Is SEO dead? No — but it stopped being the whole job

Where the claim comes from

"SEO is dead" has been published every year since at least 2008. It was declared dead by social media, by mobile apps, by voice search, by Google's answer boxes, and now by AI assistants. The claim performs well because it is frightening and short, and because the people most likely to repeat it are selling the thing that supposedly replaced it.

This time there is a real change underneath the headline, which is why the slogan has more purchase than usual. The change is not that search stopped working. It is that the step after ranking — the person looking at a list and choosing — is being done for them more often than it used to be.

Both things are true at once, and most writing on the subject picks one. Search is still the largest single source of commercial intent on the internet. The assumption that ranking equals winning is no longer safe. Holding both is the whole of the grown-up position.

What has genuinely changed

• The answer often arrives without a click. An AI summary at the top of the results page, or an assistant answering directly, settles the question in place. The impression happened; the visit did not.
• Ranking and being named have come apart. You can hold the top organic position for a phrase and still be absent from the answer to the equivalent question, because the model is leaning on a directory or a roundup rather than on the page that ranks.
• Thin informational content has lost its job. A page whose purpose was to define a term or convert a unit was always borrowing traffic from a question with a one-line answer. That loan has been called in.
• The shape of the result changed. Ten links reward being in the running. Two or three names reward being the obvious choice. Those are different targets, and the second one is less forgiving.
• Measurement broke at the edges. Impressions and clicks still exist. The recommendation that went to a competitor inside somebody's private conversation does not appear anywhere, which is the subject of the zero-click problem.

What has not changed at all

• Most traffic still comes from search. For the overwhelming majority of small and medium businesses, organic and paid search remain the largest channels by some distance. A plumber in Brisbane is still found mostly by people searching for a plumber in Brisbane.
• Crawlability is still the price of entry. If a machine cannot fetch and parse your pages, nothing downstream matters. That was true for Googlebot and it is true for the newer crawlers.
• Third-party corroboration still decides who looks credible. Reviews, citations, consistent business details, mentions in places people trust. The same signals, read by a different kind of reader.
• Clear, specific, genuinely useful writing still wins. If anything the premium on it went up, because a model quoting you is choosing the clearest available statement of a fact.
• Transactional and local intent still converts through a visit. Nobody books a dentist, buys a sofa or hires a solicitor inside a chat answer. The answer narrows the field; the site still has to close.

Why the honest answer is boring

The sober version is that one job split into two. The old job — be findable, be credible, be clear — did not go anywhere. A second job appeared alongside it: be quotable, and know whether you are being quoted.

That is a much less exciting sentence than "SEO is dead", and it is the reason the accurate account travels less well than the hot take. It is also what anybody deciding next year's budget actually needs, because the two framings lead to opposite decisions and only one of them is defensible.

Be careful with the evidence on both sides

The debate is unusually badly served by its statistics. Studies of "zero-click" rates measure different things, count different query types, and often come from companies with a position to defend. Agencies showing catastrophic traffic graphs rarely mention what kind of pages those were. Tool vendors showing no effect at all are usually averaging across clients who were never exposed in the first place.

Google has said that it sees AI features driving quality traffic, and publishers have said loudly that their referrals fell. Both can be accurate descriptions of different parts of the web. Neither tells you what happened to a dental practice in Leeds or a software company selling to finance teams.

The only dataset that settles the question for you is yours. Segment your organic traffic by page type — informational, commercial, branded, local — and look at the trend in each separately. An average across all four hides exactly the movement you need to see. In most of the accounts we look at, the damage is concentrated in one segment and the rest is steady, which is a very different problem from a general collapse and calls for a very different response.

Rollout timing matters too. Google's AI features arrived in different countries at different times, and assistant availability still varies by market. If your traffic changed sharply in one month, check whether something launched in your market that month before concluding anything about your content.

Who should actually worry

The impact is uneven, and the people loudest about it are not always the people most affected. Three profiles are genuinely exposed.

Publishers of informational content monetised by advertising, whose traffic depended on questions an assistant can now answer in a paragraph. Affiliate sites built on "best X" roundups, where the model may read the roundup and skip the site. And any business whose entire content strategy was a library of near-identical location pages, which were already a liability and are now simply ignored.

Far less exposed: service businesses with a real physical presence, anything with a complicated or high-value purchase, anything where the customer needs reassurance before committing, and anybody whose traffic was already mostly branded or referral. If that is you, the correct response is attention rather than alarm.

What to do differently from last year

Keep doing the SEO that works. Then make three adjustments, none of which require abandoning anything.

First, write at least some pages for a reader who will never arrive. If a model is going to summarise your sector, you want the clearest statement of the relevant fact to be yours. That means plain declarative sentences about what you do, for whom, where, and at what sort of price — the things people actually want to know and most sites bury.

Second, measure the second surface. Rank tracking tells you nothing about whether an assistant names you. Those are separate questions with separate answers, and the gap between them is often the most interesting thing a business learns all year. The reasons behind that gap are worth understanding in detail, which is the subject of ranking well while never being named.

Third, stop treating position as the goal. Position is a proxy that worked for two decades and has started to mislead. The goal was always to be the business a potential customer ends up choosing; ranking was just how you got in front of them.

The comparison worth reading next

This piece is the "should I panic" answer, and the short version is no, but do pay attention. The mechanical comparison — what carries over from your SEO work, what does not, and exactly why page one is not enough — is set out separately in AI visibility versus SEO, which is the more practical read if you have already accepted the premise and want to know what to change.

One last thing worth saying plainly, because it cuts against the interests of everybody writing about this topic, ourselves included. Nobody can tell you with confidence how this looks in three years. The models change, the interfaces change, and the commercial arrangements behind them are unsettled. Anyone selling you certainty about 2029 is selling you a story. The defensible position is to keep the channel that works, add the measurement you are missing, and revisit it honestly every few months.