For thirty years search handed you links and you picked one. AI assistants hand you a single answer and name a few businesses, so now the machine picks, and there is no page two of an answer. The firms that build verifiable evidence early become the names the machines learn first. That work has a name: AI search optimisation.
Somewhere in the UK right now, a customer is not searching. They are asking. They type a full sentence into ChatGPT or Google's AI Mode, something like "who's the best boiler engineer near me for a same-week fix", and back comes a short paragraph that names two firms and says why. The other forty firms in that town are not on page two. There is no page two. They are simply not in the answer, and the customer will never learn they existed. Multiply that one moment by the millions of times it happens each day and you have the whole shift in a sentence: search used to hand you a shelf of options, and now it hands you a recommendation.
Nothing this big has happened to how people find businesses since Google itself turned up, and most companies have not clocked it. So let's be precise about three things: what actually changed, why it outweighs any algorithm update you have already survived, and what an honest response to it looks like.
Thirty years of search, in four acts
You cannot see how strange this moment is without seeing where it came from. Search reinvents itself about once a decade, and every time it does, the rules of being found get rewritten under everyone's feet while they carry on as before. The shape of it never changes. The interface moves, the winners change with it, and whoever adapts early becomes the default that everyone else lives with for years.
Directories and early Google
The web was still small enough to list by hand. You found a business in Yahoo's directory, or, from September 1998, through a new engine that ranked pages by who linked to them. To be found was to be catalogued.
Ten blue links
For two decades the bargain was simple. Google gave you ten results, you fought to sit near the top, and a top spot meant a stream of clicks. A whole industry grew up around winning that click. Rank, win the click, repeat.
The first answers
Apple put Siri on the iPhone 4S in October 2011, and in January 2014 Google started lifting a single answer to the top of the page as a featured snippet. The SEO world called it "position zero." Search was trying, for the first time, to answer you rather than just point.
ChatGPT arrives
OpenAI released ChatGPT on 30 November 2022, and within two months it was the fastest-growing consumer application in history. Millions of people were suddenly getting useful answers to real questions with no list of links at all.
Google answers first
Google launched AI Overviews in the US in May 2024 and brought them to the UK that August. The AI-written summary now sat above the blue links, on the most-used search engine on earth, answering before anyone had scrolled.
AI Mode ships
Google rolled AI Mode out to the UK at the end of July 2025: a full conversational search built on Gemini. Not a summary bolted onto the results, but search itself rebuilt as a conversation that asks and answers in turn.
Search rebuilt AI-first
At Google I/O in May 2026, Google confirmed AI Mode had passed one billion monthly users and rebuilt the search box itself around AI. That June, OpenAI began running product ads inside ChatGPT. The answer, not the link, had become the product.
Four acts, and in every one the firms that moved early set the defaults everyone else lived with for years. We are standing at the start of the fifth.
The break point: the machine chooses now
Here is the line that divides the last thirty years from the next thirty. Search used to return links and a human chose between them. Now it returns an answer, and the machine has already chosen before you see a thing. Ranking was a contest you could watch and climb. Being named in an answer is a decision taken somewhere you cannot see, with no second page to catch you if you miss.
None of that is a forecast. It is already how a large slice of searches behave, and the numbers are blunt.
Most searches now finish on the results page itself. The answer was enough, so nobody clicked through. And when Google does put an AI Overview on top, the pull away from websites gets sharper still: the top organic result loses well over half its clicks against a normal result for the same kind of query.
You might assume a number one spot on Google still keeps you safe, that the AI just reads the top results and parrots them back. It does not work like that. When Ahrefs ran 15,000 prompts through ChatGPT, Gemini and Copilot in 2025 and checked the citations against Google's own rankings, only about 12% of the pages those assistants quoted also ranked in Google's top ten for the same question. The machines are drawing their answers from a different pool of sources than the one you have spent years optimising for.
For local businesses the door is narrower again. When SOCi looked at nearly 350,000 business locations for its 2026 Local Visibility Index, ChatGPT recommended roughly 1.2% of them, about one in eighty, while the Google map pack surfaced businesses far more freely. AI assistants are far pickier, and more than half the firms that win in Google local search never make it into the AI answer at all. None of this means search is shrinking. The traffic is moving, not evaporating: AI platforms sent 1.13 billion visits to the top 1,000 websites in June 2025, up 357% year on year (Similarweb via TechCrunch, 2025). The audience is right there. The only question is whether the machine says your name when they ask.
What the new discipline actually is
When the rules of being found change, a new craft grows up to meet them. This one goes by several names, and the argument over which name is right is mostly noise. What counts is the work underneath, because the work is concrete and it stays the same whatever you call it.
