The traffic pattern almost everyone is seeing

The support ticket arrives in the same shape every time. Rankings are stable or improved. Impressions are flat or up. Sessions are down twenty, thirty, forty percent. Nothing was changed, no penalty landed, and the drop does not map to any deployment.

This is not a mystery and it is not, in most cases, a problem with the site. It is the structural change that has been building since AI Overviews rolled out and has now fully arrived: rank and clicks have decoupled. Holding position three is no longer the same asset it was two years ago, because the results page increasingly answers the question before position three is ever reached.

What the 2026 zero-click numbers say

The figures are consistent enough across trackers to be treated as real, even where the exact percentages differ by methodology:

  • Roughly 60% of searches globally now end without a click, rising to about 77% on mobile.
  • The share of searches producing at least one click fell about 9.5 percentage points between 2024 and 2026 — a decline of roughly 23% in relative terms.
  • Publisher referral traffic from Google is down around 38% year over year.
  • On queries where an AI Overview appears, organic click-through has fallen sharply — some measurements put the drop above 60%.
  • AI Mode refers traffic at roughly 1.6–2.5% of queries, against 17–19% for traditional Google search.

That last figure is the one to sit with. It is not that AI search sends less traffic; it is that AI search sends an order of magnitude less traffic per query. Any strategy built on the assumption that visibility converts to sessions at historical rates is working from a broken model.

The losses are not evenly distributed

Aggregate numbers hide the useful detail. The damage is concentrated almost entirely in content whose job an AI summary can do.

Definitional content — "what is X", "how does Y work", "X vs Y explained" — has been hit hardest, because a two-paragraph synthesis genuinely satisfies that intent. Listicles and thin roundups are close behind.

What has held up: pricing pages, comparison pages where the user needs to evaluate their own situation, local and "near me" searches where the outcome is a visit or a call, service pages, product pages, and anything where the next step is a transaction rather than an understanding. An AI Overview can tell someone what a headless CMS is. It cannot get them a quote for one.

Which means the strategic response is not to write more of what is being eaten. It is to move the centre of gravity of your content programme toward queries where a summary is not a substitute for a visit.

Being the source is now a real outcome

The harder shift is accepting that a citation without a click has value. When someone asks an assistant to recommend agencies in your category and your studio is named, that is a commercial event even though your analytics will never see it. It shows up later, as a branded search or a direct visit or an enquiry that says "you were recommended to me".

This is measurable, imperfectly. Branded search volume in Search Console is the closest proxy most businesses have, and it is the metric to watch when unbranded traffic is falling. If branded search is rising while sessions fall, your visibility is working; the delivery mechanism changed. If both are falling, you have a genuine problem.

We go deeper on how to earn those citations in our guide to generative engine optimisation, and on how to check whether it is working in tracking your brand's visibility in AI search.

What to do about it, in order

  1. Re-baseline before you react. Compare like for like: segment traffic by query intent, not in aggregate. If your informational cluster is down 40% and your commercial cluster is flat, that is a completely different situation from a site-wide decline, and it calls for a completely different response.
  2. Stop judging content on sessions alone. Bring conversion rate and lead quality into the same view. Many businesses are seeing fewer visitors converting at higher rates, because the casual researchers are being filtered out upstream. That is not a loss.
  3. Shift new production toward decision-stage intent. Pricing, comparisons, "how to choose", objection handling, and genuinely specific advice that a generic summary cannot replicate.
  4. Make what you have easy to cite. Clear headings that answer the question they pose, direct answers in the first sentence under each heading, specific numbers and dates, and no burying the point under three paragraphs of throat-clearing.
  5. Invest in the channels search cannot mediate. Email, community, referral, and direct relationships. This has always been sensible and is now urgent.

What not to do about falling search traffic

Do not respond to a traffic decline by publishing more of the content that is declining. The instinct when sessions fall is to increase output, and in this particular case volume on informational keywords is close to a rounding error against an AI summary that answers the query in place.

Equally, do not conclude the site is broken and start rebuilding. If you have genuinely ruled out a technical cause — and our indexing checklist is the fast way to do that — then the platform changed, not your website. The fix is editorial and strategic, not structural.

And if the traffic you do get is not converting, that is a separate and more solvable problem — start with traffic but no leads, or send us your numbers and we will tell you which of the two you actually have.