brands appear alongside which topics, with which adjectives, how often. If your brand appears a lot and in relevant contexts, the model «knows» it. If it barely exists outside your website, for the model it practically doesn’t exist.
This knowledge has a cut-off date: whatever happened after training isn’t there. That’s why the second piece matters more and more.
2. What they search for in real time
When the question calls for it, the assistant searches the web, reads a few pages and builds the answer from them. This is known as retrieval-augmented generation (RAG). The process is usually similar everywhere:
- It rephrases your question into several more specific searches. Google calls this query fan-out in AI Mode: one question becomes many sub-queries.
- It retrieves sources from a search index: Google’s in AI Overviews and AI Mode, proprietary and third-party indexes in ChatGPT or Perplexity.
- It selects and summarises the clearest and most reliable passages.
- It generates the answer by mixing those sources with what it already knew, and often cites them.
Each platform has its nuances. Perplexity always cites its sources and favours original data. ChatGPT tends to rely on sources like Wikipedia, Reddit and the media. Google’s AI Overviews draw on the same index as the search engine, so classic SEO still counts. But in every case one idea repeats: the model prefers what many independent sources say consistently.
Why being talked about matters more now
In classic SEO, links were the currency of authority. In AI search, the currency is mentions. An Ahrefs study of 75,000 brands found that brand mentions on the web have a 0.66 correlation with visibility in AI answers, compared with 0.22 for backlinks. Three times more.
And those mentions don’t come from your website: according to Muck Rack, 82% of AI citations come from earned media: press, analysis, comparisons, forums. In other words, from what others say about you.
Your website says who you are. AI repeats what others say.
This turns brand, which many companies treated as something «soft» and hard to measure, into a direct acquisition asset. A brand with a consistent reputation in many places is a brand AI recommends.
How to build a reputation AI can see
- Define clearly who you are and repeat it the same way everywhere. The same description of what you do and for whom on your website, LinkedIn, directories, profiles and press releases. Consistency helps the model associate you with your category.
- Work on press and industry media. Articles, interviews, opinion pieces and appearances in «the best tools for…» comparisons.
- Be where people talk. Forums, Reddit, professional communities, reviews on Google, G2, Trustpilot or your industry’s equivalents. These are sources models read a lot.
- Publish your own data. Studies, surveys, benchmarks. Original content is what gets cited most, because nobody else has it.
- Write to be cited. Direct answers up top, clear definitions, FAQs, tables and lists. If a passage makes sense on its own, it’s easier to extract.
- Keep content fresh. Recently updated pages get cited considerably more than old ones.
- Don’t block the AI crawlers you want to find you. Check your robots.txt and your CDN settings.
How to measure it
AI visibility doesn’t have a Search Console yet, but you can track it:
- Make a list of the 20 or 30 questions your ideal customer would ask and run them every month in ChatGPT, Perplexity, Gemini and AI Overviews. Note whether you appear, in what position and next to whom.
- In GA4, create a channel for traffic coming from chatgpt.com, perplexity.ai, gemini.google.com or copilot.microsoft.com.
- Track your brand mentions in media and social, not just your links.
- There are dedicated AI visibility tools that automate this tracking; useful once the volume grows.
In short
AI doesn’t replace brand: it amplifies it. The brands people mention, review and cite are the ones models recommend. Investing in reputation, useful content and presence in your industry’s conversations is no longer just branding: it’s how you keep showing up when your customer asks.
If you’ve checked how your brand shows up in AI answers, I’d love to hear what you found: get in touch.
Sources: compilation of studies by Ahrefs, Similarweb and Muck Rack · Jasper: GEO vs AEO
