Somewhere in a Christchurch or Auckland inbox this week is a cold email offering a new service called GEO, or possibly AEO, promising to get a business cited inside Google's AI answers before competitors catch on. The pitch usually arrives with urgency attached. Search is changing, the old rules are dead, and the businesses that move first on generative engine optimisation will own the new landscape while everyone else disappears.
In late June 2026, Google published a document that quietly took the ground out from under most of that pitch. The official guide to optimising for generative AI features on Google Search is the search team's own account of what actually influences whether a page shows up in AI Overviews and AI Mode. Its central message is short. Optimising for generative AI search is optimising for search, and that is still SEO. The acronyms are new. The work is not.
This matters more in New Zealand than the timing suggests, because the AI features are already here. AI Mode launched in New Zealand in August 2025, AI Overviews have been running locally for over a year, and industry trackers estimate the overviews now appear on a large share of Google searches. Kiwi businesses are being asked to pay for a solution to a problem, at exactly the moment the platform behind that problem has published the answer for free. This post walks through what the guide actually says, what it tells you to stop worrying about, and what a New Zealand business should genuinely do about search in the age of AI answers.
What Google actually published
Google's generative AI guide is a reframing of existing search best practice, not a new rulebook. Generative AI search optimisation is the work of making a website genuinely useful and technically clean so that Google's ranking systems, which now feed its AI features, can find, trust, and surface the content. It works because AI Overviews and AI Mode are built on the same core ranking and quality systems that have always driven Google Search, not on a separate machine that rewards different behaviour. For a New Zealand business, the practical implication is that money spent on foundational search quality serves both traditional results and AI answers at once, while money spent on AI-specific tricks tends to serve neither.
The guide explains the two techniques underneath the AI features in plain terms. The first is retrieval-augmented generation, also called grounding, where the model relies on Google's ranking systems to pull relevant, current pages from the index and then generates a response supported by clickable links to those pages. The second is query fan-out, where the model quietly generates a set of related searches around the original question and gathers results from across them. Ask how to fix a lawn full of weeds and the system may also search for the best herbicides and how to prevent weeds returning, then assemble an answer from the spread.
Both techniques reward the same thing. Pages that are genuinely relevant, genuinely useful, and technically accessible get retrieved and cited. There is no separate door into the AI answer that bypasses the quality bar. That single fact reshapes how a business should read every GEO pitch that follows.
Why "AEO and GEO are just SEO" is the line that matters
Google addressed the acronyms directly, which it rarely does. Answer engine optimisation and generative engine optimisation are terms used to describe work focused on visibility in AI search experiences, and from Google's perspective, that work is still search engine optimisation. The guide points readers who are considering third-party AEO or GEO services toward its guidance on evaluating third-party SEO advice, which is about as pointed as Google's documentation gets.
For a New Zealand business owner weighing up a proposal, this is the sentence that should shape the budget conversation. A local agency offering a GEO retainer is not necessarily selling something dishonest. Search visibility work genuinely does help a business appear in AI answers, because the two run on the same systems. The question to ask is narrower and more useful. Is this a real search programme with a new label on the cover, or is it a set of AI-specific tactics that Google has said do not do anything? The difference decides whether the spend compounds or evaporates.
The distinction is easy to test. Ask any provider what they would actually do. If the answer is content quality, technical health, genuine expertise, and measurement, that is search work worth paying for, whatever it is called. If the answer leans on special files, machine-only markup, or engineered mentions, the next section is the reason to be sceptical.
The tactics Google told you to ignore
The most useful part of the guide for a cautious buyer is the section on what does not work. Google listed several popular generative AI tactics and said plainly that they carry no weight in Google Search.
The first is the llms.txt file, along with other special machine-readable files created specifically to feed AI. Google does not use them. Creating and maintaining one will neither help nor harm visibility in Google Search, because the system ignores it. A business is free to keep one for other tools, but it should not appear as a line item on a Google-focused invoice.
