Three Findings Foragentis Published Before Google's Own AEO Guidance Said the Same

On Friday, May 15, 2026, Google published a guide called "Optimizing your website for generative AI features on Google Search." It expands the AI-features documentation Google had already published in 2025, and it carries no byline: John Mueller announced it on the Google Search Central blog, and the guide itself is unsigned Search Central documentation. What makes it new is the vocabulary. This is the platform's first explicit statement on what does and does not constitute Answer Engine Optimization, Generative Engine Optimization, and adjacent practices. It is, predictably, restrained: a glossary, a reframing of foundational SEO best practices, guidance on local, shopping, image and video content, a section of "mythbusts," a Search Console measurement section, and one brief part on agentic experiences.

Restrained or not, it is the first time Google has named the category and taken a position on what works inside it. That makes it the most consequential single document the AEO/GEO discourse has produced since Aggarwal et al.'s November 2023 academic paper.

We published our own research on this category in late April 2026, three weeks before Google's guidance appeared. The white paper, The State of AEO and GEO in 2026: A Field Report on Answer Engine Optimization, Its Consensus, and What the Data Actually Supports, surveys the discourse, audits the most-cited evidence, and presents proprietary findings from ForIntel research across fifteen verticals. It was produced from primary YouTube transcript analysis, public web content from a dozen vendors, and original SERP and LLM citation research.

Three specific findings from that paper now read as having anticipated Google's official position. We are not claiming we predicted Google. We analyzed independent evidence and reached conclusions that align with what Google has now confirmed. The chronology is the chronology, but the more substantive point is the methodology: the same questions, examined honestly and with adversarial review, reach the same conclusions whether you sit at Google or at an independent research practice in Sacramento.

Here are the three alignments, with the receipts.

Alignment 1: "AEO" and "GEO" are contested vocabulary, not separate disciplines

What Google said (May 15, 2026):

"AEO" stands for "answer engine optimization" and "GEO" for "generative engine optimization". These are both terms you may see used to describe work specifically focused on improving visibility in AI search experiences. From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.

What our white paper said (late April 2026):

The paper's Part 1.1, Taxonomic Inventory, catalogs five terms: SEO, AEO, GEO, LLMO, AIO, and analyzes their practical overlap. The functional distinctions are thinner than the acronym proliferation suggests. Most of the foundational practices each acronym claims are already present in the others. The genuine points of divergence are at the newest surfaces: LLM citation and conversational AI search.

We did not pick a side in the SEO-vs-AEO/GEO debate. We treated it as a discourse phenomenon worth describing accurately, with the editorial framing that a serious operator should focus on the work and let the vendor community fight over the naming. Google has now resolved that debate from the platform's perspective, and the resolution is the same one a careful reader of the field would have reached: the practices are substantially continuous with SEO, the new surfaces matter, and the acronyms are mostly marketing.

The chart we published in the paper (Figure 1) is captioned to say that the five acronyms actually diverge at the newest surfaces, LLM citation and conversational AI search, and not at the traditional SEO practices they all claim. That figure is the visual statement of the same claim Google has now made in prose: this is mostly SEO with new surface area. (The figure appears in the PDF edition of the paper, not in the web edition.)

Why this alignment matters for operators: Vendors who have invested heavily in branding their AEO/GEO offering as a distinct discipline will face a harder sales conversation with sophisticated buyers in the weeks ahead. The buyer who has read Google's guidance will ask, reasonably, what the vendor is doing that traditional SEO investment does not already cover. Operators who have framed their work as "still SEO, with attention to new surfaces," the framing our paper recommends, will not face that friction.

Alignment 2: Brand authority is a proxy for an upstream variable, not a directly controllable lever

What Google said:

Seeking inauthentic "mentions" across the web isn't as helpful as it might seem. Our core ranking systems focus on high-quality content while other systems block spam. Our generative AI features depend on both.

What our white paper said:

Finding E of the white paper analyzed the relationship between domain authority and LLM citation, drawing on ForIntel research that tested 720 query-model probes across four leading answer engines. The headline finding was that domain rating shows a large effect size (Cohen's d = 1.12) as a correlate of LLM citation, while raw referring-domain count shows a negligible effect (d = 0.09). Read that number with the limit the paper puts on it. Only seven distinct cited domains produced it, measured against a non-cited comparison pool of 458, which is too few for any inferential test. The paper is explicit that d = 1.12 is a descriptive effect size and not a tested predictor, and that a reader who treats it as predictor strength has misread the analysis. The paper's counter-signal block carries the second warning:

Correlation is not causation. Large effect sizes describe how strongly domain rating co-occurs with LLM citation. They do not prove domain rating causes citation. The signal may be a proxy for a latent variable, brand recognition already embedded in the LLM's training data, and the practical implication for an operator may be closer to "the LLM already knows your brand from its training" than to "building domain rating will cause future citations."

Translated: the brands that get cited by LLMs are largely the brands the LLMs already know about from their training data. Domain authority correlates with citation because both correlate with the underlying fact that the brand is widely recognized. Operators who treat the correlation as a direct causal lever, who buy backlinks, seed inauthentic mentions, or invest in PR firms whose value proposition is "we will get you cited", are pulling on a string that may not be connected to anything.

