On June 5, 2026, Google did two things in the same documentation update. It published a new page, "Google Search's guidance on using third-party SEO tools, services, and advice," and it updated the long-standing "Do you need an SEO?" page with new questions specifically about evaluating AI-search-focused advice. Together, these amount to Google formally naming AEO and GEO as recognisable categories of SEO service — three weeks after the May 15 guide that some read as dismissing the entire premise of AEO and GEO as distinct disciplines.
The two documents are not in tension, and reading them together is more useful than reading either alone. The May 15 guide told site owners what Google's own systems do and do not require. The June 5 update tells site owners, in Google's own words, how to tell a credible AEO or GEO practitioner from a vendor selling vapor. That second document is the more immediately useful one if you are currently paying, or considering paying, anyone for AI-search optimization work.
What Question Does Google Now Tell You to Ask Any SEO Vendor?
The updated "Do you need an SEO?" page adds a specific, quotable question to the standard vendor-evaluation checklist: "If an SEO has advice on optimizing for AI experiences (also known as 'AEO' or 'GEO' services), is their advice aligned with Google Search's official guidance on optimizing for generative AI features?" This is a direct instruction to cross-reference any AEO pitch against the May 15 guide covered in the companion article in this series, and it puts the burden of proof explicitly on the vendor, not on you as the buyer trying to interpret an unfamiliar acronym.
The practical application is straightforward and takes about ten minutes per vendor claim. When an agency or tool pitches a specific AEO tactic, ask them directly which part of Google's own documentation supports it. If the pitch is selling llms.txt generation, content chunking services, or proprietary AI-specific schema as a Google-visibility lever specifically, the May 15 guide already tells you Google does not confirm any special treatment for these — the vendor is either unaware of Google's own stated position or is deliberately not mentioning it.
What Is the Difference Between a Claim Grounded in Documentation and a Proprietary Ranking Factor?
This is the sharpest and most useful distinction in Google's framework. A credible vendor grounds specific claims in Google's own published documentation at developers.google.com/search, not in a proprietary "AI ranking factors" model that nobody outside the vendor's organisation can independently verify. If a pitch cannot point to a specific Google page behind a given recommendation, the correct posture is to treat that recommendation as the vendor's opinion or their own research finding, not as an established fact about how Google's systems work.
This distinction does not mean independent research and correlation data have no value — the extensive body of multi-engine citation research cited throughout this series is genuinely useful and, in most cases, transparent about its own methodology and sample size. The distinction Google is drawing is narrower and more specific: a vendor claim about how Google specifically ranks or selects content for AI Overviews should be checked against what Google itself says, not accepted purely on a vendor's authority. A vendor claim about how Perplexity or ChatGPT behaves is a different category entirely, since Google's own documentation makes no claims about competitor systems at all, and independent third-party research is the only evidence available for those engines.
How Should You Evaluate an "AI Visibility Score" or Similar Proprietary Metric?
Google's guidance includes a specific, pointed caution here: no external tool has access to Google's internal ranking data, so any tool reporting a proprietary "AI visibility score" is producing a model or estimate, not a direct measurement of Google's actual behaviour. The right question to ask a vendor selling this kind of score is what underlying data it is built on, and whether the vendor presents the number honestly as an estimate rather than implying it reflects verified, ground-truth ranking information Google does not actually disclose to anyone outside the company.
This does not mean citation tracking and visibility monitoring tools are without value — quite the opposite, since the entire measurement discipline covered in the AEO measurement guide in this series depends on exactly this kind of tracking. The distinction is between a tool that transparently reports what it directly observed — "your brand was cited by ChatGPT on this exact prompt, on this date, with this exact wording" — and a tool that produces an opaque composite score with no disclosed methodology, implying a precision about Google's internal systems that no outside party can actually possess. Verifiable, observation-based tracking passes Google's own bar for credibility. A black-box proprietary score generally does not.
What Does This Mean for AEO Agencies and Consultants Specifically?
For an agency or freelance consultant offering AEO or GEO services, Google's June 5 update is, on balance, a validating development rather than a threat — it formally acknowledges these as legitimate service categories worth naming, which several agencies had been hesitant to do given the earlier ambiguity about whether Google considered the terms meaningful at all. The AEO for agencies guide in this series covers the operational side of running this kind of practice; Google's June 5 documentation is now the correct, citable authority to point a skeptical prospective client toward when they ask "is this a real thing, or is it a rebrand of SEO to justify a new invoice line item?"
