Generative engine optimisation is being sold as brand new. It mostly isn't. What gets a brand recommended by ChatGPT and AI search, and what to do first.
Every agency deck this year has a GEO slide. Generative engine optimisation, AI search optimisation, answer engine optimisation: pick your acronym. The pitch is that AI has changed search so completely that you need a whole new discipline, and probably a whole new retainer.
Some of that is true. Most of it is SEO with a new scoreboard.
The shift is real
People increasingly ask ChatGPT, Gemini, Perplexity or Google's AI answers instead of scrolling through ten blue links. Ask an assistant for a hotel in Woolloomooloo, or the best business account for paying overseas suppliers, and it gives you a short list with a reason for each.
If you are not on that list, you are not in the consideration set. And your analytics will not tell you, because nobody clicked.
That blind spot is the genuinely new part. Traditional search leaves a trail of impressions and clicks. AI answers mostly don't.
Mentioned is not recommended
The metric most tools report is whether your brand is mentioned. That is the wrong bar. Plenty of brands are mentioned in an answer as background, or as the option the assistant suggests people avoid. What matters is whether you are recommended when someone asks the question your customer asks.
So test it the way a customer would. Write down the ten questions a buyer asks before they choose you. Ask them in two or three assistants. Note who gets recommended, and the reason the assistant gives. That reason is your brief.
What actually moves it
When we dig into why one brand is recommended and another is not, the answers are rarely exotic:
- The basics are done. Pages are crawlable and fast, and the content is in the page rather than loaded later by scripts. If a crawler cannot read your page, neither can a model.
- The facts are on the page. Prices, locations, what you do and who it is for, in plain words and structured data, not buried in a PDF or an image.
- Other people talk about you. Assistants lean heavily on third-party sources: reviews, comparison articles, partner pages, press, directories and forums. A brand that only talks about itself on its own site is easy to overlook.
- Content answers real questions. Not AI-written filler, which the assistants are getting better at ignoring, but genuinely useful answers to the questions people ask.
Every one of those is good SEO. That is the point.
A word on llms.txt
There is a lot of noise about adding an llms.txt file to your site. It is not a standard yet, and adding one is cheap and harmless. It is a sign a team is paying attention. It is not a strategy. If your pages cannot be read, a text file pointing at them will not save you.
The better bet is to control your own web layer: pages you can change quickly, with content and structure you own. Then whichever convention wins, you can adapt in a day rather than a quarter.
Where to start
- Fix the fundamentals. Technical SEO and content first. AI visibility work on a site that cannot be crawled is wasted money.
- Audit your AI visibility. By question, by assistant and by competitor: recommended, mentioned or absent.
- Close the third-party gap. Partner pages, reviews, PR and comparisons that point to your most valuable pages.
- Measure it separately. Track AI visibility as its own number, because your normal analytics will not see it.
- Turn it into a plan. A list of fixes with owners and dates beats a GEO-themed slide deck every time.
Summing up
AI search is real and growing, and the brands that ignore it will feel it. But you do not need to throw out what works. Get the fundamentals right, find out whether you are actually recommended, and earn your place on the list the way good brands always have: by being the clearest, most useful answer.