SEO
SEO in 2026: what still works and what does not
· 4 min read · Oddbird
Zero-click, AI Overviews and ever smarter algorithms. A sober inventory of what has fallen away, what now weighs more, and where you put your time.
What you take away
- Positions have become a worse yardstick than impressions and enquiries. Measure at the bottom, not the top.
- Technical work has not become less important, but it has shifted: from speed to readability for machines.
- The biggest gain usually sits in pages you already have that are just underperforming.
Much has been written about the death of SEO in recent years. That is not true, but the feeling underneath it is: anyone working the way they did four years ago sees returns fall without doing anything wrong. The field has shifted, not disappeared.
The shortest summary: search engines increasingly answer themselves, so the traffic that does come through is smaller and better. That calls for different choices, and above all for different numbers to steer on.
What has fallen away
| Used to work | What happens now |
|---|---|
| Thin pages per keyword variant | They compete with each other and dilute your own authority. One good page beats five half ones. |
| Bought links and link exchanges | Largely ignored, and a risk once patterns are recognised. Doing nothing has become expensive. |
| Writing for the algorithm | Systems have become good enough to recognise that a text was not written for people. |
| Steering on average position | With an answer at the top, position three says little about what still reaches you. |
| Blog volume as a goal | Without distinctive substance, publishing more mainly yields more pages that do nothing. |
The common thread: everything that leaned on quantity rather than substance has become worth less. That is not a moral move by search engines, it is a technical one: they can now judge what is there.
What now weighs more
Search intent over search volume
A keyword with ten thousand monthly searches where nobody buys anything is worth less than a search with forty a month where someone has their wallet out. We always knew that, but now that the top of the funnel is largely absorbed by AI answers, the difference has become painfully visible.
In practice: sort your keywords not by volume but by how close they sit to a decision. "What is ceramic coating" is a different kind of question from "ceramic coating cost", and the second deserves your best page.
Pages that are just underperforming
The quickest gain rarely sits in new content. It sits in pages ranking eight to fifteen for a search that actually matters to you. Those pages are already found and already judged; they usually lack just one thing, and that is often that they do not fully answer the question.
- Look in Search Console for pages with many impressions and few clicks.
- Establish per page which exact question it wants to answer.
- Read it as if you had that question and note what you would still have to Google afterwards.
- Add exactly that, and nothing else. Do not expand for the sake of expanding.
Readability for machines
Technical work has shifted. Speed and Core Web Vitals remain important, but at most sites they are now in order. Where it goes wrong is the layer above: structured data, one clear main heading per page, a title that says one thing, and a sitemap that is correct.
That is no longer an SEO detail. It is the basis on which both Google and an AI model decide what your page is about. An eighty-six character title listing six services gets cut off in search results and gives an AI model no usable summary.
What you steer on now
The most common mistake in this new field is that people keep steering on positions while positions say less. The following set works better.
| Metric | Why this one | Where you find it |
|---|---|---|
| Impressions per query | Shows whether you appear at all, even when nobody clicks. | Search Console |
| Click-through rate on your key pages | Measures whether your title and description do their job. | Search Console |
| Enquiries from organic traffic | The only number attached to the till. | Analytics, with goals set up |
| Mentions in AI answers | The new visibility, and still barely measured by anyone. | Manually, with fixed questions |
If you may track only one number, track enquiries from organic traffic. Everything above it is an intermediate step.
What this means for a small team
For a business without its own marketing department the main conclusion is reassuring: you do not have to publish more, you have to choose better. Four good pages around the questions your customers actually ask do more than forty pieces nobody reads.
And start with measurement. Without analytics and Search Console everything above is a guess, and then it is impossible to see whether anything works. At a surprising number of businesses, that is the first thing missing.