What is happening to SEO?
Updated: May 2026. We refresh this page regularly to keep pace with fast-moving AI platforms and policies.
When we wrote the first version of this piece in late 2024, the question was still “what is happening to SEO?” – with all the uncertainty that the present-continuous tense implies. A year and a half later, the answer is clearer. Search didn’t vanish. It got eaten.
For brands, the question has shifted. It’s no longer “how do we get found?” – it’s “how do we get cited?”.
The data that wasn’t there before
In late 2024, we cited Dotdash Meredith reporting that Google search traffic to its sites had halved since 2020. Google denied at the time that AI search features were killing website traffic overall. Eighteen months on, the picture is unambiguous.
The Reuters Institute’s Journalism and Technology Trends 2026 report, drawing on Chartbeat data from over 2,500 publishers globally, found Google search traffic to publishers dropped a third in the year to November 2025. Google Discover referrals were down 21% over the same period.
The damage at individual outlets is starker. Business Insider reported a 55% decline in organic search traffic between April 2022 and April 2025. HuffPost lost half of its search referrals over the same window. The Verge, ZDNet and HowToGeek each lost more than 85% of their Google traffic. Digital Trends went from 8.5 million monthly clicks in early 2024 to fewer than 270,000 by January 2026.
Pew Research, tracking nearly 69,000 real Google searches, found that when AI Overviews appear, users click traditional results 8% of the time – versus 15% without an AI summary. That’s roughly a 47% drop in clicks. Similarweb data shows zero-click searches rose from 56% in May 2024 to 69% a year later.
So a third of search traffic to publishers has gone, and the trajectory continues. The Reuters Institute’s survey of media leaders projects a further 43% decline over the next three years.
The browser is the agent now
The 2024 piece predicted that AI-powered browsers – Arc, Comet, ChatGPT’s Atlas – would shift from tools to active agents. That has happened. Google’s AI Mode, launched in May 2025, is now processing over a billion queries a month with 75 million daily active users. In April 2026 Google replaced the familiar “Search” prompt on Android with “Ask Google”.
For users, the implication is simple: many never see the open web at all. They ask, the AI synthesises, and they act on the answer. We’ll be unpacking what this means more broadly in our forthcoming piece on AI agents.
For brands, the implication is that being indexed isn’t enough any more. You need to be present – consistently and credibly – in the sources the AI is drawing from.
A new vocabulary, with limits
A new layer of acronyms has built up over the past year: GEO (generative engine optimisation), AEO (answer engine optimisation), LLMO (large language model optimisation). The labels differ; the underlying practice is broadly the same – making your content easier for an LLM to retrieve, understand and cite.
These practices are real. Structured content, clear formatting, schema markup, factual depth and authoritative sourcing all improve AI visibility. The discipline matters.
What’s harder is measurement. LLM responses are non-deterministic – ask ChatGPT or Google’s AI the same question 100 times and there’s a less-than-1-in-100 chance of getting the same brand list twice. There is no “position 1” in ChatGPT. Visibility is mention frequency across many prompts, not a fixed ranking. Any tool that promises a single, clean “AI ranking” should be treated as directional rather than definitive.
The vocabulary is useful shorthand. The measurement layer is still maturing. Both are worth engaging with.
From visibility to legibility, with evidence
In 2024 we argued that the brands which would win in AI-first discovery were the ones most legible – clear, credible, and structured well enough for both humans and machines to recognise. The data now backs this up.
Recent research has found that around 88% of users in Google’s AI Mode took the AI’s shortlist without checking elsewhere. The AI’s top pick became the user’s top pick 74% of the time. About a quarter of users overrode the AI’s order – and when they did, it was usually because they recognised a brand lower down the list.
That last point matters most. It means the brands that endure in AI-first discovery will be both legible (cleanly retrievable by the systems doing the synthesising) and recognisable (memorable enough to override the AI’s default order). One without the other isn’t enough.
Earned media is doing more work than ever. Recent analysis suggests distributing content across multiple publications can increase AI citations by up to 325% compared to publishing only on your own site. A feature in a respected outlet now feeds two systems at once: human credibility, and the AI citation graph that decides what gets recommended.
Durable principles for AI-first discovery
The principles haven’t changed dramatically. The leverage points have.
Be clear. Focus on usefulness, not fluff. AIs prefer answers; humans do too.
Be structured. Semantic HTML, proper headings, schema markup, accessible formats. Make it easy for the systems indexing the web to understand what you’re saying.
Be consistent. Align tone, messages and facts across every channel. LLMs cross-reference – inconsistency reads as unreliability.
Be cited. Earned media, partnerships and third-party mentions matter more now than they did a year ago. They feed the citation graph.
Be recognisable. Brand still wins overrides. The work of brand building – distinctive voice, memorable identity, repeated exposure – directly shapes whether your name gets picked from an AI’s list.
Be adaptive. Revisit the strategy regularly. The interfaces will keep changing. The principles probably won’t.
What this means for marketing leaders
A few practical shifts to consider.
Rebalance the channel mix. Organic search is no longer a reliable single channel. Diversify into earned media, partnerships, communities, owned audiences (newsletters, podcasts) and creator collaborations.
Invest in brand. The data is unambiguous: when an AI gives a list, the names people already know win the click. Brand-building work that felt indulgent five years ago is competitive infrastructure now.
Track AI visibility. The category of tools that monitor brand mentions in LLMs is maturing fast. They’re worth piloting – with a clear-eyed view that the metrics they produce are useful as a direction of travel, not a fixed scoreboard.
Don’t panic-buy. The pace of new AI-optimisation services means there’s a lot of urgency in the market. The fundamentals – clear content, structured data, brand recognition, earned credibility – are stable enough to invest in confidently.
The brands that endure won’t be the ones gaming the next algorithm. They’ll be the ones that are easy to understand and impossible to ignore.
If you spot a change in the platforms or measurement tools that affects this guidance, tell us. We keep this page updated so it stays practical and current.
Last updated: May 2026