The practical implication for how an agency should pitch its own work going forward: every specific technical recommendation should be traceable either to Google's own documentation (for claims about Google's AI Overview and AI Mode specifically) or to a named, methodologically transparent independent study (for claims about Perplexity, ChatGPT, Claude, or Gemini). A pitch deck that blends both categories without distinguishing them is exactly the kind of pitch Google's new framework is designed to help a buyer catch and question.
How Do You Apply This Framework to Your Own Internal AEO Programme?
The same accountability standard applies whether the "vendor" is an external agency or your own internal team running an AEO initiative. Before your next quarterly AEO planning cycle, take your current tactic list and sort each item into one of three buckets: confirmed by Google's own May 15 documentation (foundational technical SEO, content quality, avoiding spam), supported by independent, transparent third-party research for a specific non-Google engine (schema stacking for Perplexity, entity consistency for Claude, and similar), or unverified assumption carried over from an older playbook with no clear source behind it. Most teams that run this exercise honestly find a meaningful share of their current activity sits in the third bucket, and Google's own framework gives you the language to challenge it internally without needing to invent your own audit criteria from scratch.
How NotionCue Aligns With Google's Verifiability Standard by Design
Google's core critique of low-credibility AEO tooling is the opaque, unverifiable proprietary score — a number with no visible methodology standing in for actual observed behaviour. The NotionCue Citation Tracker is built the opposite way: every data point is a direct, dated observation of what a specific AI engine actually said in response to a specific tracked prompt, with the full response text captured alongside it. There is no black-box "visibility score" standing between you and the underlying evidence — you can read exactly what ChatGPT, Perplexity, Claude, Gemini, or Google's AI surfaces said, on which date, in response to which query, and judge the significance of that yourself.
This matters directly for the vendor-evaluation exercise Google is now encouraging. If you are assessing whether an AEO investment — internal or external — is working, the correct evidence is the kind the NotionCue Prompt Tracker produces: a verifiable, week-over-week record of citation presence and citation text across named prompts and named engines, not a composite index number nobody outside the tool's own team can audit.
Start your free NotionCue trial and use the underlying prompt-and-response data, not a summary score, the next time you need to demonstrate to a skeptical stakeholder — or to yourself — that an AEO investment is producing observable, checkable results rather than an unfalsifiable claim.
Google's June 5 update also revised the broader "Do you need an SEO?" page to remove several outdated examples and simplify its language generally — a reminder that this entire category of documentation is actively maintained rather than static. Bookmark developers.google.com/search/docs and check it directly before renewing any significant AEO vendor contract, rather than relying on secondhand blog summaries (including this one) as a permanent substitute for the primary source.
Frequently Asked Questions About Vetting AEO Vendors
Does Google's naming of AEO and GEO as legitimate services mean it endorses any specific vendor or tool?
No. Google's June 5 update names the service categories as real and worth evaluating carefully; it does not endorse, certify, or recommend any specific agency, consultant, or software tool. The framework it provides is explicitly a set of questions for the buyer to ask, not a seal of approval any vendor can claim to have received from Google.
What if a vendor's advice contradicts Google's May 15 guide but seems to work for other AI engines?
This is not automatically disqualifying, and it is one of the more nuanced parts of applying Google's framework correctly. Google's guidance describes Google's own systems specifically. A vendor recommending schema markup, for instance, is not wrong simply because Google's guide says schema is not required for its own AI Overview inclusion — the vendor should be transparent, however, that the recommendation is aimed at Perplexity, ChatGPT, or another non-Google engine, and should be able to point to the independent research supporting that specific claim rather than implying it is also a confirmed Google ranking factor.
Should I stop working with a vendor who cannot cite Google documentation for every recommendation?
Not necessarily, but you should ask why. A capable AEO practitioner may have well-founded, independently-researched recommendations for engines Google's own documentation does not address at all, since Google's guidance has zero jurisdiction over how Perplexity or Claude select citations. The red flag Google's framework is designed to help you spot is a vendor implying Google-specific authority for a claim that is actually about a different engine entirely, or presenting an unverifiable proprietary metric as though it were an objective measurement of Google's internal systems.