Google's mythbust on "seeking inauthentic mentions" is the same warning, stated from the platform side. Their version is operational: don't buy mentions, our systems will catch you. Our version is structural: even authentic mentions may not produce the citation lift you expect, because the upstream variable driving both is brand recognition already embedded in the training corpus.

The two warnings are complementary. Google is telling operators what not to do. We are explaining why the obvious thing not to do is also unlikely to work even if you could do it without getting caught.

Why this alignment matters for operators: A meaningful slice of AEO/GEO consulting advice in 2026 is structured around the implicit claim that domain authority and brand mentions are levers operators can directly influence in commercially relevant timeframes. The honest read of the data, and now of Google's guidance, is that they mostly cannot. Brand-mention density accrues over years through earned coverage, real product launches, podcast appearances, and credible editorial placements. A six-month engagement designed to "build your AEO presence" is unlikely to move the needle on the variable that actually matters.

Alignment 3: Schema markup is not a meaningful AEO lever

What Google said:

Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add. However, it's a good idea to continue using it as part of your overall SEO strategy, as it helps with being eligible for rich results on Google Search.

What our white paper said:

The paper's Part 1.3, Divergence 2: Does schema markup matter for AEO?, treats schema as one of the most over-claimed AEO levers in the vendor playbook. The paragraph, with the passage on the Ahrefs experiment elided for the moment:

A substantial subset of the consensus playbook recommends FAQ schema, Article schema, and Organization schema as core AEO practices. Google's own documentation, however, explicitly states that no special schema is required for AI Overview inclusion, and that the same structured-data practices that serve traditional search are sufficient. [...] The honest read: schema markup is recommended by many and rigorously measured by none we could find. It is not harmful, and we could find no published controlled test establishing its causal role in AI citation.

That paragraph as first published carried one more sentence, citing Ahrefs' Xarumei experiment as evidence that fabricated Reddit and Medium narratives outperformed schema-rich commercial pages. The sentence was wrong about the experiment, and the correction belongs in public.

What Ahrefs actually ran, published by Mateusz Makosiewicz on December 10, 2025, was a narrative-injection test. He built a fictional luxury paperweight company, xarumei.com, on an AI website builder, planted three conflicting invented backstories on a blog, a Reddit AMA and a Medium article, then put 56 questions to eight AI products and watched most of them repeat the invented details. Gemini, Google's AI Mode, Perplexity, Grok and Copilot went with the seeded stories over the brand's own official FAQ, while ChatGPT-4 and ChatGPT-5 cited the FAQ in 84% of answers. His summary: in AI search, the most detailed story wins, even if it is false. The experiment had no schema arm and no comparison against real commercial pages, and it was not a measurement of Google AI Overviews specifically. So what it measures is which source an assistant trusts, and it says nothing either way about schema. Ahrefs' own recommendations at the end of that piece in fact tell brands to add schema markup to their FAQ pages. The web edition of the white paper now carries that correction. The PDF edition, which is the copy sent to anyone who requests the paper by email, still carries the original sentence and has not yet been reissued.

What is left standing, once the Xarumei experiment stops doing work it cannot do, is the absence of evidence. We could find no published controlled test showing that schema markup lifts AI citation. Our paper's recommendation holds on its own terms: schema is worth continuing for traditional SEO reasons (rich results eligibility, structured data benefits in Search Console, accessibility), and operators should not price it as a proven AEO lever. Google's May 15 guidance is the same position, with the same caveat: schema "is a good idea to continue using" but is not "required" or "special" for AI features.

Why this alignment matters for operators: Vendors whose AEO offering is essentially "we will audit and implement your schema markup at a premium" are selling a service whose primary value is traditional-SEO rich-results eligibility rather than AEO citation lift. Buyers should price the engagement accordingly. The Foragentis white paper had reached this conclusion three weeks before Google made it official.

What this means for Foragentis, and for the buyer

These three alignments are not the only places our paper overlaps with Google's guidance. The paper's treatment of unique, non-commodity content as the durable lever, what we call the authority horizon in Part 3.3, aligns directly with Google's emphasis on first-hand, expert-led, people-first content. The paper's three-currencies framework (citation, referral, revenue) is a more operationally useful version of the same caution Google gives in less structured form: do not conflate the distinct things AI features are doing on your site.

The three we have highlighted here are the strongest single alignments: places where the paper named a specific contested practice, took a specific defensible position, and was published before Google made the same statement.

What this should mean for the reader who is choosing a research partner: independent research conducted with proper methodology and adversarial review can reach the same conclusions the platform reaches, without privileged access. The white paper was produced from public discourse analysis, vendor case-study auditing, and original SERP and LLM citation research using commercially available infrastructure. Any operator with sufficient discipline could have done the same work. The fact that few did is the gap Foragentis exists to close.

For verticals not covered in the white paper, and Google's guidance, being Google-only, has nothing to say about ChatGPT, Claude, or Perplexity citation dynamics specifically, ForIntel custom reports start at $649 per vertical. The methodology that produced the alignments above is the methodology that produces those reports.

Read the full white paper: The State of AEO and GEO in 2026

Request a custom vertical report: ForIntel custom reports, starting at $649 per vertical.

Foragentis is an AI research and product company based in Sacramento. ForIntel is its research arm, producing original intelligence on AEO, GEO, and adjacent practices across commercially distinct verticals. For methodology questions, contact forintel@foragentis.com.