17 SEO myths that turn bad advice into busywork

Google has no magic word count, does not automatically penalize AI-assisted content, and does not require a special file for AI Overviews. Start by challenging the myths used to justify large changes without a diagnosis, especially bulk disavows and mass production of near-identical pages.

I’ve spent 20 years building publishing websites. My test for an SEO recommendation is whether we can name the problem it fixes and the evidence that would show it worked. A more impressive audit score is a weak answer on its own.

These entries focus on Google Search. Advice about Google’s AI features should not automatically be applied to every other search engine or AI assistant. The order follows the work affected, from content commissioning to technical fixes and AI visibility.

1. “Google automatically penalizes AI-generated content”

AI use alone does not establish a Google policy violation. Google’s guidance on AI-generated content distinguishes appropriate automation from content produced to manipulate rankings.

That distinction leaves plenty of room for bad publishing. Its current generative AI guidance warns that generating many pages without adding value may breach the scaled content abuse policy. Having an editor click “approve” does not explain what those pages contribute.

For a practical editorial check, pick a finished article and identify the claims someone verified, the information contributed by your publication, and the task a reader can complete with it. If the only improvement over existing results is different wording, the production method is a distraction.

Apply that check to human-written copy too. A person can produce a derivative article quite efficiently without assistance from a machine.

2. “Every keyword variation needs its own page”

A different phrase does not necessarily mean a different reader need. “CRM for small companies” and “small-business CRM software” could be served by the same comparison. A CRM implementation checklist probably needs a different page because it supports a different task.

Before adding a URL, compare the results for both queries and write a one-sentence purpose for each proposed page. If the purposes are interchangeable, begin with one useful page. Similar results are a clue, not a mechanical rule; product differences, language and local requirements can justify separate content.

There is also a policy boundary. Google’s doorway abuse policy covers pages made for similar queries that funnel visitors toward a more useful destination. Its scaled-content rules address mass production of low-value material, however it is made.

For practices that cross those lines, see our guide to 22 black hat SEO techniques to avoid, ranked by potential damage. A crowded content calendar is not, by itself, a spam violation. The purpose and value of the pages matter.

3. “Longer content ranks better, so aim for 2,000 words”

There is no universal minimum, maximum or ideal word count. Google says this explicitly in its SEO Starter Guide.

A returns policy might need a deadline, eligibility rules and instructions. A guide to migrating a large ecommerce site needs considerably more explanation. Padding the first to match the second would make it worse.

Use competing pages to find missing questions and useful evidence. Do not turn their average length into a requirement. Your brief should specify the decisions, examples and limitations the article must cover; the resulting length is something an editor can judge.

The qualification matters: a short page can still be incomplete. Removing filler and removing necessary explanation are different editing decisions.

4. “There is an ideal keyword density”

A percentage cannot tell you whether a page answers a query. Repeating a target phrase until an optimization tool approves it can leave the same information expressed several times while the useful detail remains missing.

Google’s Search Essentials recommends using the words people search for in prominent, descriptive places such as the title and main heading. It also prohibits keyword stuffing. Neither is an instruction to hit a percentage.

For a page about ecommerce returns, a better review asks whether it explains the return window, delivery costs, exclusions and refund process. Keep the language customers recognize. Remove repeated phrases that contribute nothing.

Keywords still help establish the subject. The mistake is treating their frequency as a substitute for answering it.

5. “Changing the date or publishing daily makes Google favor the site”

Freshness depends on the query. Google describes systems that favor fresher information when the search calls for it. A current software-pricing page and a guide to calculating a percentage have different maintenance needs.

That does not create a general daily-publishing requirement. Google’s helpful-content guidance specifically challenges changing dates without substantial changes and adding or removing content merely to make a site seem fresh.

Maintain a change list for important pages: which price, screenshot, product behavior or recommendation is now wrong? Fix those things. Update the date when the revision warrants it.

If nothing material changed, leave the timestamp alone. A newer date cannot make an obsolete screenshot accurate.

