Search myths, checked against the sources

Search myths persist for a structural reason. Advice outlives the system it described. A tactic that worked in 2011 gets written down, the system changes, and nobody goes back to retract the post. The claim keeps circulating because it keeps being repeated, not because anyone rechecked it. Each myth below is answered against a primary source, in most cases Google's own documentation, with the link next to the answer so you can read it yourself.

Meta keywords still matter

They do not. Google's SEO starter guide lists the meta keywords tag among the things not to focus on, and states it plainly: "Google Search doesn't use the keywords meta tag." Google published that position in a September 2009 Search Central post explaining that the tag had been abused into uselessness, and has never reversed it. Filling the tag costs nothing and does nothing.

There is a minimum word count

There is not. The starter guide answers this directly: "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)." Longer pages sometimes correlate with better rankings because thorough answers tend to run long. The length is a side effect of the coverage, not a target to hit.

Duplicate content carries a penalty

It does not. The starter guide names this one as a myth in so many words: "If you have some content that's accessible under multiple URLs, it's fine; don't fret about it. It's inefficient, but it's not something that will cause a manual action." What actually happens is described in Google's canonicalization documentation: when several URLs serve the same content, Google selects one as canonical and consolidates signals onto it. The cost is wasted crawling and split signals, not a penalty. Copying other people's content is a separate matter, handled by the spam policies.

Keyword density has an optimal percentage

Google has never published a target percentage, and no Google document names one. What does exist is the opposite boundary: keyword stuffing is a named violation in Google's spam policies, defined as "filling a web page with keywords or numbers in an attempt to manipulate rankings in Google Search results." The starter guide adds the readable version: excessively repeating the same words, even in variations, is tiring for users. There is a floor of relevance and a ceiling of spam, and no dial in between.

Heading tags must run in strict order to rank

Semantic heading order is worth doing, but not for the reason usually given. The starter guide is explicit: "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." The same passage adds that there is no ideal number of headings for a page. Order your headings correctly because screen reader users navigate by them, and stop auditing it as a ranking defect.

A subdomain ranks differently from a subdirectory

Google treats this as a business decision rather than a ranking one. The starter guide covers subdomains versus subdirectories under things not to focus on: "From a business point of view, do whatever makes sense for your business," noting that subdirectories can be easier to manage while subdomains can make sense for partitioning topics. Case studies showing traffic gains after a migration usually change several things at once, including URL structure, internal linking and consolidation, so they do not isolate the subdomain variable.

Google penalizes AI content

Not for being AI-written. Google's February 2023 statement is one sentence: "Appropriate use of AI or automation is not against our guidelines." The line it draws is purpose, not production method, and automation used primarily to manipulate rankings violates the spam policies. Measurement agrees: a study of 1,000,000 pages found top-10 positions containing between 8.4 and 11.7 percent pages that were at least 80 percent AI generated. The full evidence is in does Google penalize AI content.

llms.txt improves AI visibility

There is no measured support for it. Google states in its AI features guide that you do not need to create AI text files or markup to appear in Google Search, because Google Search ignores them. Server logs point the same way: a study of 137,000 sites found that 97 percent of published llms.txt files were never fetched by anything. The proposal and the measurements are laid out in llms.txt.

Structured data is required to appear in AI answers

Google says otherwise in the same guide: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." Structured data still earns rich results in classic search, which is a good reason to keep using it, but that is a different claim from the one being sold. Where it helps and where it does not is covered in structured data.

More pages means more traffic

Only when each page earns its place. Google's spam policies name scaled content abuse: "when many pages are generated for the primary purpose of manipulating search rankings and not helping users." The AI features guide adds that creating a separate page for every query variation falls under that policy. Volume is not the offense; sameness at volume is, and the consequence lands on the whole site rather than on the weakest page. See thin content and the scaled content abuse policy.

Paying for a tool gives you access to Google's ranking systems

It does not. Google's AI features guide says it flatly: "No third-party tool has access to our internal ranking or AI systems." Any external number about your rankings or your AI visibility is an outside-in estimate built by sampling, whatever the dashboard implies. Estimates are useful when you know what they are, and they are not readings taken from inside Google. The first-party numbers live in Search Console.

How to check a claim yourself

Three steps settle most of these. Find the primary source, meaning the vendor's own documentation rather than a summary of it. Check its date, because search advice ages badly and a correct answer from 2016 can be wrong now. Then prefer documentation over commentary, and where documentation is silent, prefer a published measurement with a stated sample size over an assertion with none. If nothing turns up at any of those levels, the honest answer is that the question is open, which is a better place to stand than a confident wrong one.