The second is chunking, the practice of breaking content into small pieces on the theory that AI understands fragments better. Google's systems handle multiple topics on a single page and surface the relevant part without help. There is no ideal page length, and pages should be built for the reader rather than reshaped for a machine. The third is rewriting content specifically for AI systems, which is unnecessary because the models already understand synonyms and intent, so a business does not need to capture every phrasing a person might use. The fourth is the pursuit of inauthentic mentions across the web, which Google's core ranking and spam systems are built to see through. The fifth is over-focusing on structured data, which is not required for generative AI search, though it remains worthwhile for rich results in traditional search.
Read together, these five items describe a large share of what gets sold as GEO in the current market. That is the value of a primary source. When the platform itself publishes the list of things that do not move the needle, a New Zealand business gains a clean way to separate a credible proposal from an expensive one.
What actually earns a place in AI answers
Stripping out the tactics that do nothing leaves the work that does, and Google is unambiguous about what that is. The single strongest influence on long-term visibility in generative AI search is unique, non-commodity content made for a real audience.
Google draws a sharp line between commodity and non-commodity content, and the examples are worth borrowing. Commodity content is the "7 Tips for First-Time Homebuyers" article, built from common knowledge that could have come from anyone and adds little a reader could not find in a hundred other places. Non-commodity content is the piece titled something closer to "why we waived the building inspection and what it cost us", grounded in first-hand experience and a point of view that a generative model cannot manufacture from the rest of the internet. The AI systems look across many sources, so the content that stands out is the content that says something the other sources do not.
This is where New Zealand businesses hold an advantage they routinely undersell. A Hawke's Bay vehicle dealer knows things about financing a used car in a regional New Zealand market that no global content mill can assemble. A Christchurch clinic understands the specific consenting and privacy obligations of operating under the Privacy Act 2020 and the Health and Disability Commissioner in a way that generic health content never will. First-hand local knowledge is exactly the non-commodity signal the guide describes, and most Kiwi operators have far more of it than they publish.
The second pillar is a clear technical structure. To appear in AI features, a page must first be indexed and eligible to appear in Google Search with a snippet, which means the ordinary technical requirements still decide the ceiling. Content needs to be crawlable, since the models learn from publicly accessible pages. Semantic HTML helps, though Google is relaxed about imperfect code and focused on human readability. JavaScript-heavy sites need to follow JavaScript SEO practice so their content is not hidden from the crawler. Page experience, reduced duplicate content, and verification in Search Console round out a foundation that has not changed in character, only in stakes.
The uniquely New Zealand read on all of this
The guide is written for a global audience, so the local translation is where a New Zealand business should concentrate. Three things follow from the New Zealand context specifically.
The first is that thin content loses harder in a small market. New Zealand search volumes are smaller than the United States or the United Kingdom, which means AI answers here draw from a shallower pool of genuinely local sources. A business with real first-hand content about operating in Auckland, Wellington, or Ōtautahi is competing against fewer credible local pages for citation, not more. The scarcity works in favour of anyone willing to write from actual experience.
The second is that local entities are a trust signal the models can read. Referencing the Privacy Act 2020, the Commerce Commission, Consumer Guarantees Act obligations, Inland Revenue for GST questions, or Stats NZ for a genuine figure does two things at once. It sharpens local search relevance, and it signals the kind of grounded, specific expertise that both readers and AI systems increasingly reward. Vague content that could describe a business anywhere is precisely what the non-commodity standard filters out.
The third is that the New Zealand market is currently rich in GEO sales pressure and poor in primary-source literacy. The businesses that read what Google actually published, rather than what a vendor deck claims Google wants, will make calmer and cheaper decisions. That is not a technical advantage. It is a judgement advantage, and in a small business economy it compounds.
