Search as we knew it — type a query, get ten blue links, pick one — is being fundamentally restructured. The shift started quietly with Google's Featured Snippets in 2014, accelerated with voice search, and has now reached a tipping point with the mass adoption of AI assistants and Google's own Search Generative Experience (SGE).
The numbers are stark. According to multiple industry studies, AI-generated answer boxes now appear for over 40% of informational queries on Google. On ChatGPT, users send over one billion messages per day — many of them product research, service comparisons, and "best of" queries that would previously have been answered by clicking through to a website.
What has actually changed
Traditional search was an index and rank problem. Google's job was to crawl the web, index content, and rank pages by relevance. Your job, as a website owner, was to be the most relevant and authoritative page for a given query. The game was fundamentally about ranking.
AI search is a synthesise and cite problem. ChatGPT, Perplexity, Gemini, and Google's AI Overviews don't present a list of pages — they generate an answer and then optionally cite their sources. Your job now is to be cited as a trusted source in that synthesised answer. The game is fundamentally about citability.
This is not an incremental change. These are two completely different optimisation disciplines requiring different strategies, different content structures, and different technical implementations.
How AI answer engines work
To optimise for AI search, you need to understand how AI systems decide what to include in their answers. The process, simplified, works like this:
- Query interpretation: The AI parses the user's intent — not just keywords, but the actual question being asked and what kind of answer would satisfy it.
- Source retrieval: The system retrieves candidate content from its training data, live web index (for systems with web access), or knowledge graph.
- Credibility filtering: Sources are evaluated for authority, consistency, entity recognition, and content quality. Pages with clear entity signals, structured content, and strong E-E-A-T patterns pass this filter. Most don't.
- Answer synthesis: The AI generates a response drawing from qualifying sources. Those sources may be cited explicitly (Perplexity, Google SGE) or influence the answer without citation (ChatGPT training data).
The critical insight is step 3. Most websites fail the credibility filter not because their content is poor, but because it isn't structured in ways AI systems recognise as authoritative. This is entirely fixable.
Key takeaway: AI systems don't rank pages — they synthesise answers from sources they trust. Getting into those answers requires a different kind of optimisation than getting onto page 1 of Google. Both matter. Neither alone is enough.
The traffic impact: winners and losers
Early data on traffic patterns under AI search shows a clear bifurcation. Brands that are cited in AI answers are seeing a new traffic channel emerge — high-intent visitors who have already been "warmed up" by the AI's recommendation. Brands that aren't cited are seeing organic traffic decline as AI overviews intercept their target queries.
The sites losing traffic fastest share common characteristics: thin content, no schema markup, no clear entity presence, and content written purely for keyword stuffing rather than genuine expertise. These sites ranked because they gamed ranking signals — and AI systems are largely immune to those games.
The sites gaining are those with deep topical authority, well-structured content, clear expertise signals, and consistent entity presence across the web. These are characteristics that were always valuable for SEO but are now essential for AI visibility.
What to do about it right now
There are five immediate actions every website owner should take in 2025:
- Audit your AI citation presence. Search your brand and key queries on ChatGPT, Perplexity, Gemini, and Google with SGE enabled. Understand where you appear and where you don't.
- Implement comprehensive schema markup. JSON-LD Organization, Article, FAQPage, and HowTo schema are the minimum. These are the structured signals AI systems use to understand your content.
- Build entity authority. Ensure your brand entity is consistent across your website, Google Business Profile, Wikipedia (if eligible), Wikidata, and major industry directories.
- Restructure content for AI readability. Clear factual assertions, defined entities, logical hierarchies, and FAQ sections all improve AI citability significantly.
- Strengthen E-E-A-T signals. Author credentials markup, editorial processes, external citations, and expert attribution are increasingly important signals for both Google and AI systems.
The brands that act on these changes now will establish citation authority that compounds over time. Those that wait will find themselves competing for AI visibility in an increasingly crowded and expensive landscape.
The rules of search haven't just changed — they've been rewritten. The good news is that the new rules reward genuine expertise and well-structured content, which is exactly what every quality brand already produces. The work is in making that content visible to the AI systems now shaping how customers discover you.
Is your brand invisible to AI search?
Get a free AI Visibility Audit — we’ll check your SEO, AEO & GEO scores and deliver a prioritised action plan within 3–5 days.
Get AI Audit →