Search stopped sending traffic, and that changes what you publish

The numbers on search referral have moved far enough that the old strategy is no longer conservative — it is simply wrong.

Zero-click searches reached 68% in early 2026. Google referral traffic to publishers fell 38% year on year as of January 2026. Where an AI Overview appears, organic click-through drops by around 61%, and the top result loses roughly 58% of its clicks.

Most instructive of all: "how to do X" queries are cannibalised almost completely, while transactional queries — "buy X", "price of X", "best X for Y" — are barely touched.

That last split tells you what to do.

The tutorial is dead as an acquisition channel

If your content strategy is "write helpful how-to articles and rank for them", the channel you built it on is closing. Not because your writing got worse, but because the answer now appears above the results and the user's question is resolved before a click exists.

This is worth stating without drama. It is not a penalty and it is not recoverable by better SEO. The query intent is being satisfied at the surface. For a generic how-to, that is arguably correct behaviour — the user wanted the answer, not your page.

What survives is content where the page itself is the point: something with data the model cannot synthesise, a judgement that carries a name and a track record behind it, or a decision the reader has to make and wants a specific opinion on.

Being cited is not the same as being visited

The strategic shift is from optimising for clicks to optimising for citation. Those are different goals with different tactics, and conflating them wastes effort.

A model citing you needs to be able to attribute a claim. That rewards things a keyword strategy never did: specific numbers with a stated source, clear authorship with a real role, dates, and structured markup that says unambiguously what this document is and who wrote it. Vague, comprehensive, keyword-hedged prose — the format that ranked well for a decade — is close to unusable as a citation because there is no crisp claim to attribute.

The uncomfortable part is that citation without a click produces no analytics event. You will be recommended in conversations you cannot see. Teams that judge content purely on sessions will conclude the work is failing at exactly the moment it starts working.

What that means practically

Write fewer, more specific pieces. A post that says "we measured this and here is the number" can be cited. A post that surveys a topic comprehensively cannot, because there is no single claim to lift.

Put your name and role on it. Attribution requires an entity to attribute to.

Mark it up properly — author, publication date, headline — so a machine reading the page does not have to guess. This is unglamorous and it is most of the technical work.

Stop treating traffic as the only success measure. Direct visits, branded search volume, and people arriving already knowing what you do are the signals that survive the shift. So is the plainest evidence of all: a client saying they were pointed to you.

The honest uncertainty

Nobody outside the model providers knows exactly what earns a citation, and anyone selling certainty about it is guessing. The published statistics vary by source and methodology, and this is a fast-moving area where a number six months old may already be wrong.

What looks durable is the direction: the intermediary that used to send you visitors is increasingly answering on your behalf. Content whose only value was being the page that answered a common question has lost its distribution. Content that contains something specific enough to be worth quoting has, if anything, gained.

That is not a comfortable transition for anyone whose traffic came from the first category. But it does favour writing something true and particular over writing something comprehensive and safe, which is not the worst incentive the web has ever had.