SEO Myths and Facts: Meta Keywords, Word Count, AI Content and "Guaranteed Ranking" Promises

Version note: This post is based on the Google Search Central documentation (the SEO (Search Engine Optimization) Starter Guide, the spam policies, the generative AI (Artificial Intelligence) content guidance and the third-party SEO guidance published in June 2026) as of November 2026.

Guess First

An agency sends the Bookstore team a proposal. It has five items: (1) meta keywords will be added to every page, (2) the density of the main keyword in the text will be raised to 2–3%, (3) every blog post will be at least 2,000 words, (4) all AI-written content will be rewritten "because Google penalizes it", (5) the agency guarantees the top position within six months. How many of these items do you think can be backed by Google's own documentation? Which one could even harm the site if done?

The Anatomy of a Myth: Where Does It Come From?

Most SEO myths aren't pure invention; they are claims that were once true in some way, that derive false causation from a correlation, or that are born when a tool's own metric is mistaken for "Google's metric". Meta keywords were used by some search engines at one point. The observation that long content ranks well may come from the fact that well-ranking topics are comprehensive. A tool's "domain authority" score is the tool's own estimate.

The method in this post is simple: take every claim to Google's own documentation. If it's in the documentation, write it with its source; if not, say "Google doesn't confirm this". Google's third-party SEO guidance published in June 2026 recommends the same test: credible advice labels its claims as opinion or experience, or bases them on Google's official documentation.

Meta Keywords

In a Search Central (then called Webmaster Central) blog post dated September 21, 2009, Google announced that it doesn't use the keywords meta tag in web ranking. Today's SEO Starter Guide has the same sentence: Google Search doesn't use the keywords meta tag. The same blog post notes that the description meta tag, on the other hand, is sometimes used as snippet text; that's the topic of the fourth post in this series.

What the developer should do: if the CMS (Content Management System) has a "keywords" field, remove it so editors don't waste effort, or at least make it optional. Printing the tag does no harm, but it contributes nothing to ranking either; it does nothing except show competitors which words you target.

Keyword Density and "LSI Keywords"

Google's documentation has no keyword density target. On the contrary, keyword stuffing is a spam policy violation: filling a page with keywords or numbers in an unnatural way. The SEO Starter Guide also says that repeating the same words excessively both goes against the spam policies and tires the reader. Google's ranking systems guide says BERT (Bidirectional Encoder Representations from Transformers), neural matching and RankBrain match words to meanings and concepts; repeating the same word adds nothing to meaning matching.

The LSI (Latent Semantic Indexing) keywords claim belongs to the same family. LSI is an information retrieval technique published by Deerwester and colleagues in 1990. There is no "LSI keyword" concept in Google's Search Central documentation; Google doesn't confirm this claim. The systems Google describes for meaning matching are the ones listed above.

What the developer should do: don't add a "keyword density" check to the CMS or the CI (Continuous Integration) pipeline. Instead, check what Google's key best practice says: are the words people would search for present in prominent places such as the title, main heading, alt text and link text?

The Ideal Word Count

The SEO Starter Guide's sentence is clear: there's no magical word count target. Google's generative AI search guide published in May 2026 repeats the same thing: there is no ideal page length. The agency's "every post at least 2,000 words" item therefore isn't based on Google; padding the length with filler text goes against the "Why" part of the helpful content questions, that is, why the content was made.

How long should a post be? Long enough to cover the topic completely. The posts in this series are between 1,300 and 2,000 words because the topic requires it; the answer to "what does status code 404 mean?" is two paragraphs.

What the developer should do: don't put a minimum word count rule in the CMS.

Domain Age, Keyword Domains and the Duplicate Content "Penalty"

There is no statement in Google's documentation that a domain's age is a ranking factor; Google doesn't confirm it. On keyword domains, Google has two clear sentences. According to the SEO Starter Guide, keywords in the domain name or URL (Uniform Resource Locator) path have hardly any effect beyond appearing in breadcrumbs. The ranking systems guide also describes the exact match domain system, which limits the credit given to domain names chosen to exactly match a query. For the TLD (Top-Level Domain, e.g. .com, .com.tr), the Starter Guide says it only plays a role in country targeting, and even there it's usually a weak signal.

