What Google's Own Guidance Says About AEO, GEO and the Tools Selling Them

A marketing lead in fintech, web3, tech, or SaaS now receives several pitches a month promising better visibility inside AI answers. They arrive carrying their own marketing speak on answer engine optimisation and generative engine optimisation, and usually a dashboard packaged with score metrics meant to represent how visible a brand is to an AI system. MarcomFintech is a personal brand working across marketing, communications and fintech, and the useful development this quarter is that Google has published something a buyer can hold those pitches against. The page sits on Google Search Central and carries a last-updated date of 5 June 2026.

What Google Has Published

The guidance sits at developers.google.com/search/docs/fundamentals/third-party-seo and covers how to evaluate third-party SEO services, tools and advice. What makes it notable is that Google names the new vocabulary itself, referring to third-party advice about search listings and AI experiences by both of its industry labels, answer engine optimisation and generative engine optimisation. Google rarely acknowledges industry coinages inside its own documentation, and naming them is a way of claiming the ground they describe.

Three sentences carry the weight. On tools: "Third-party tools don't have access to our internal ranking data. They can't guarantee performance." On vendors: "Google doesn't evaluate third-party services, so be wary of such claims and those making them." On advice in general, the test Google offers is that good advice "either qualifies their claims as opinion based on data or experience, or backs up their claims by citing official Google Search guidance."

That last sentence is the practical one. It gives a non-specialist something to apply in a meeting without needing to know how any of the systems work.

Why this is Uncomfortable for Anyone Selling AI Visibility

The point deserves a direct answer rather than a dodge. The founder writing this builds an AI visibility audit tool, LITV AI SEO Agent v2.0, which checks five categories, Technical SEO, SXO, GEO, AEO and AVI, with AVI scored across six AI platforms. Google's sentence about third-party tools applies to that product exactly as it applies to every competitor. No external tool reads Google's ranking systems, and any score one produces is an assessment against published guidance and observable behaviour rather than a reading taken from inside the machine.

The honest position for a vendor is to say what the number actually is. A diagnostic score describes how well a page meets criteria the tool can inspect directly: whether the entity is clear, whether claims are attributable to a named source, whether the structure lets a passage be extracted and cited on its own. Those are checkable statements about the page. A predicted ranking or a promised citation would be a statement about Google's systems, which no third party is in a position to make. A buyer should establish which of the two is on offer.

The Idea That Refused to Wait

It started the way most difficult ideas do: with a problem that did not yet have a product to solve it. Marketers were being told they needed to optimise for AI search engines, GEO, AEO, and a dozen other emerging frameworks, but no tool audited all of it comprehensively at a price point or scope that made sense for a working practitioner.

On 28 February 2026, with VS Code open and Claude Code loaded as a co-pilot, the project began. There was no product brief and no roadmap document, only a blank file, a terminal, and two decades of compacted instinct about what digital practitioners actually need.

Designing the Five-Category Audit Engine

Before a single user-facing pixel was drawn, the logic had to be right. The agent was designed to audit across five distinct AI-search categories: Technical SEO, SXO (Search Experience Optimisation), GEO (Generative Engine Optimisation), AEO (Answer Engine Optimisation), and AVI (AI Visibility Index). Each category needed its own scoring matrix, weighting, and recommendation language.

The coding experience from before the 2010s earned its keep here. Understanding how HTML structures, crawlers, and schema interact is not something anyone prompts their way into overnight, and that muscle memory shaped every architectural decision that followed.

What No Special Optimisation Means

Google's separate documentation on AI features, last updated 10 December 2025, goes further than most vendors would like. It states: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." It adds: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add."

Read quickly, that sounds like the whole category is invented. Read properly, it says something narrower and more demanding. Google is describing the absence of a secret. The work itself does not disappear anywhere in that sentence. The same page directs readers to apply "the same foundational SEO best practices for AI features as you do for Google Search overall", and those fundamentals are precisely where most sites are weakest: crawlability, indexable text, clear structure, and content that answers the question a person actually asked.

This blog argued in July that Technical SEO, GEO and AEO are one discipline rather than three separate checklists. Google's documentation now describes the same position from the other direction. Anyone being sold a distinct AI optimisation programme, priced separately from the technical work it depends on is being sold a division Google does not recognise.

Where Google's Guidance Stops

One limit matters and is easy to miss. Google's documentation governs Google. It says nothing about how ChatGPT, Perplexity, Claude or Copilot select and cite their sources, and those systems carry their own retrieval behaviour and their own preferences, some published and some not. A brand treating Google's guidance as the full specification for AI visibility has mistaken one large platform for the entire surface.

The commercial reason this still matters was covered here in August, when zero-click search reached 68% in 2026. Once the answer arrives without the click, the question that decides whether a brand is seen at all is which sources the answer draws on, and that question resolves differently on each platform.

How to Judge the Next Pitch

Four questions, all answerable in a first call, and all drawn from what Google has now put in writing.

Ask what the score measures. A page-level assessment against published criteria is a legitimate product. A predicted ranking or a guaranteed citation is something else, and Google's sentence about internal ranking data is the reason to say so.

Ask which platforms are covered and how each one is checked. Google, ChatGPT, Perplexity and the rest behave differently, and a single blended number covering all of them conceals more than it reports.

Ask for the source behind each recommendation. Google's own test is the right one to borrow here: is the claim qualified as opinion drawn from data or experience, or is it backed by official guidance.

Ask what happens if the technical fundamentals are broken. A proposal for AI visibility that does not start with crawlability, indexation and content quality is selling candies rather than the shop that sells candies.

FAQs

Does Google's guidance mean GEO and AEO are not real disciplines? No. It means Google does not treat them as separate from search optimisation on its own surfaces, and that no third party can claim inside knowledge of its ranking systems. The underlying work of making a page clear, attributable and extractable remains real, and it matters more on platforms outside Google that publish far less about how they choose sources.

Should we remove llms.txt from our site? Not necessarily, though nobody should sell it as a Google ranking factor. Google's documentation states directly that AI text files are not needed to appear in AI Overviews or AI Mode. The file costs almost nothing to maintain and other systems may make use of it, so keeping it is a reasonable low-cost hedge.

Can any tool tell us whether ChatGPT or Perplexity cites our brand? A tool can query those systems and record what they return, which is genuine observed evidence. What no tool can do is explain why a source was chosen or guarantee the result repeats, since retrieval varies by phrasing, timing and user context. Treat the output as a sample rather than a score.

What should a marketing budget for AI visibility actually cover? The same technical and editorial fundamentals that serve traditional search, done properly, plus measurement across the platforms that matter to the business. If a line item cannot be traced back to a fundamental or to a measurement, ask what it buys.

Being pitched AI visibility work and unsure which parts are real? Reach out at hello@marcomfin.tech.