You will hear it called AI SEO, answer engine optimisation (AEO), or generative engine optimisation (GEO). Peel the labels off and the job comes down to three things. Evidence a machine can verify: clear, consistent facts about your business that an AI can check against other sources and trust. Answers a machine can lift: content written so a model can quote it cleanly, in the shape a real question takes. And presence in the sources a machine trusts: the reviews, directories, and third-party pages the models actually read when they build an answer, what we do as citations and authority work. We wrote a plain-English decoder of the three terms in SEO vs AEO vs GEO, and set out the full method in our guide to AI search optimisation in the UK. Search is not dead. It is changing shape, and the craft is changing right along with it.
Why this wave will be sold badly
Every platform shift breeds snake oil, and this one will be no different. In the ten-blue-links era the con was "guaranteed number one rankings", a promise no honest agency could keep because nobody controlled Google's algorithm. The same pitch is back in a new costume: "guaranteed placement in ChatGPT." Nobody can honestly promise that one either, because nobody controls what a model says.
The tell never changes. Be wary of anyone selling certainty over a system they do not own, and of tactics dressed up as secrets. Two of the most-hyped "secrets" doing the rounds right now simply do not do what the sellers claim. An llms.txt file, sold as a way to make AI favour your site, has no confirmed ranking effect at all: Google clarified in June 2026 that these files are not used by Search and hand you no visibility benefit either way. Schema markup on its own does not buy citations either. When Ahrefs tracked 1,885 pages that added schema in 2026, the citations barely moved, because schema tends to sit on sites that were already strong rather than being what made them strong.
What sets us apart is not a better adjective. It is behaviour you can verify. We publish our sources so you can check the claims for yourself. We tell you plainly what does not work, the two examples above included, because a firm that only ever says "yes, that helps" is selling to you, not advising you. We make the fixes ourselves, on the website you already have, instead of posting you a report and walking off. Our prices are fixed and the terms roll month to month, so we have to earn next month by being useful this one. And every month you get an evidence report showing what the AI assistants actually say about your business, in their own words, so progress is something you can see rather than something we assert. We do not guarantee outcomes, because nobody honestly can, and anyone who does is telling you something about their honesty.
The window is open now
The answers are being written now, this quarter, out of the evidence that exists today. Models learn who to trust from patterns that stack up over time, so an early, consistent presence compounds while late arrivals start from further back. Being early has never counted for more, because the defaults set this year are the ones customers will hear repeated for years.
This is the part most businesses will get wrong simply by doing nothing. The shift is quiet. No algorithm-update email lands, no ranking drop gives you a fright, just a slow, invisible reallocation of who gets recommended and who gets left out of the sentence. By the time it is obvious, the early movers are already the names the machines reach for by default. The age of answers has begun, most businesses have not noticed, and the ones who act while it is still early will be the names the machines learn first.
Questions people ask
Is SEO dead now that AI answers questions directly?
No. Search is changing shape, not vanishing. Traditional ranking still decides plenty of queries, and the fundamentals of clear, trustworthy content carry straight over. What has changed is that ranking on Google no longer buys you a place in the AI answer: only about 12% of pages cited by AI assistants also rank in Google's top ten for the same prompt (Ahrefs, 2025). You now have to earn a spot in both places, which is why the work has broadened rather than ended.
Can anyone guarantee my business will appear in ChatGPT or AI Overviews?
No, and be wary of anyone who says they can. Nobody controls what an AI assistant says, just as nobody controlled Google's ranking algorithm back in the days of "guaranteed number one" promises. What can honestly be done is concrete: build the verifiable evidence, quotable answers, and trusted third-party presence that make a model more likely to name you, then measure what it actually says month after month. We report on real behaviour, not guarantees.
What is the difference between SEO, AEO and GEO?
They are three names for closely related work. SEO is optimising to rank in traditional search results. AEO, answer engine optimisation, is about being the answer an AI lifts and quotes. GEO, generative engine optimisation, is about being cited by generative systems like ChatGPT and AI Overviews. In practice the tasks overlap so heavily that the label matters far less than the work. We pull the three apart properly in our decoder article.
Sources
- SparkToro, In 2026, Less than One Third of Google Searches Still Send a Click (2026)
- Ahrefs, Update: AI Overviews Reduce Clicks by 58% (2026)
- Ahrefs, Only 12% of AI-cited URLs rank in Google's top 10 for the original prompt (2025)
- SOCi, 2026 Local Visibility Index (2026)
- Similarweb via TechCrunch, AI referrals to top websites were up 357% year over year in June, reaching 1.13B (2025)
- Ahrefs, We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved. (2026)
- Search Engine Land, Google says llms.txt files won't harm or help your search rankings (June 2026)
- Google, AI Mode on Google Search in the UK (July 2025)
- Google, Google Search updates at I/O 2026 (May 2026)
Stagg Studios does this for you
Schema, structure and the technical foundations that decide whether Google and AI can read you, kept in order every month on the site you already have. One simple price: £100 a month or £1,000 a year, no lock-in.
Get startedSee pricing