Google's SEO starter guide is right, and it is answering one question

Google publishes an SEO starter guide, and the most useful part of it is a section headed “Things we believe you shouldn’t focus on”. Every item on that list is correct. It is also a list about ranking, and three of its entries still decide whether a machine that is summarising your page can use it.

What the document actually says

It is worth being precise about the guide’s own framing, because most summaries of it are more assertive than the original. It opens by telling you there are no secrets in it that will rank your site first, notes that some of its suggestions might not apply to your business at all, and observes that you usually do not need to do anything except publish your site on the web.

On timing it says that some changes might take effect in a few hours and others could take several months, and that you probably want to wait a few weeks before deciding whether your work had any beneficial effect. That range is wider than almost any agency will quote you, and it comes from the people who run the index.

And it puts the weight where the industry mostly does not: creating content people find compelling and useful, it says, will likely influence your presence in results more than any of the guide’s other suggestions.

That is the whole document, honestly summarised. It is short, it is accurate, and it is free.

The list of things not to focus on

The section is the most quoted part of the guide and the least read past the headings. Verbatim, on the items that matter here, read from developers.google.com on 20 August 2026:

• The keywords meta tag. “Google Search doesn’t use the keywords meta tag.”
• Content length. “The length of the content alone doesn’t matter for ranking purposes (there’s no magical word count target, minimum or maximum, though you probably want to have at least one word).”
• Keywords in the domain or URL. “The keywords in the name of the domain (or URL path) alone have hardly any effect beyond appearing in breadcrumbs.”
• Subdomains versus subdirectories. “Do whatever makes sense for your business.”
• Number and order of headings. “Having your headings in semantic order is fantastic for screen readers, but from Google Search perspective, it doesn’t matter if you’re using them out of order.”
• E-E-A-T as a ranking factor. “No, it’s not.”

Every one of those is true, and several of them have been true for longer than most people working in this field have been doing it. The keywords meta tag has been ignored since 2009. There has never been a word count target. There is no duplicate content penalty, only wasted crawling.

The question the list is answering

The guide is a document about ranking in Google Search. Read as an answer to “what makes this page rank”, the list is right on every entry.

The reason it needs reading carefully in 2026 is that ranking is no longer the only outcome a page is competing for. When a query gets answered in place, by an AI Overview or by an assistant, the thing being contested is whether a passage of your page is selected and whether your name appears next to it. That is a different mechanism with different inputs, and Google’s guide is not written about it, does not claim to be, and is not wrong about it.

Three items on that list are load bearing in the second mechanism while being genuinely irrelevant to the first.

Heading order

Google’s statement is about ranking and it is accurate: out of order headings do not hurt you in Search.

What headings do is divide a document into passages. Anything that has to quote part of a page rather than link to the whole of it has to first decide where the parts are, and the heading outline is the most reliable structure available for that. A heading level that skips a rank describes a subsection belonging to a parent that does not exist, and an empty heading tag is a boundary with no label on it. Neither costs a position. Both make the page harder to cut cleanly.

This one is also worth attention for a reason that has nothing to do with argument: it is the most commonly broken thing we measure. In our audit of 98 agency homepages, heading problems appeared on 64% of the sites we could fetch, more than any other on-page item, and those are professional marketing sites. The full breakdown of what fails and what does not is in the on-page piece.

So the practical position is that Google is right and you should fix them anyway, at a cost of about ten minutes per template.

E-E-A-T

“Thinking E-E-A-T is a ranking factor: No, it’s not.” This has been said repeatedly by Google in several forms and the industry keeps not hearing it. E-E-A-T describes qualities human raters use when assessing whether results are good. It is not a score computed for your page.

What is separately, mechanically true is that a system summarising a page has to decide whether to name a source, and a page with no identifiable author or organisation in its markup makes that decision easy in the wrong direction. There is nothing to attribute to.

This is not E-E-A-T through a side door. It is a different mechanism that happens to reward one of the same behaviours, and the distinction matters because it tells you what to do. You do not do “E-E-A-T work”. You put a real author and a real organisation into the markup, once, in a template, and then it is done.

Content length

“No magical word count target, minimum or maximum” is correct and is routinely misread in both directions. It does not mean length is free of consequences. It means length is a consequence rather than an input.

What actually gets rewarded when a machine is choosing a passage is structure: enough sections bounded by real headings that a document can be divided at all, and a short direct answer at the top of each one rather than three paragraphs of preamble. A tight page organised that way outperforms a long one written as unbroken prose, and the word count is an output of covering the question.

The section the guide does not have

There is one gap that is not a matter of emphasis. The guide has a section on promoting your website and a section on things not to focus on, and it has nothing at all about whether the systems that now answer queries can reach your pages, because that is not what it is for.

That is not a hypothetical concern. In our second study, of the 810 sites in the sample that publish a robots.txt at all, 28.3% disallow at least one major AI crawler and 1.5% disallow Googlebot. Almost none of those are decisions anybody made. They are defaults inherited from a security plugin, a CDN rule or a template, and the site owner following Google’s guide correctly would never encounter the question, because Google’s guide is about Googlebot and Googlebot is allowed.

You can do everything in the starter guide, do it well, and be invisible to the thing that answers your customer’s question.

How to actually use it

Read the guide first, and take the “shouldn’t focus on” list at face value. It will save you from most of the expensive nonsense in this field: the word count targets, the keyword density tools, the exact match domain, the E-E-A-T score somebody wants to sell you.

Then treat it as complete for the question it answers and add the three things it does not cover, none of which take an afternoon:

1. Fix heading order in the template. Levels descend one at a time, no empty headings. Heading structure checker.
2. Put an author and an organisation in the markup. Not styled text, machine readable. It is one template change.
3. Read your own robots.txt and decide what it says about AI crawlers. Decide it deliberately, either way. AI crawler checker.

Everything on that list is compatible with the guide. None of it contradicts anything Google says. It is simply the part of the job that a document about ranking in Google Search was never going to describe, and the guide is more honest about its own scope than most of the people quoting it.

Sources

• Google, “SEO Starter Guide”, developers.google.com/search/docs/fundamentals/seo-starter-guide, read 20 August 2026. All quotations above are from this document.
• visibility100x, GEO agency audit, 98 homepages, 88 fetchable, published 6 August 2026. Raw data public.
• visibility100x, AI crawler access study, 1,048 reachable sites drawn from the Tranco top 1,500, 810 of which publish a robots.txt. Published 11 August 2026. Raw data public.