6. “E-E-A-T is a score you can raise by adding an author bio”

Google describes experience, expertise, authoritativeness and trustworthiness as concepts that help assess content quality. E-E-A-T itself is not a specific ranking factor, and quality raters do not directly determine individual page rankings.

A genuine biography helps explain who is responsible for an article. It cannot establish that a claimed product test took place. A review needs evidence appropriate to its verdict: what was evaluated, how, and with which limitations.

For a software review, for example, an accurately described workflow and screenshots from that work would be stronger editorial evidence than a paragraph calling the author an expert. If the review is based on documentation, say so.

Check the claims before polishing the biography. Readers should be able to understand why they can trust the material on the page.

7. “Domain Authority and Domain Rating are Google ranking factors”

Moz’s Domain Authority and Ahrefs’ Domain Rating are third-party metrics. Ahrefs explains that DR compares backlink profiles within its database. It is not a score supplied by Google.

The necessary qualification is that Google does use site-wide signals as well as page-level assessments. Rejecting a vendor metric as a Google input does not prove that every assessment happens in isolation at the page level.

Use these metrics to organize research, then inspect the pages and links behind them. A link prospect’s subject, readership and editorial standards matter more to a publishing decision than whether its score clears an arbitrary threshold.

Ahrefs itself cautions against making improvement of the authority score the primary goal. Put the commercial objective in the brief: relevant visitors, useful referrals or stronger visibility for a particular set of pages.

8. “Backlinks no longer matter”

Google continues to list link analysis and PageRank among its ranking systems. Declaring links obsolete is not supported by that documentation.

It also does not tell you that every site needs the same number of links, that links outweigh all other considerations, or that the competitor with the largest backlink count deserves to win.

A useful planning question is what would make another publisher cite your page. A reproducible test, a maintained dataset, a specialist explanation or an original interview can give them a reason. Merely writing a longer version of their article is a less persuasive proposition.

Buying links to manipulate rankings remains covered by Google’s link-spam policy. Paid sponsorships need appropriate link qualification. The continued relevance of links does not validate every method sold for acquiring them.

9. “Every toxic backlink needs to be disavowed”

A tool’s warning is a reason to investigate, not an automatic instruction to upload a disavow file. Google says most sites do not need its disavow tool and warns that incorrect use can harm search performance.

Its documented criteria combine a substantial number of artificial, spammy or low-quality links with a manual action, or a likelihood that those links will cause one. It recommends trying to remove problematic links first.

That makes the history relevant. Links you commissioned through a ranking-manipulation scheme require a different investigation from a random collection of scraper sites that appeared without your involvement.

Before acting, record which links you believe are problematic, why, and whether your site or supplier arranged them. Check Search Console for a manual action. Do not let a recurring “link cleanup” task become a substitute for that assessment.

10. “A traffic drop means Google penalized the site”

A fall in clicks describes an outcome. It does not identify the cause. Google’s traffic-drop guidance covers technical problems, ranking changes, changing demand and seasonality, among other explanations.

Check the Manual Actions report for a manual action. A clean report does not rule out an algorithmic reassessment, an indexing problem or a weaker result against competitors.

Compare the affected pages and queries across equivalent periods, with country and device filters where relevant. Falling impressions, falling positions and lower click-through rates suggest different lines of investigation. A site-wide average can conceal which pages changed.

Keep a record of your own releases too. A navigation change or accidental indexing control deserves attention before a theory about an unnamed update.

11. “Duplicate content always triggers a penalty”

Ordinary duplication, such as the same content being accessible through several URLs, is not automatically a manual-action offense. Google makes that distinction in its starter guidance.

There can still be a technical problem worth fixing. You may want a particular URL shown in search, or need consistent signals about which version represents the content.

Google documents redirects and canonical annotations as strong canonicalization signals, with sitemap inclusion a weaker one. Choose the method that fits the URLs: a retired duplicate and a variant that users still need are different cases.

Inspect a representative URL to see which canonical Google selected before rewriting an entire site. This guidance about duplicate URLs also does not grant permission to copy someone else’s work or publish mass-produced variations.

