Google publishes an SEO Starter Guide inside Search Console. It is short, it is free, and it is the closest thing this industry has to an official rulebook.
It is worth reading twice. Once for what it says, and once for what it never mentions.
What it endorses is unglamorous and correct
Strip out the reassurance and the guide asks for a specific list of things. Let crawlers reach your pages and see what a visitor sees. Write content that is genuinely useful rather than a rehash of someone else's. Keep one URL per piece of content, and canonicalize when you cannot. Write titles that are unique to the page and accurately describe it. Use structured data so your pages become eligible for richer result types. Link out with anchor text that says where the link goes. Describe your images properly.
That is a structural checklist. Almost none of it is about keywords. All of it is about whether a machine can reach your content, read it, and understand what it is looking at.
One detail is worth pulling out for local businesses. The guide explicitly says a good title can carry the name of the business and its physical location. Geography in a title is not a trick or a growth hack. It is documented advice from the search engine itself.
What it quietly retires
The guide also has a section on what not to spend energy on, and it kills several things that are still actively being sold.
- The keywords meta tag does nothing.
- E-E-A-T is not a ranking factor. Google states that plainly.
- Duplicate content is not a penalty. It is inefficient, not punitive, and it will not trigger a manual action.
- There is no ideal word count, minimum or maximum.
- Heading order matters for screen readers, not for ranking.
- Keywords in the domain name or the URL path barely register.
If a proposal leans on any of those as the reason to hire someone, the proposal is arguing against the documentation it claims to follow.
The word that never appears
Now the absence.
Across the entire guide there is nothing about being cited by ChatGPT. Nothing about Perplexity. Nothing about Claude. Nothing about how a business becomes the source an assistant reaches for when a buyer asks it who to hire.
The natural assumption is that Google has not gotten to it yet, and that a future revision will cover it. That assumption is wrong, and planning around it wastes the window.
Why the gap is structural rather than temporary
Google documents Google. The only generative surface it has any reason to write about is AI Overviews, because AI Overviews is its own product. It has no incentive whatsoever to publish a playbook for being cited well inside a competing answer engine.
Run that logic across the rest of the field and the vacuum becomes obvious. The organizations best positioned to write the authoritative guide to answer engine optimization are the answer engines. Every one of them is disqualified by the same conflict of interest. OpenAI is not going to explain how to get recommended by Perplexity. Perplexity is not going to document how Claude decides what to cite.
So the guide that would settle these arguments does not exist, and there is no party that has both the knowledge to write it and a reason to.
This is a different problem from the one covered in the piece on telling AEO tools, agencies, and systems apart. That one is about pricing chaos among vendors. This one is about why no authority exists to resolve the chaos, and why none is coming.
What that means when you are buying the work
In a field with no certification, credentials are not available as a filter. That is uncomfortable, but it is also clarifying, because it forces the evaluation back onto method and evidence.
Questions that actually separate practitioners:
- What will you measure, and what does the baseline look like before you start?
- How will we know in ninety days whether this worked?
- What would you fix first on my specific site, and why that first?
- What do you refuse to promise?
That last one carries the most information. Someone who names the limits of their own influence is describing the real shape of the work. Someone who guarantees citations is describing something nobody controls, because the engines decide what to cite and they change their minds on their own schedule.
Measurement is the part most often skipped. If nobody wrote down what the baseline was, no one can tell you later whether anything moved. Tracking AI citations is the unglamorous half of this discipline and the half that makes the rest accountable.
The honest split
Half of this work is documented. Crawlability, structure, canonicals, titles, schema, links. Google wrote it down, it is free, and it is correct. Anyone charging premium fees for that half as though it were secret knowledge is overcharging for public information.
The other half is not documented and will not be. How engines resolve which business an ambiguous query refers to, how much weight independent corroboration carries against self-description, why a retrieval fix lands in weeks while a training-shaped one takes months. There is no manual for that. There is only what practitioners have actually observed, and the discipline to admit which parts are still uncertain.
That second half is where judgment is the product. It is also where the honesty shows, because the people doing it well are the ones willing to say which parts they cannot control.
If the structural half of your site has never been repaired, start there rather than with the interesting half. Entity consistency across surfaces is usually the highest-leverage place to look, and the SEO, AEO and GEO Foundation exists to do that work in one pass.