The first time I saw an SEO strategy built around keywords, I was twenty-three and it was already out of date. It took the industry another decade to notice. It'll take another three years for the rest to catch up, and in that window, companies that understand what replaced the keyword will win disproportionately.

This is not a piece arguing that "keywords don't matter anymore" — that claim has been made badly by a lot of people, and the backlash against it is mostly justified. Keywords still matter. What's changed is what a keyword is, and how you should organise work around it.

The short version: the atomic unit of SEO work used to be the keyword. The atomic unit of SEO work in 2026 is the intent cluster — a group of semantically and behaviourally related searches that a single well-built piece can answer. If your team is still building content briefs around single keywords, you're doing three units of work to produce one unit of value.

What a keyword actually is

Back up. What is a keyword, really? For most of SEO's history, a keyword was a string of text a user typed into a search box. That was the unit because that was what the search engine could see. If you wanted to rank for "best running shoes for flat feet," you needed a page that matched that exact phrase, or close variants of it.

Google's understanding of that string was shallow. It could tokenise it, stem it, expand it with synonyms, and match it against your page's text. But the engine didn't really know what "flat feet" meant. It knew the words often appeared near words like "pronation," "arch support," "orthotic." The match was statistical, not semantic.

This changed in stages. BERT in 2019 taught Google to read sentences in context. MUM in 2021 extended that to understand queries across languages and formats. Then the LLM era arrived and Google stopped pretending it was a string-matching engine. It became something closer to a reading engine with a search index attached.

Google isn't matching strings anymore. It's reading the query, reading your page, and deciding whether your page answers the query. That's a completely different task.

This matters for your content team because the old playbook — "one page per keyword, match the query text, hit the word count, ship it" — was optimised for a version of Google that no longer exists. The pages that rank in 2026 are not the pages that most precisely match the query string. They're the pages that most completely answer the underlying question, including the adjacent questions a reader is likely to have next.

What an intent cluster actually is

An intent cluster is the set of queries a single underlying question generates. The question is the unit; the queries are surface variations.

Take "best running shoes for flat feet." That's a query. The underlying question is closer to: "I have flat feet and I want to start running. What should I wear and what do I need to know?"

The cluster of queries around that question might include:

In 2015, those would have been seven different pages. In 2020, they would have been one page with seven H2 sections, each targeting a different query. In 2026, the right answer is one page that reads like a single coherent piece, built to genuinely answer the underlying question, and which happens to cover all seven surface queries naturally along the way.

The distinction is subtle but it matters: you're not writing to target seven queries. You're writing to answer one question, and the seven queries are a signal that you got the question right.

Why your current process can't do this

Most SEO teams I audit are running a process that looks like this:

  1. Run keyword research. Get a big list of keywords.
  2. Filter by volume and difficulty. Prioritise.
  3. Write a brief for each prioritised keyword. Hand to writer.
  4. Writer produces content. Publish.
  5. Track rankings for each keyword. Repeat.

Every stage of this process is organised around the keyword as the unit. The tools reinforce it: your keyword research tool outputs keywords, your rank tracker tracks keywords, your brief template has a "primary keyword" field, your performance dashboard shows keyword rankings.

The process isn't wrong so much as misaligned. It's optimised for an era when the keyword was the unit. It now produces three structural problems:

Problem 1: Fragmented content

When the unit is the keyword, you end up with a lot of pages that all sort of answer the same underlying question from slightly different angles. The result is a content library where your own pages compete with each other for the same intent cluster. Google picks one of them (usually not the one you'd have picked), and the others dilute your site's authority rather than adding to it.

I see this most often on B2B SaaS blogs that followed the HubSpot playbook religiously in 2016-2020. They'll have forty posts on subtle variants of the same topic, and Google will rank exactly one of them for everything.

Problem 2: Shallow content that matches the query

When the brief is organised around matching a specific keyword, the writer produces content that matches the keyword. That sounds tautological, but what it means in practice is that the page answers the surface query rather than the underlying question. You get 800 words explaining what the keyword means, and then two sentences of actual advice. The page ranks for a week, then gets outranked by anyone who took the trouble to answer the real question.

Problem 3: Performance signals that don't match reality

When you track rankings by keyword, you miss the shift in how users actually find your content. You'll see a ranking drop on one keyword and panic, not realising that the underlying intent cluster is still sending traffic — just through a different variation of the query. Or you'll see a ranking improvement on a keyword and celebrate, not realising the traffic is going down because the cluster has shifted.

The right unit of measurement is traffic and conversions from the intent cluster, not rankings on any individual keyword inside it.

What to do instead

Three changes, in the order I'd make them:

1. Switch your briefing unit from keyword to intent cluster

Your content briefs should be organised around a question, not a keyword. The brief identifies the underlying question, lists the cluster of queries that represent it, specifies the reader (ICP), and describes what a complete answer looks like. The writer's job is to produce a piece that reads as a single coherent answer, not to stuff a keyword at a particular density.

In practice this means rewriting your brief template. The "primary keyword" field becomes "underlying question." The "related keywords" field becomes "queries in this cluster." You'll be shocked how much better the drafts come back.

2. Audit for fragmentation and consolidate

If you've been running the keyword-as-unit process for a few years, you almost certainly have fragmented content — multiple pages covering the same intent cluster from slightly different angles. These pages are hurting you, not helping.

Run a fragmentation audit: for each intent cluster you're trying to rank for, identify every page on your site that touches it. In 80% of cases you'll want to consolidate multiple pages into one authoritative piece, then redirect the others to it. This is one of the highest-ROI SEO tasks you can do right now, and most teams haven't done it.

3. Change what you measure

Track performance by intent cluster, not by keyword. Group your queries into clusters first, then look at aggregate traffic, clicks, and conversions for each cluster. Individual keyword rankings become diagnostic — useful for troubleshooting when a cluster underperforms — but they stop being the primary KPI.

Your rank tracker probably doesn't support this natively, which is itself revealing about where the tooling sits. In the meantime, most teams I work with export raw query data from Search Console and cluster it themselves, usually with a simple script.

The bigger picture

Every decade or so, search goes through a shift that changes what the actual work of SEO is. 2005 to 2012 was about links. 2012 to 2019 was about on-page optimisation and technical fundamentals. 2019 to 2024 was the transitional period where the old playbook still half-worked and the new one wasn't fully articulated.

We're now in whatever the next era is. I think it's organised around intent clusters as the unit of work, comprehensive answers as the mode of content, and the reading experience of the page as the core ranking signal. I could be wrong about the details. I'm not wrong about the direction.

The companies that figure this out first will build SEO moats that are genuinely hard to compete with. The companies that don't will spend another three years producing content libraries that their own pages compete with, measuring rankings on individual keywords, and wondering why the traffic isn't growing.

If you'd like to talk about what this looks like for your company specifically, you know where to find me.