Traffic is falling on pages that are performing better than ever. That is not a contradiction. It is a reporting problem, and it has a fix.

In short

• Visibility and clicks have separated. A brand can be more visible than ever and send less traffic than last year.

• Click-based reporting now understates performance, and the gap is widest at the top of the funnel where discovery happens.

• Four measures replace the traffic line: branded demand, share of answer, qualified enquiry volume and self-reported source.

• The teams handling this well changed their reporting before they changed their tactics.

  
  

Somewhere in most marketing dashboards there is a line going down that nobody can fully explain. Sessions are off. Impressions are flat or higher. Enquiries are steady, or even up. The team is doing better work than last year and the headline chart disagrees.

This is the zero-click economy, and it is not a passing anomaly. When a search result page answers the question above the links, a summary satisfies the reader, or a comparison is settled inside a platform the buyer never leaves, the click that used to prove your work existed simply does not happen. The influence still happened. The measurement did not.

What the data actually shows

The most useful study on this so far tracked real browsing behaviour rather than opinions. Pew Research Center followed nearly 69,000 searches from 900 US adults and compared what people did when a result page carried an AI-generated summary against when it did not.

Read those three numbers together and the shape of the problem is clear. The answer is being consumed, not clicked. The reader is satisfied and leaves. Your brand may well have been named in that answer, and your analytics will record nothing at all.

The other half of the picture is more encouraging, and most teams have not noticed it. Adobe's analysis of US retail sites found that visitors arriving from AI assistants converted 42% better than other traffic in March 2026, spent 48% longer on site and viewed 13% more pages. A year earlier the same source converted at roughly half the rate of everything else.

Fewer visits, each worth considerably more. That is not a decline. It is a change in the unit of measurement.

Why traffic reporting breaks first at the top

The damage is not evenly spread. Transactional pages are largely fine, because someone ready to buy still needs to reach a checkout, a form or a booking screen. It is the informational layer that empties out: the explainer, the comparison, the cost breakdown, the eligibility page.

Those pages were never valued for their conversion rate. They were valued because they introduced the brand early and cheaply. When their traffic falls, a purely click-based reporting model concludes that the content programme has failed, and the budget moves to whatever still shows a measurable click. Six months later the pipeline thins out, and nobody connects the two events because the reporting was never built to connect them.

The four measures that replace the traffic line

None of these are exotic. All of them can be built with tools a mid-sized marketing team already has.

1. Branded demand

   Count how many people search for your brand by name, month on month, and split it out from your total organic numbers. Branded search is the cleanest available proxy for whether your upper funnel is working, because nobody types your name unless something put it in their head. If your explainer traffic falls while branded demand rises, your content is doing its job and the click is simply happening later, somewhere else.

   Track it as a trend against your own history, not against a competitor benchmark. Seasonality matters more than you expect, so compare like months.

   
   

2. Share of answer

   Write down the fifty to two hundred questions your buyers genuinely ask, in their words. Ask those questions across the assistants and summary surfaces your market uses, on a schedule, and record whether your brand is named, which sources the answer was built from, and whether what was said about you is accurate.

   
   

   This gives you a percentage you can trend and defend: out of the questions that matter to our business, how often are we in the answer? It is the closest thing the zero-click world has to a ranking, and unlike a ranking it also tells you what is being said.

   
   

3. Qualified enquiry volume, matched in the CRM

   If the click is unreliable, the enquiry is not. Every form, call and chat still lands somewhere. The work is joining it back: agreeing one definition of a qualified enquiry with your sales team, then tracking that number monthly alongside spend.

   
   

   Most organisations discover an uncomfortable gap here. The platform reports one number, the CRM holds a smaller one, and the number sales would call real is smaller still. That gap is not a measurement failure. It is the most valuable diagnostic you own.

   
   

4. Self-reported source

   Add one optional question to your enquiry form: how did you first hear about us? Free text or a short list, never mandatory.

It is imperfect, biased towards whatever the person remembers most recently, and still worth doing. It is the only instrument that can see the podcast, the community thread, the colleague's recommendation and the answer engine that named you, none of which will ever appear in an analytics platform. Run it for a quarter and compare the pattern against your attribution model. Where the two disagree, your attribution model is usually the one that is wrong.

Practical

A board report that survives the change

One page, four blocks, no session counts on it anywhere.

• Demand. Branded search trend and share of answer, both as a twelve-month line.

• Efficiency. Cost per qualified enquiry and cost per closed customer, by channel.

• Quality. Enquiry to qualified rate, and qualified to closed rate, so a cheap but useless source is visible as the expense it is.

• Evidence. Self-reported source for the quarter, quoted rather than summarised.

What not to do

Three responses are common and all three make the problem worse.

• Cutting the content that stopped sending traffic. If those pages are being read inside summaries and answers, cutting them removes you from the answers too. Check whether you are being cited before you decide you are being ignored.

• Blocking machine access to protect the content. Understandable, and usually applied with a blunt instrument. The crawler that collects data for model training and the one that fetches your page to build a live answer are frequently different user agents. Blocking both removes you from the answer while doing very little about the training question.

• Moving the whole budget to the channels that still report clicks. This is the expensive one. It concentrates spend on people who are already in the market, which is the smallest and most competitive audience available, and quietly stops the work that creates demand in the first place.

The order to do this in

Change the reporting before you change the tactics. A team that shifts strategy while still being measured on sessions will be punished for doing the right thing, and will revert within two quarters.

A workable sequence:

• Agree the definition of a qualified enquiry with sales, in writing, before anything else.

• Add the self-reported source question this month. It costs nothing and it starts collecting history immediately.

• Separate branded from non-branded search in every report from now on.

• Build the question set and start recording answers, even manually at first. A spreadsheet updated fortnightly beats a perfect system you never build.

• Only then reallocate budget, using the new numbers rather than the old ones.

  
  

The uncomfortable truth is that the click was always a proxy. It was a convenient, countable stand-in for interest, and for twenty years it was good enough. It is becoming less good, gradually, and the teams that notice early get to rebuild their reporting calmly rather than defending a falling chart they cannot explain.

Frequently asked questions