The SEO work that is not worth doing

Every item below was either genuinely useful once or sounds like it should be. That combination is what keeps them alive: an idea that was correct in 2009 and an idea that has never been correct are indistinguishable to somebody encountering both in the same blog post, and search advice is republished far more often than it is rechecked. Tools and plugins add momentum by continuing to display a measurement long after it stopped meaning anything, because removing a feature customers ask about is harder than leaving it.

The cost is not that these tasks are harmful, though a couple are. It is that they are absorbing, measurable and satisfying to complete, which makes them extremely effective at displacing the work that would have helped. A business that spends its available hours adjusting phrase frequency and tuning a third-party score has done SEO all month by its own account, and has not touched the page that still lacks a title or the service page nothing links to. That is the actual damage, and it is why a list of things not to do is worth as much as a list of things to do.

The idea is that a target phrase should make up some particular percentage of a page's words, and that hitting the figure helps. It was never anything but folklore — no search engine has published such a target and none has confirmed one exists — and it survives because plugins display a percentage with a colour beside it, which feels like a measurement of something.

What is real is much simpler and needs no arithmetic: a page about a subject should use the words people use for that subject, because otherwise it is not obviously about it. That happens naturally when somebody knowledgeable writes about the thing. Working towards a percentage inverts the process and produces prose that repeats a phrase where a pronoun belongs, which reads badly to the customer you were writing for. Modern retrieval works on meaning rather than frequency counts, so a page discussing related concepts, answering adjacent questions and using natural synonyms is better served than one hitting a ratio. Write the page, then reread it and ask whether a stranger would know what it is about from the first paragraph. That question replaces the entire exercise.

The meta keywords tag is a list of terms in the page's head, invisible to visitors. Google confirmed it does not use the tag for ranking, and that confirmation is well over a decade old. Filling it in achieves nothing. The only argument for touching it at all is mildly against: the tag publishes your target term list to any competitor who views your source, which is a small gift for no return. If your platform generates the field, ignoring it is fine; there is no need to remove it from every page, and no benefit to populating it.

Submitting your site to search engines belongs in the same category and is stranger, because a whole industry of submission services still exists. Search engines find pages by following links. If any indexed page anywhere links to your site, you will be found; if nothing links to your site at all, that is the actual problem and submission does not address it. There is a legitimate version — requesting indexing for a specific URL in Search Console after publishing or genuinely fixing a page, which is reasonable once. What is not legitimate is any service offering to submit your site to hundreds of search engines, since the search engines that matter number very few and none of them requires or benefits from it.

The disavow tool lets a site owner tell Google to ignore specified inbound links. It exists for a narrow and genuine purpose: a site that engaged in manipulative link building, received a manual action for it, and needs to demonstrate cleanup. That is who it is for.

What happens instead is that a business installs a tool, sees a list of inbound links from sites it does not recognise, and disavows them defensively. This is unnecessary and carries real risk. Google has stated for years that it ignores obvious low-quality links automatically, because otherwise any site could damage a competitor by pointing rubbish at it — a system that fragile would not survive. Meanwhile the disavow file is a blunt instrument: nothing warns you if you disavow a domain that was sending you genuine value, and businesses have removed real links this way while trying to be careful. The rule worth following: if you have not received a manual action, and you have never engaged anybody to build links for you, this tool is not for you and the correct number of hours to spend on it is none. An unfamiliar linking domain is usually a scraper, an aggregator or a directory, and none of those needs your attention.

Domain authority, domain rating and similar metrics are estimates invented by the companies that sell SEO tools. They are not produced by any search engine, they are not inputs to any ranking system, and each vendor's version is a different number computed differently from that vendor's own crawl. Google has said plainly it has no such site-wide authority score of the kind these metrics imply.

They are not worthless: as a rough comparative signal, they can be a reasonable shorthand, and vendors are generally clear in their own documentation about what the figure is. The problem is what businesses do with it. Setting a target for the number, reporting it monthly as progress, choosing where to seek coverage based on it, or paying anyone whose offer is framed around raising it — all of these treat a vendor's estimate as an objective the search engine is scoring. It is not. The nearest thing to a legitimate use is comparing two sites roughly, with the vendor named and the word estimate attached. A monthly report tracking it as a headline figure is measuring one company's model, and any movement in it may reflect a change in the model rather than anything that happened to your site.

The redirect is the point of the whole exercise, and the replacement work is mostly unglamorous and immediately checkable. Read your most commercially important page's source and confirm the basics: it has a title that describes it and front-loads the subject, it has a description, it has one h1, it is not accidentally marked noindex, its canonical points at itself, its images have real alt text and dimensions. That is minutes per page and it is where genuine defects actually live. Then look at whether your important pages are linked from anywhere other than the menu, because a commercially critical page with one internal link pointing at it is a real problem that no score will reveal.

Then the work that is not technical at all. Answer the questions customers actually ask, sourced from your inbox and your phone log rather than invented. Put your prices, or state honestly what determines them. Make the practical facts on your contact page unmissable. Resize the photographs that are making your pages slow on a phone. If you have Search Console connected, find the queries with impressions and few clicks, which name pages where people saw you and chose somebody else. None of that produces a number to report, which is precisely why it loses to the items above — and it is the work that changes whether a visitor becomes a customer.