- Front-loaded titles, internal links to commercial pages and answering early are the three items from 2020 that still work.
- Keyword density never worked, and the prose it produces is now a liability rather than a neutral cost.
- Bought links moved from low-yield and low-risk to low-yield and risky; the footprints are trivial to detect.
- Third-party authority scores are a prospecting tool, not a target, and we dropped them from client reports in 2019.
- The item that did not exist in 2020 is being identifiable: one name, one address, named authors, a clear statement of what you do.
This page has been on our site since January 2020. It has picked up links and it still gets traffic, which is exactly the situation where most agencies change the date at the top and call it updated.
Instead, here is the original list with an honest verdict on each item, six years and roughly forty core updates later. Where we were wrong, we say so.
Still works: writing the title for the query, not the page
The original advice was to put the term people actually search at the front of the title tag. That has survived everything. Google rewrites titles more often than it did in 2020, but it rewrites bad titles. A clear, specific, front-loaded title is usually left alone.
What changed is the source of truth. In 2020 we took the term from a keyword tool. Now we take it from Search Console: filter to the page, read the top ten queries by impressions, write the title against those. Better data, same technique.
Still works: internal linking to your money pages
The 2020 version said "link from your blog posts to your service pages". Still correct, still the most reliable improvement available on an established site, and still the thing almost nobody does systematically.
One refinement. Vary the anchor text, and write links into sentences a reader benefits from. A block of six identical exact-match links at the bottom of every article is a pattern, and patterns get discounted. The current version of this advice sits in our first-day SEO checklist.
Still works, differently: targeting featured snippets
Answering the question in the first forty words under a clear heading still gets you selected. What has changed is what selection is worth. Sometimes it is a snippet and a click. Increasingly it is a sentence inside an AI summary with a citation link most readers will not follow.
Our position now: keep doing it, because it is free and it is also just clearer writing, but stop counting the traffic as the return. Count it as being present in the answer. Advanced SEO has the structural detail.
Stopped working: keyword density
We hedged on this in 2020 and we should not have. Aiming for a percentage of keyword occurrences has no useful effect, and it produces prose that reads like a machine wrote it, which is a far worse thing to look like now than it was then.
Write the page for someone deciding whether to hire you. Use the words they use. That is the whole rule.
Stopped working, now harmful: bought links
In 2020 a batch of directory submissions and a few paid guest posts was a low-yield but low-risk tactic. It is now low-yield and genuinely risky. The patterns, same anchor, same footprint, sites that exist only to sell links, are the easiest thing in the world to detect at scale.
What replaced it in our own work: publishing data or tooling someone has a reason to reference, writing up client work where we have permission, and local relationships, sponsorships, associations, suppliers. Slower, smaller, and it never needs undoing.
We still get the same question twice a month: someone has been offered 50 links for a flat fee. The honest answer is that the offer is priced correctly for what it is worth.
Do not reach for the disavow tool first. Check for a manual action in Search Console. Without one, disavowing at scale usually removes more value than it recovers.
Stopped working: chasing a third-party authority score
Domain authority is a model built by a tool vendor to approximate Google. It is useful for comparing two link prospects. It is not a target, it is not a ranking factor, and buying links priced by it is how a lot of small businesses have wasted a lot of money.
We removed authority scores from client reports in 2019 and nobody has asked for them back.
Still works: answering the question in the first paragraph
The original list called this "write for the snippet". The technique is unchanged and the reason has broadened: a reader scanning on a phone, a search engine choosing a snippet, and a language model looking for an extractable answer all reward the same thing. Say the answer, then explain it. Bury it in paragraph four and all three move on.
Neutral now: exact-match domains, meta keywords, image geotags
These were fading in 2020. They are now irrelevant. Meta keywords have been ignored for well over a decade, geotagged photo metadata does nothing for local rankings, and an exact-match domain is neither help nor hindrance, though it can make a business sound less credible than its own name would.
The new item that did not exist in 2020
Being identifiable. AI systems and classic search both reward a business with one consistent name, one address, named authors, and a clear statement of what it does and since when. In 2020 we would have filed that under branding. It is now an SEO input, and it is the cheapest one on the list.
What this exercise taught us
Everything on the original list that survived was not a trick. It was ordinary craft: describe the page accurately, connect it to related pages, answer the question early, be a real business. Everything that failed was an attempt to signal quality without producing it.
That is a boring conclusion, and it is the one we would bet on for the next six years.
If you inherited a site built on the old list
The order we work in: check Search Console for a manual action, then re-title the pages that already rank in positions 8–20, then fix internal links, then decide what to do about the thin pages. Rewriting everything at once is what people reach for and it makes attribution impossible. If the site carries link damage from an old campaign, that diagnosis is worth paying for rather than guessing at, and it is where most of our SEO engagements start.
Questions
Are there any SEO shortcuts that still work in 2026?
Title rewrites against real query data and internal linking are the closest thing: cheap, fast, and effective on established sites. They are not shortcuts so much as work most sites have never done.
Should I remove old blog posts that use 2020 tactics?
Rewrite the ones with links or traffic, merge the ones that overlap, and remove only genuinely dead pages. Deleting a page that still earns clicks to tidy up a tactic is a net loss.
Does keyword research still matter?
Yes, for deciding what to publish. It matters much less for how to write once you have decided. Use tools for demand and Search Console for language.
How do I know if a link is worth having?
Ask whether a relevant human might click it. If the page exists to host links, the answer is no, regardless of what any metric says about the domain.