Google Published a Way to Vet Your SEO. Use It on Me.

On 5 June 2026 Google published a new page of guidance on using third-party SEO tools, services and advice, and updated its long-standing “Do you need an SEO?” document with a section on how to evaluate what an SEO recommends.

It is the first time Google has given buyers a structured way to check whether the person they are paying is telling them the truth. It also, for the first time, names AEO and GEO explicitly as a service category worth scrutinising.

I sell these services. I am going to walk through the framework anyway, and then apply it to my own work, because a vetting guide you only point at competitors is marketing rather than advice.

The Three Questions

Stripped of diplomacy, Google’s framework asks three things about any SEO recommendation.

1. Does the advice cite official Google documentation? Not a blog post about a tweet about a podcast. The actual source.
2. Do the AI optimisation claims match what Google has actually said? Specifically, is AEO or GEO advice aligned with Google’s own published guidance on generative AI features, or invented?
3. Does the data imply a connection to Google’s internal systems? Because it does not have one.

That third point is the sharpest thing in the document. Google states plainly that it does not evaluate or endorse third-party SEO tools, and that those tools “don’t have access to Google’s internal ranking data”.

Read that against how tool metrics are usually sold to clients and you can see why Google bothered to publish it.

What This Means About Tool Scores

Every third-party authority score, difficulty score, visibility index and optimisation grade is a model built by that vendor from data they collected themselves. None of them read anything from inside Google.

That does not make them worthless. A well-built proxy can correlate usefully with reality and help you prioritise. I use tools daily.

It makes two specific things wrong:

• Presenting a vendor score as though it were a Google metric
• Treating an increase in a vendor score as a result, when nothing measurable happened to traffic, rankings or revenue

If a report shows your “SEO score” rising from 68 to 91 and does not show what happened to traffic or conversions, you have been shown a picture of a tool’s opinion of a page. That is a process metric at best. It is not an outcome, and it should never be the headline of a monthly report.

The AEO and GEO Part Is the Point

Google could have written this guidance in 2019. It wrote it now, and it specifically added a question about AEO and GEO advice, because that is where the unverifiable claims currently live.

The field is roughly two years old, the measurement is thin, and until Search Console added generative AI reporting in June there was almost no first-party data to check anything against. That combination is ideal conditions for confident nonsense.

Things I would ask any AEO or GEO vendor to substantiate:

A competent vendor will answer these comfortably, and will tell you where their confidence ends. Discomfort at being asked is itself the answer.

Turning It On Myself

Applying the same three questions to my own work, honestly:

Do I cite official documentation? Where it exists, yes, and I try to name dates. Where it does not exist (which is much of AI search) I should be labelling that as reasoning rather than fact, and I am not always as explicit about the difference as I should be.

Do my AI claims align with Google’s guidance? On the substance, yes: I have argued against llms.txt on the basis that Google said it does not use it, and against treating AEO as separate from good content structure. Both positions match the published guidance.

Do I imply access to internal data? No, and I would rather lose work than pretend otherwise. Nobody outside Google knows the ranking systems. Anyone who says they do is selling certainty, which is a much easier product than SEO.

The uncomfortable one is the first. The honest framing for a great deal of AI search advice, mine included, is “this is consistent with how these systems appear to work and with what Google has said, and it has not been independently verified”. That is less sellable than a framework with a name. It is also true.

Using This as a Buyer

Four things to do with the framework, in order of how much money they save you.

1. Ask for the source on any specific claim. “Where did that come from?” is a complete audit in five words.
2. Check that reports lead with outcomes. Traffic, rankings, conversions, revenue. Tool scores belong in an appendix.
3. Treat guaranteed rankings as disqualifying. Nobody controls the ranking systems. This has been true for twenty years and Google repeats it in the hiring guide.
4. Ask where their advice stops. Anyone who cannot name the limits of their own method has not reached them yet.

Why Google Bothered

Publishing a vetting framework is a slightly unusual move, and the likely reason is straightforward: bad SEO advice creates support burden, damages trust in search, and produces sites that make results worse. Arming buyers is cheaper than policing vendors.

The side effect is good for anyone doing this properly. A framework that separates verifiable advice from invented advice raises the floor, and the practitioners who lose from it are the ones who were relying on you not asking.

Ask.

Written by Shahzaib Ul Hassan, senior AI SEO consultant and founder of ShazzSEO. Ranking sites since 2009. 500+ websites optimized, 3,000+ students trained.