Somebody asked me at a meetup last month what they should do about keyword density on their landing page. I had to physically stop myself from sighing.
It is not that they were stupid for asking. It is that the advice they had clearly read was from 2014 and nobody had told them to stop reading it. There is a whole archaeology of SEO tips that used to be true, then became neutral, then became actively bad, and the blog posts explaining them are all still ranking. So you read them. So you do them. So your site gets a little worse.
Here is the stuff I used to tell people, and what I think now.
Keyword density
Around 2010 I genuinely thought there was a magic number. Two percent? Three? I had a little tool I used (I think it was called SEOQuake or something similar) and I would tweak copy until the percentages looked right. I made clients add their target phrase to paragraphs where it did not belong. Reader, I am sorry.
Google has not cared about keyword density in any meaningful way for over a decade. The search engine reads context now, not counts. If your page is about replacing a kitchen tap and you mention “kitchen tap” once and “the tap” four times and “this fitting” twice, that is fine. That is how humans write. The page is about what the page is about.
What I tell people now: write the sentence the way you would say it out loud. If the keyword fits, great. If forcing it makes the sentence weird, do not force it.
Meta keywords
I am putting this here just so I can laugh. Yes, in 2026, I still occasionally see <meta name="keywords"> tags filled in. Google publicly stopped using them in 2009. Bing followed not long after. They do nothing. They have done nothing for seventeen years. Pull them out of your template, save the bytes.
The only thing meta keywords are good for these days is showing your competitors what you think your keywords are, which is a small gift to give them for free.
Exact-match anchor text
This one I gave as advice well into the 2010s. “When you link to your services page, use the exact phrase you want to rank for.” Buy a few links from a directory, all with the same anchor text. Tidy.
Then Penguin happened in 2012, and the sites that did this loudest got obliterated. The pattern of every inbound link saying the exact same three words is the most obvious unnatural-link signal there is. Real people link with phrases like “this guide,” “Marin’s site,” “here,” or, embarrassingly often, just the URL. A link profile that is 80% “best plumber Glasgow” looks fake because it is fake.
I tell people now: if you control the anchor (your own internal links, a guest post you wrote), pick something descriptive but natural. If somebody else is linking to you, do not ask them to change the wording. It will read better than anything you would script.
Stuffing locations into footer text
You have seen this. “We serve London, Manchester, Birmingham, Liverpool, Leeds, Sheffield, Bristol, Cardiff, Glasgow, Edinburgh, Newcastle, Belfast…” running across the footer in 9pt grey text.
I do not know who started doing this but they have a lot to answer for. It worked, briefly, around 2008. It does not work now. It looks spammy to users, it looks spammy to Google, and it tells visitors that you are the kind of business that thinks they are clever. If you actually serve multiple cities, build proper local pages with real content for each one. If you do not, do not pretend you do.
Long content for the sake of length
The “1500+ words because Google likes long content” school of thought is the one I feel worst about. I gave this advice. I had clients pad answers, repeat themselves, add a “frequently asked questions” section that nobody had ever asked, all to hit some imagined word count threshold.
Length correlates with rankings, sometimes. The causation is not what people think. Longer pages tend to rank because longer pages tend to cover topics more thoroughly, not because the word counter sends Google a signal. A 600-word page that fully answers the question will outperform a 2,200-word page that buries the answer under three paragraphs of throat-clearing.
If you can answer the user’s question in 400 words, answer it in 400 words. Save everyone’s time.
Heavy schema everywhere
I went through a phase where I added Schema.org markup to everything. Every paragraph an Article. Every footer link a BreadcrumbList. Every heading wrapped in WebPage properties. FAQPage markup on pages that were not actually FAQs.
Most of it did nothing. The FAQ rich results that everyone chased got rolled back hard around 2023, and pages stuffed with fake FAQ schema ended up looking worse than ones with no schema at all. I still use schema where it represents reality (a recipe is a recipe, an event is an event, a product has a price), but I have stopped sprinkling it like seasoning over content that does not need it.
What I actually say now
This is the boring part. The advice that has held up.
- Write the page the reader actually came for. Answer their question first, in the first paragraph if you can.
- Link sensibly. Link to your other pages where the link is genuinely useful, not because you are passing “link equity.”
- Fix the speed issues. Big images, render-blocking scripts, fonts loading from three CDNs.
PageSpeed Insightsstill tells you most of this for free. - Make sure the headings actually describe the sections.
- Do not block the crawl with garbage in
robots.txt.
That is roughly it. I used to have a forty-point checklist. Now I have about six things, and most of them are just “write a decent page that loads quickly.” The job got smaller and clients are sometimes disappointed when I tell them that. They want the magic. There is no magic. There never was, but at least the old magic generated invoices.