Brand demand holds up where generic traffic erodes

65.3% of Google searches run from France end without a single click. The answer sits in the results page and nobody leaves it. That is what SparkToro measured on the Similarweb panel (opens in a new tab) between January and April 2026, in a study published in June. In 42.3% of cases the session stops entirely right after the query, one of the highest shares among the six countries studied.

Read at face value, the number describes a general erosion of search traffic. It actually lumps together two kinds of traffic that follow different rules. Only one of them is shrinking.

When someone searches for an answer, the results page can supply it instead of you. When someone searches for a name, nobody can answer in your place. That asymmetry has become the most structural fact in search in 2026. It can also be measured.

Amsive analysed 700,000 keywords (opens in a new tab) across ten websites and five industries, including finance, healthcare and education, isolating those that trigger a Google-generated summary. On those keywords, click-through rate falls by 15.49% on average. The breakdown matters more than the average. On keywords without a brand name, the drop reaches 19.98%. On branded keywords, click-through rate rises by 18.68%.

A second figure from the same study explains the first. Only 4.79% of branded queries trigger a summary at all. An automated summary exists to answer a question. "Lenoir practice Bordeaux" is not a question, it is an address. There is nothing to condense and nothing to substitute. In the rare cases where a summary does appear, it usually leans on pages belonging to the brand being searched for.

These measurements date from April 2025 and summary coverage has expanded since. What does not move is the mechanism. The gap between the two families of queries comes from their nature rather than from a Google setting. A query that names its destination leaves no room for an intermediary, whatever the results page looks like.

Informational traffic, on the other hand, is now negotiated with systems that answer before you do. What a citation inside those answers actually returns is the subject of another article (opens in a new tab). The short answer is far less than what gets sold around it.

A branded query does not measure search performance. It measures a memory. Someone retained a name, attached it to a need then typed it. The work that produced that memory happened elsewhere, often months earlier.

Research from the Ehrenberg-Bass Institute on mental availability describes exactly this mechanism. A brand is not memorised in the abstract, it is memorised attached to a situation. Nobody remembers "an osteopath", people remember "someone for a locked back on a Saturday". Jenni Romaniuk calls these situations category entry points (opens in a new tab) and shows that growth comes from the number of situations a name is linked to, more than from the strength of any single link.

That shift changes the question put to search. Being findable when someone looks for a solution is no longer enough. The goal is being the name that surfaces when the situation arises. The website is what turns that memory into a customer. It is also what manufactures the memory. A visit either leaves a trace or it does not. It remains the only surface a business controls entirely, the one every other channel leans on.

On 12 October 2021 the IPA think tank devoted to share of search published its findings (opens in a new tab) from 30 case studies covering 12 categories in seven countries. A brand's share of search, meaning the number of searches for its name divided by searches for every name in its competitive set, represents 83% of its market share on average. The group is careful to note these are correlations rather than causal relationships.

The value of the indicator is not in that percentage. It is in the lag. Share of search moves before market share does, as the work Les Binet presented in 2020 (opens in a new tab) and James Hankins later extended demonstrated. A business watching its share of search fall has several months before revenue carries the same signal.

The calculation needs two things. An honest competitive set, five to ten names your customers would mention spontaneously if asked who else they looked at. And a source of search volume, which depends on your size.

Google Trends returns no data below a certain search volume

That volume threshold rules Google Trends out for most small practices and SMEs. A neighbourhood practice will never reach it, however strong its local reputation. The fallback is no less useful, it is simply less comparative.

Google announced on 20 November 2025 (opens in a new tab) a filter that automatically separates branded queries from the rest in the performance report, available to all eligible sites since 11 March 2026. A card in the Insights report gives the split of clicks between the two families directly. The filter applies across every search type, images and video included.

Three details matter before relying on it.

For those sites, the regular expression filter (opens in a new tab) does the same job with a hand-written list. Search Console uses RE2 syntax, matches partially and ignores case by default.

The same filter set to does not match gives the non-branded side. These are two distinct curves to follow rather than a single ratio.

The ratio between the two families is the number everyone looks at. It is also the more misleading of the two. It rises when brand demand grows. It rises just as well when generic traffic collapses. Over the past twelve months the second explanation has become the more common one, which makes a rising ratio a poor reason to celebrate.

What should be read are both volumes in absolute terms, over a rolling twelve months. Branded impressions climbing while non-branded falls describes a business gaining recognition in a shrinking search market. Flat branded impressions with non-branded in freefall describes an entirely different situation, calling for a different response.

Direct traffic is no substitute for this measurement. It aggregates arrivals from messaging apps, email clients, PDF documents and everything the analytics tool cannot attribute to a source. A rise can come from an email campaign as easily as from growing recognition.

A business watching its branded traffic grow often wonders whether it should secure that name with ads. The question received an unusually rigorous experimental answer. eBay switched off ads on its own branded keywords across part of its market then measured what happened. The results, published in Econometrica in 2015 (opens in a new tab) by Thomas Blake, Chris Nosko and Steven Tadelis, show no measurable short-term benefit on branded keywords. The traffic lost on the paid side shifted onto the organic result. The working paper (opens in a new tab) is freely available.

The mechanism is easy to follow. Someone typing your name has already chosen you. The ad does not persuade them, it charges for a click that would have happened anyway. The same study establishes that real returns on paid search fall well below non-experimental estimates, because paid clicks and purchase intent are correlated by construction.

One case changes that calculation. When a competitor bids on your name, the ad no longer defends an incremental click, it defends an entry point. The spend is then justified. It is steered on impression share rather than cost per click, and the Google Ads auction insights report says within seconds whether the case applies.

In December 2025 Ahrefs measured, across 75,000 brands (opens in a new tab), which factors correlate with a brand's visibility in generated answers. Branded web mentions come out well ahead with a coefficient of 0.664. Backlinks top out at 0.194. The study covers correlations and says so explicitly. The order of magnitude between the two families of signals is hard to ignore all the same. Being named weighs more than being linked.

That result moves the work onto ground a small business can actually reach. Earning a link takes a negotiation with a publisher. Earning a mention takes being present where your category gets discussed. Four levers contribute directly.

The work Les Binet and Peter Field did for the IPA, built on 996 campaigns covering 700 brands and 83 categories (opens in a new tab), separates two effects clearly. Activation produces a sharp, short spike. Brand building produces a slow effect that compounds year after year. The balance observed across the most effective campaigns sits around 60% brand building to 40% activation.

That split does not transfer as is to a business with no media budget. What transfers is the tempo. Share of search is read as a rolling average over six to twelve months, never over a single month. Three weeks of data say nothing. A dashboard showing brand demand as a weekly variation mostly produces noise.

Generic traffic is negotiated with systems that change their rules without warning and keep a growing share of the clicks. The demand that names you has no intermediary to convince. It builds slowly, it is measured for free inside a tool you already own. It belongs to the business rather than to the platform.

It is the one line of search nobody else holds the key to. The website remains the foundation that turns it into customers, and the experience it produces is what makes a name stick long enough to be typed one day into a search bar.