The "duplicate content penalty" is another frequently heard myth. The Starter Guide is clear: having the same content at multiple URLs is inefficient, but it doesn't lead to a manual action. Copying other people's content is a different matter; that falls under the scraping spam policy. On the subdomain versus subdirectory debate, the Starter Guide says "do whatever makes sense for your business".

What the developer should do: choose the domain and URL structure for users and maintenance, and consolidate URL variants of the same content with a canonical. That matters for crawl efficiency, not out of fear of a penalty.

"Google Penalizes AI Content"

This sentence gets misread in both directions. According to Google's generative AI content guidance, generative AI can be useful for researching a topic and adding structure to original content; the content must still comply with Search Essentials and the spam policies. The line is in the scaled content abuse policy: producing many pages without adding value is a violation "no matter how it's created". The helpful content page also cites producing lots of text with generative AI without manual oversight or curation as an example of "little to no effort".

The guidance has two practical additions: manually check all AI-generated content before publishing, including titles, meta descriptions, structured data and image alt text, and explain how automation was used where readers would expect it. For e-commerce sites, AI-generated images must carry IPTC (International Press Telecommunications Council) DigitalSourceType metadata set to TrainedAlgorithmicMedia.

Back to the agency's fourth item: rewriting "because it was written with AI" is the wrong reason. The right question is whether the content is accurate, original and useful to the reader.

What the developer should do: if you write a job that produces content, put a human approval step and a volume limit on its output; keep "generated with AI" as a metadata field in the publishing pipeline.

"Guaranteed Rankings" and Third-Party Tools

Google's "Do you need an SEO?" page is clear on this in one sentence: no one can guarantee a #1 ranking on Google. The same page warns against those who claim a "special relationship" with Google, offer a "priority submit", won't explain their methods, or promise submission to thousands of search engines; it recommends granting read-only access for an audit. In June 2026 a section on evaluating an SEO's recommendations and the tools they use was added to the page.

The third-party SEO tools and services guidance published on the same date is even more concrete: Google doesn't evaluate third-party services, so be skeptical of a "Google approved" claim; no tool or service guarantees ranking success; third-party tools have no access to Google's internal ranking data, and their predictions are their own. The guidance specifically recommends scrutinizing these services: sitemap generation and indexing directives, "SEO-optimized" content generation, claims to improve the rankings of existing content, and AEO (Answer Engine Optimization) or GEO (Generative Engine Optimization) promises for AI experiences. A tool's "authority" score is therefore the tool's own model, not Google's. The one place recommended for data directly from Google is Search Console.

What the developer should do: turn every recommendation in a tool or agency report into a ticket by asking for its source. If the source is Google documentation, apply it; if not, label it an "experiment" and measure it with Search Console data.

Production Notes

Collecting the myths from this series and Google's answers in a single table shortens team discussions:

Put the table on a wiki page and add the link to the Google documentation on every row. Google's documentation changes; checking the change log at developers.google.com/search/updates once a month keeps the table from going stale.

Common Pitfalls

Taking correlation for causation. Observations like "top-ranking pages average 2,000 words" may describe a common trait of good pages on that topic, not the cause of the ranking.

Taking an old Google statement for today's rule. Changes such as the helpful content system joining core in 2024 make old advice obsolete; always look at the current version of the documentation.

Falling into the opposite myth while fighting one. "Meta descriptions are useless" is also wrong; they don't affect ranking, but they can be used as the snippet.

Taking a third-party tool's chart for Search Console data. Google's guidance draws attention to exactly this confusion.

A Small Experiment

★☆☆ Pick an SEO tip you read recently and look for its source on Google Search Central. If you find it, write down the link; if not, write "Google doesn't confirm".

★★☆ Is there a field, rule or check in your own CMS or project that rests on one of the myths in this post? If so, write a short justification for removing it and attach the Google documentation to it.

★★★ Write a reply that evaluates the five items of the opening agency proposal one by one: for each item, add what Google says, its source and what you recommend instead.

The SEO Fundamentals series ends here: we've seen how a search engine processes a page, the rules, ranking, appearance, the diagnostic tools and the myths. The next step is the GEO (Generative Engine Optimization) series: how AI answer engines find and cite sources, and how different from SEO, and how similar, that really is.