12. “Blocking a page in robots.txt removes it from Google”

For web pages, robots.txt controls crawling; it is not a reliable removal mechanism. Google explains that it can still index a blocked URL discovered through links, even without fetching the page’s content.

For an accessible page you want excluded from search, use an appropriate indexing control. Google must be able to crawl the page to read its noindex tag or response header. Blocking that crawl can prevent it from seeing the instruction.

Confidential material needs authentication or another genuine access restriction. Excluding a page from a search index does not stop someone opening its URL.

Write down the requirement before touching the configuration: fewer crawl requests, exclusion from search, or restricted access. Those are three separate jobs.

13. “A sitemap guarantees indexing and replaces internal links”

A sitemap helps Google discover URLs. Google explicitly says it does not guarantee they will be crawled or indexed.

Internal links do additional work: they connect relevant resources for readers and help Google discover and interpret pages. Google’s link guidance recommends that every page you care about has a link from at least one other page on your site.

After publishing a useful new guide, identify existing pages where a reader would genuinely need it and add contextual links there. Checking only whether the new URL appears in the sitemap misses that opportunity.

If an important page remains unindexed, examine that page’s status and accessibility. Repeatedly submitting the same sitemap will not explain the problem.

14. “A 100/100 PageSpeed score will get the page to the top”

Good performance matters. A perfect score does not establish that your page is the best answer. Google’s page-experience documentation says Core Web Vitals are used by its ranking systems, while also warning against chasing perfection purely for SEO.

Its Core Web Vitals guidance focuses on loading, responsiveness and visual stability. Those are useful things to investigate: can visitors see the main content promptly, interact with the page and avoid unexpected layout movement?

Prioritize problems on important templates and devices. A purchase button that shifts as an advertisement loads is a concrete defect. Spending another week on an already usable page solely to make a test score greener requires a better explanation.

Keep the content diagnosis alongside the performance diagnosis. A fast page that omits the information behind the query still omits it.

15. “Valid structured data guarantees a rich result”

Passing a technical validation test does not guarantee display. Google’s structured-data policies make that explicit. The page must also meet the relevant feature and content requirements.

Choose markup that accurately describes the visible content and a supported feature. Then validate it, check eligibility and inspect whether Google actually displays that feature. Keep those questions separate when reporting results.

An audit can truthfully say “the syntax is valid” while leaving the other questions unanswered. A developer has completed one piece of the work, not secured an entitlement to extra space in the search results.

Use accurate markup where it serves the page. Do not add unsupported claims or invisible material because a competitor’s result looks more elaborate.

16. “Google’s AI Overviews require llms.txt or special AI schema”

Google says its AI search experiences do not require new machine-readable files or special schema. A mandatory “AI file installation” sold as the price of admission to AI Overviews misstates that requirement.

If a supplier proposes a file or integration, ask which system consumes it, what it is expected to change, and how that change will be measured. That is a reasonable implementation discussion. A guarantee of Google inclusion is not.

Check ordinary eligibility and whether important content can be accessed and understood before commissioning extra files. This conclusion is specifically about Google’s documented requirements. A different AI product may have its own integration or crawler behavior, which needs separate evidence.

17. “Ranking first guarantees a citation in an AI answer”

Google explains that AI Overviews and AI Mode may use multiple related searches, called query fan-out, while generating a response. Their supporting links can therefore differ from the results for the original query.

Being indexed and eligible for a search snippet meets a documented eligibility condition. It does not guarantee selection. Nor should you assume AI Overviews and AI Mode will display the same sources.

Consider a hypothetical search for email software for a charity. A comparison page may help with the overall choice; a provider’s nonprofit eligibility policy may answer a narrower question. Those pages contribute different information. This is an illustration of why a single ranking cannot describe every possible supporting source, not a claim about a live result.

Track the questions, sources and referrals that matter to your business. When useful information is missing from your site, improve it because it helps the decision. Do not promise that reformatting a paragraph will force an AI system to cite it.