Every week someone tells us SEO is dead (cue the yawns and eye rolling) and AI is the whole game now. Usually it’s a LinkedIn post – the business soap box. Occasionally it’s a prospect who has been advised to move their entire budget into “AI visibility” by someone who was selling short-form video strategy eighteen months ago.
We get the instinct – “don’t do this and risk missing out”. Search is changing faster than it has in a decade, and nobody wants to be the last team optimising title tags while the world moves on.
But there’s a difference between turning a ship and spinning it.
Will AI replace SEO?
The short answer: no, not yet, and probably not in the way people mean.
AI is a growing share of how people find things, and it will keep growing. But right now it is a small channel sitting next to a very large one. The honest move is to start working on it without gutting the thing paying the bills.
Is SEO dead? Here’s what the data says
Conductor’s 2026 benchmark covered 13,770 domains and 3.3 billion sessions. AI referrals averaged 1.08% of site traffic. Ahrefs, across roughly 82,000 sites, put it nearer 0.1%. Chartbeat found AI chatbots still send under 1% of publisher page views, even after ChatGPT referrals grew more than 200% across 2025.
The most generous figure we’ve seen comes from First Page Sage, which tracked 218 sites and 3.4 billion sessions and had AI platforms at 6.24% of all sessions by mid-2026, up from 0.11% at the start of 2023.
Organic search, over the same window, is still doing about half the work. BrightEdge puts it at roughly 53% of trackable traffic.
So: somewhere between 1% and 6%, growing quickly, next to a channel doing 53%.
That’s not a reason to ignore AI. It’s nowhere near a reason to bet the house on it.
The AI number is bigger than it looks, and that helps our case
Two caveats, because context is king and pretending 1% is the whole story would be lazy.
First, a lot of AI traffic doesn’t announce itself. One 2026 analysis found roughly 70% of AI referrals land in GA4 as direct. If you’re reading a standard channel report, you’re undercounting.
Second, and this is the important one: the biggest AI influence on your traffic isn’t ChatGPT. It’s Google. Previsible’s study of 6.77 million LLM-driven sessions found AI Overviews and AI Mode already drive more AI-influenced traffic than ChatGPT, Claude, Gemini, Perplexity and Copilot combined.
Read that again, because it’s the entire argument. The dominant AI surface sits inside the search engine you are already optimising for.
GEO vs SEO: mostly the same job
Getting cited by a model isn’t a separate discipline with its own dark arts. Models pull from pages that are crawlable, well structured, unambiguous and corroborated somewhere else. Clear entities. Clean markup. Content that answers the question rather than circling it for 1,200 words. Mentions and links from sources the model already trusts.
Ultimately, that’s SEO. It has been for years. The output has changed a lot. The inputs have changed far less than the self-proclaimed ‘experts’ would like you to believe.
Rip out your SEO programme to fund a GEO programme, and you defund the thing feeding your AI visibility.
Small channel, sharp intent
That small number still earns attention on its own merits.
Onely found AI search platforms accounted for 0.5% of traffic but 12.1% of signups. Adobe data showed AI-referred retail traffic converting 42% better than non-AI traffic in March 2026, having converted worse a year earlier. These visitors arrive pre-qualified. They’ve done the research in the chat window and turn up ready to evaluate.
Low volume, high intent. That’s a channel you build for. It just isn’t one you remortgage for.
What turning the ship actually looks like
A few degrees, held steady:
- Measure it properly. Segment AI referrers so they stop hiding inside direct. You can’t argue about a channel you can’t see.
- Track visibility, not just clicks. If the answer mentions you and nobody clicks, that still happened.
- Fix the fundamentals. Structured data, internal linking, clear entity signals. They serve both surfaces at once.
- Publish things worth citing. Original numbers, real client results, actual opinions. Models don’t cite paraphrase.
- Keep the SEO programme running. The 53% is what pays for the experiment.
The uncomfortable bit
A lot of the loudest AI-first advice is sold by people who need it to be true. A brand new channel is a brand new retainer, and it conveniently arrives without a decade of benchmarks to be judged against.
We’d rather tell you the boring thing. Move 10 to 15% of your effort across. Hold the rest. Review it quarterly and turn further as the numbers turn. If AI referrals hit 20% of your traffic next year, turn harder and turn fast. Your own analytics will tell you when, long before LinkedIn does.
Ships that turn slowly still get where they’re going – arguably more safely and securely.
Words by: Andy Robson
Agency Founder & Director
Andy founded SQUIDGY and has spent the best part of two decades working in search. Over that time he's seen the industry reinvent itself more than once, sat through every "SEO is dead" prediction going, and learned to tell the lasting shifts from the noise. He writes from experience rather than theory, drawing on years spent in Search Console, technical audits and the unglamorous detail of what actually makes sites rank.
