The “SEO is dead” crowd has been wrong for a decade. They will be wrong next year too. But the SEO that worked in 2020 does not work now — and a lot of brands are finding that out the hard way.
The traffic drops are real. The panic is understandable. What most brands are getting wrong is the diagnosis.
What Google actually changed — and why most people got it backwards
Google’s helpful content updates were not about rewarding longer articles or penalising AI writing. They were about intent. Content built to rank rather than to answer something — that is what took the hit. We had a B2B client lose 40% of their organic traffic in a single core update. Their content was not thin. It was thorough. But it was written for a crawl bot, not a person. When we rewrote with an actual reader in mind, most of that traffic came back within three months.
The uncomfortable truth is that a lot of SEO best practices were just reverse-engineering ranking signals — and Google got better at spotting that.
The mistake most teams made was reading the update notes and optimising for the stated criteria. Author bios. E-E-A-T signals. That kind of thing. Some of that matters. But the brands who recovered fastest were not the ones who added an author photo and called it expertise. They were the ones who looked at their content and asked a harder question: would someone who already knew the answer to this recommend it to someone who did not? For most SEO content produced between 2018 and 2023, the honest answer is no.
We went through a client’s entire content library — 340 articles — and scored each one against a single question: does this actually help the person who searched for it? Around 60% scored poorly. Not because the information was wrong, but because the structure was designed around a keyword rather than a question. The intro stalled. The answer was buried. The conclusion went nowhere useful. Fixing those 200 articles properly took four months. The alternative was watching the traffic graph continue in one direction.
There is a related mistake that almost nobody talks about: the canonicalisation of mediocre content. A lot of sites grew their libraries fast between 2019 and 2022, publishing three posts a week on adjacent topics, letting them compete with each other for the same searches. We had one client — a software company in the HR space — with eleven articles all targeting variations of the same intent. None of them ranked. The combined traffic across all eleven was about 300 sessions a month. We consolidated them into two, redirected the rest, and within six weeks both of the consolidated pages were on page one. The site had less content. It performed better. That runs against every instinct people have about SEO, but it is what the data said to do.
What still holds up
Technical SEO is not glamorous but it still matters more than people give it credit for. Core Web Vitals improvements alone — particularly page speed on mobile — still move rankings more than most people expect. Clean site structure and crawlability matter too, but the speed gap between a well-optimised and a neglected site is where we see the fastest wins.
Long-form content still outperforms short-form for competitive terms. The bar has just moved. A 2,000-word article now needs to genuinely be the best thing on the topic — not just the most thorough-looking. There is a difference, and Google has gotten better at telling them apart.
Links still matter, but the calculus has shifted. A handful of links from genuinely relevant, well-regarded sources does more than a hundred from content farms. We have watched clients chase link volume for years and plateau — then pick up three good editorial links and move three positions. The quality gap is wider than it has ever been.
Local SEO, for the businesses it applies to, is still underused. Google Business Profile optimisation, consistent NAP data, genuine local citations. Not exciting. Still effective. Most of our local clients who invest here see returns faster than they do from content.
One thing that is quietly working: genuine subject matter expert content. Not thought leadership in the LinkedIn sense. Actual expertise. A dermatologist writing about tretinoin protocols. An accountant explaining R&D tax credits for specific business structures. That kind of content performs because it cannot easily be replicated, it earns links naturally, and it builds the topical authority Google has been trying to reward for years. The hard part is that it requires real input from people with real knowledge. You cannot brief an AI to be genuinely expert. You can brief a person.
What is not working the way people think
Publishing at scale. The idea that more content equals more traffic made sense when Google was less sophisticated. Now a site with 40 well-maintained, genuinely useful articles will often outrank one with 400 thin ones. We have seen it happen. The instinct to produce more is understandable — it feels like progress — but the maths has changed.
Keyword density. Nobody with a current understanding of how search works is still optimising for keyword frequency. And yet we still see briefs that specify a keyword should appear exactly eight times in 1,000 words. That is not how ranking works in 2026. Write for the topic, not the term. Cover related questions. Use natural language. The keyword will appear organically because the content addresses the subject properly. Nobody has told the brief writers.
Social signals are the other thing people keep waiting to matter more than they do. The theory has been floating around for years — if something gets shared a lot, surely Google notices. Maybe it does, at the margins. In practice, we have never seen a direct correlation strong enough to change how we plan work. What social does reliably is generate referral traffic and, occasionally, attract the kind of link that does move rankings. Do it because your audience is there. If it helps search, that is a bonus — not the mechanism.
What we are actually doing right now
We have shifted almost all of our SEO work toward topical authority. Instead of targeting individual keywords, we build clusters — interlinked content that covers a subject from every useful angle, so the brand becomes the reference point rather than just a result.
The way we structure clusters has changed too. Two years ago the approach was mostly hub-and-spoke — one pillar page, several supporting articles pointing back in. That still works, but it is not enough on its own. What we are doing now is building what we call answer depth: for any given topic, we map out every question a real person might have at different stages of understanding, then check whether we have content that genuinely covers each one. Not surface-level. Not rephrased the same three points. Actually different angles, different contexts, different levels of prior knowledge assumed.
We ran this audit for a legal services client last year. They had 90 articles on employment law, all written at roughly the same level, all targeting roughly the same reader. Someone who had never heard of unfair dismissal and someone preparing for a tribunal would both land on the same pages and find them equally unhelpful. We rebuilt the content around reader stage rather than keyword. Traffic went up 34% in five months. More importantly, enquiry quality improved — people arriving already understanding what the process involved, which shortened the sales conversation considerably. That second number never shows up in an SEO report, but it mattered more to the client than the traffic lift.
It is slower than chasing terms. It also does not evaporate after the next algorithm update. Most of the brands we see scrambling right now were playing the short game. That is not a strategy failure — it is a priorities failure. Fixable, but not quickly.
The honest summary: SEO in 2026 rewards expertise that is real, relevance that is earned, and trust that is built over time. We have clients who have been publishing genuinely useful content for three years and are now untouchable in their category. They did not do anything clever. They just did not stop.
What we tell clients who are frustrated: stop measuring rankings and start measuring whether your content is actually useful. If you would not share a specific page with someone who had the exact question it claims to answer — because it is padded, or vague, or structured around a keyword cluster rather than a real problem — then you already know what the issue is. Google is not the audience. It is the gatekeeper. Write for the person, get through the gate. In 2026 that sequence is the only one that reliably works.
The brands gaining ground in search right now are not doing anything radical. They are publishing things people actually want to read, they are keeping their sites in good technical shape, and they have built up relationships with credible voices in their space — not all three at once, usually, but steadily over time. In 2026, the margin for coasting on outdated tactics has effectively closed. The distance between brands doing this properly and those that are not keeps growing.