Where this sits in Launch, Grow, and Scale
Search visibility is a Grow-phase concern. It belongs to the stage where a business already has a working presence and now needs activity to turn into measurable progress rather than guesswork across scattered channels. Sant offers search visibility, SEO, and answer engine optimisation as a Grow service, and the position Sant takes on it is the same one the Google guide arrives at from the other direction.
Sant sells GEO, and Sant will tell a client that GEO is not a separate discipline with its own secret tactics. It is disciplined search work: content grounded in genuine expertise, a clean technical foundation, honest measurement, and the patience to let authority accumulate. The Sant litmus test, "will this still make sense in twelve months", disqualifies most of the tactics Google listed as ineffective. An llms.txt file bolted on to chase an AI trend does not survive that question. A body of first-hand, genuinely useful content about a business's actual field does, because it keeps working regardless of which acronym the market is selling next quarter.
For a New Zealand business deciding what to do about AI search, the honest recommendation is unglamorous. Do not buy a GEO product built around the tactics Google has already dismissed. Invest instead in the search fundamentals that feed both traditional results and AI answers, and make sure the content carries expertise a machine cannot fake. That is slower than a silver bullet, and it is the only approach the platform itself endorses.
Frequently asked questions
What is the difference between SEO, AEO, and GEO?
SEO is search engine optimisation, the long-standing practice of making a website findable and useful for search engines. AEO stands for answer engine optimisation and GEO for generative engine optimisation, both newer terms for improving visibility inside AI-driven search experiences such as AI Overviews and AI Mode. In its 2026 guide, Google stated directly that optimising for generative AI search is still SEO, because the AI features run on the same core ranking systems as regular search. The three acronyms describe the same underlying work, not three separate services.
Do New Zealand businesses need to do anything special to appear in Google's AI answers?
No special technique is required. A page becomes eligible for AI Overviews and AI Mode by being indexed and eligible to appear in normal Google Search with a snippet, which means ordinary technical health and genuinely useful content are what decide visibility. The most effective step for a New Zealand business is to publish first-hand, non-commodity content grounded in real local expertise, since that is the signal Google's systems reward and the one generic competitors cannot replicate.
Is it worth paying an agency for GEO or AEO services?
It depends entirely on what the service actually involves. If the work is content quality, technical SEO, real subject-matter expertise, and measurement, it is worthwhile search work regardless of the label. If it centres on tactics Google has said carry no weight, such as llms.txt files, chunking content for AI, or engineering mentions across the web, the spend is unlikely to return anything in Google Search. Ask any provider to describe the specific actions they would take, then check that list against what Google's guide says works.
Does an llms.txt file help with Google Search?
Google's guide states that it does not use llms.txt files or other special machine-readable files for Search, including its generative AI features. Maintaining one will neither help nor harm rankings in Google Search. A business may still choose to keep one for other AI tools and services that do read it, but it should not be presented as a way to improve Google visibility.
How is generative AI search different in New Zealand compared to larger markets?
The mechanics are the same, but the competitive context differs. New Zealand has smaller search volumes and a shallower pool of genuinely local content, so businesses that publish real, first-hand material about operating in New Zealand compete for AI citation against fewer credible local sources. Referencing local entities such as the Privacy Act 2020, the Commerce Commission, or Stats NZ strengthens both local relevance and the expertise signal that AI systems weight. In a smaller market, thin content is easier to beat and genuine local knowledge counts for more.
Google's generative AI guide is not the disruption the GEO sales cycle needs it to be. It is a quiet confirmation that the durable way to appear in AI answers is the durable way to appear in search, which is to be genuinely useful and technically sound. For New Zealand businesses, the opportunity is not to chase a new acronym before competitors do. It is to read the primary source, ignore the tactics the platform has already dismissed, and invest in expertise that neither a competitor nor a language model can counterfeit. The businesses that do that will still make sense in twelve months, which is the only test that has ever mattered.
If you want a clear-eyed view of what actually earns visibility in AI search rather than a pitch built around the tactics Google has ruled out, Sant's approach to search visibility and GEO and the methodology behind it are where that conversation starts.