A decade ago, SEO was about ranking on a page of blue links. Five years ago, it was about featured snippets and People Also Ask boxes. Today, it is about whether Claude, ChatGPT, Perplexity, and Google's AI Overviews will cite you at all — and if they do, whether anyone will bother clicking through.
The hard numbers
Data from the second half of 2025 shows organic click-through rates on informational queries have dropped 30-50% year over year on queries where AI Overviews appear. The search volume is still there. The clicks are not. For publishers dependent on ad revenue from organic traffic, this is an extinction event. For brands using content to build a pipeline, it is a reorientation.
The framework we use with clients
Split your content into three buckets:
Commodity informational content — "what is X," "how does Y work" — is dead, or at least is now unpaid training data. No one is clicking through. Stop investing in it unless it serves another purpose (authority signals, linking strategy).
Distinctive point-of-view content — takes, analysis, opinions backed by evidence — is more valuable than ever. AI systems need diverse sources to cite, and they preferentially cite content with clear authorship and specific claims. A generic "Complete Guide to Email Marketing" gets nothing. A specific "Why we stopped sending welcome sequences in 2025 and what replaced them" gets cited.
Transactional and high-intent content — pricing pages, comparison pages, case studies — still drives clicks because users have a specific action in mind that an AI chat cannot fulfil. Invest disproportionately here.
Generative Engine Optimisation is a real thing
The practices that make you more likely to be cited by LLMs and AI search overlap with traditional SEO but are not identical. A working list:
- Author structured content — clear H-tags, scannable, with definitions near the top. AI systems chunk and retrieve by structure.
- Make original claims with numbers. "Conversion rates increased 23%" is citable. "Conversion rates improved" is not.
- Include direct quotable answer sentences early in paragraphs. LLMs love the short, confident, declarative sentence that can be lifted.
- Be findable beyond your site. Podcast appearances, guest posts, and industry publication citations feed the training data just as much as your own domain.
- Build topical authority, not keyword portfolios. AI systems care about the totality of what you publish on a subject.
What to stop doing
Stop measuring SEO success by raw organic traffic. Start measuring it by share of branded search, share of AI citations (tools like Perplexity and ChatGPT have APIs you can use to monitor mentions), direct traffic from people who heard about you somewhere in the AI ecosystem, and conversion rate on the traffic you do get. Raw visits are a vanity metric in this new world; informed visits are the currency.
SEO is not dead. It is just that the job has fundamentally changed from "rank on page one" to "be the thing AI systems want to cite, and catch users who still want to click through." Teams that keep running 2021 SEO playbooks will watch their traffic halve over the next 18 months. Teams that adapt will find the game is actually easier — there is less noise, and specificity wins.