Generative engine optimisation
Show up in Google, AI, and everywhere customers search
Ten years ago there was one question worth asking about your business online: where do we rank on Google?
Now there are five.
- 1Where do we rank on Google?
- 2What does AI Overviews say about our category?
- 3Does ChatGPT name us?
- 4Does Perplexity cite us?
- 5Does Gemini know we exist?
Most businesses are still answering the first and assuming the other four take care of themselves. They don’t — but they’re closer to it than most people think, and the reason why is the most useful thing to understand about this whole shift.
The machines are all reading the same room
There’s a temptation to treat AI search as a separate channel with separate rules. A new algorithm to reverse-engineer, a new set of tricks, a new agency line item.
It isn’t. Every one of these systems has the same underlying problem: it has to decide which businesses are worth putting in front of a person who asked a question. And none of them can visit your office, call your customers, or take your word for it.
So they do what any careful stranger would do. They look at what everyone else says about you.
Google built a twenty-five year business on that principle. Language models — trained on the open web, and increasingly retrieving from it live — arrived at the same place from a different direction. When a model gets asked who’s the best commercial roofer in Phoenix, it isn’t consulting a ranking table. It’s assembling an answer out of everything it has read. If the trade press, the local directories, the industry roundups, the review aggregators and the supplier blogs all mention you, you’re in the answer. If they don’t, you aren’t — no matter how good your site is.
Google relies on input. AI relies on input. The input is largely the same input.
Why “get found” is the wrong pitch
The industry has spent two years selling AI visibility as a discovery problem: customers can’t find you, here’s how to get found.
That framing gets the relationship backwards. These platforms are not reluctant gatekeepers you have to sneak past. They are desperate to recommend good businesses. A model that confidently names the right roofer is a model people come back to. A model that names the wrong one, or hedges into a list of generic advice, is a model people stop trusting.
The platforms want to promote you. What they lack is confidence — corroboration, from sources they already weight, that you are what you claim to be.
That’s the actual job. Not gaming a system. Giving it grounds to back you.
What that looks like in practice
Three things determine whether you turn up in an AI answer, and they’re not exotic.
- Corroboration
- You are mentioned, by name, on sites the model already reads and trusts. Not one mention — a pattern of them, across independent sources, over time. This is the single largest lever and the one most businesses have no deliberate programme for.
- Consistency
- The claims about you line up. Same services, same service area, same specialisms, same name, wherever you appear. Models are pattern-matchers; contradictory inputs make you an unsafe answer, so they route around you.
- Legibility
- Your own site states plainly what you do, who for, and where — in prose a machine can lift, not just in images, sliders and brand copy. A surprising number of companies have never written the sentence they want quoted back to a buyer.
Get those three right and you don’t optimise separately for Google, ChatGPT, Perplexity and whatever ships next year. You compound across all of them, because they are all drinking from the same well.
The uncomfortable part for anyone already buying links
If you’re spending money on backlinks — and plenty of companies are spending $1,200, $5,000, $20,000 a month — you are already paying for the input layer. The question is whether you’re getting the AI-side return on it, and almost nobody can answer that, because almost nobody is measuring it. If you are also spending on paid, the same argument applies there — see does PPC spend help you show up in ChatGPT and Perplexity?
Two programmes can produce identical link counts and wildly different AI visibility. It depends on whether the placements are on sources these systems actually read and weight, whether your name appears in a context that says something about what you do, and whether the mentions accumulate into a coherent picture or a scattered one. Link volume tells you nothing about that. Neither does domain authority.
The right diagnostic isn’t “how many links did we build.” It’s: for the twenty questions our buyers actually ask, how often are we in the answer — and is that number moving?
That’s measurable now. You can track citation share across the major AI surfaces the same way you’ve tracked rankings for years. Most companies simply haven’t started, which means most companies are running an expensive programme with a blind spot where half the outcome sits.
Where to start
Pick the questions first. Not keywords — questions, phrased the way a buyer would actually type them into ChatGPT. Twenty is enough. Run them. Write down who gets named.
That baseline usually settles the argument on its own. Either you’re in the answers, in which case you now know what’s working and can do more of it — or you’re absent while three competitors are named, in which case you’ve found the gap and you know roughly what it will take to close.
The businesses that move on this in the next twelve months are going to look, to a machine, like the obvious safe answer in their category. The ones that don’t will be invisible in a channel they never noticed they’d lost.
Read next
Want to know where you stand?
We’ll run your category’s twenty buyer questions across Google, ChatGPT, AI Overviews, Perplexity and Gemini, show you exactly where you’re named and where your competitors are, and tell you what it would take to change that.
30 minutes, no charge, no pitch deck.