What did the South African zero-click study find?

Algorithm, a performance marketing agency, analysed 16 months of Search Console data from 110 South African websites run by 71 businesses. Comparing 19 May to 30 June 2026 with the same weeks of 2025, the typical business's clicks per 100 Google appearances fell from 4.77 to 3.03, a stated 29% fall, while appearances rose and clicks fell.

How reliable is the study?

It is useful and unusually candid about its limits, but it is an agency's analysis of its own clients, so it does not represent all South African search. Appearances are not searches, sector groups are small, the AI Overview check is a single snapshot per term, and Search Console does not show what people did instead of clicking.

Why does the stated 29% not match 4.77 to 3.03?

Because the stated percentage is the typical change within each business, not a calculation from the two medians shown. Taken directly, 4.77 falling to 3.03 is about 36%. Both numbers describe the same data; they answer different questions, which is why we quote the definition whenever we quote the figure.

What should a South African business change in its reporting?

Report visibility and clicks as separate lines, split branded from non-branded searches, and give one person ownership of SEO, answer engine optimisation and generative engine optimisation so they are planned and reported together. A rising impressions line beside a flat clicks line is not necessarily a ranking problem.

South African Businesses Are Being Seen More and Clicked Less. What a South African Dataset Shows, and What It Cannot.

Algorithm, a South African performance marketing agency, published a South African dataset on zero-click search in October 2026. Its central finding is that the typical business in the sample is appearing in Google more often and getting fewer clicks for each appearance. The study is careful about its limits, so this piece sets out what it supports, what it cannot show, and what we saw on our own, much smaller, site.

What was studied

The sample is 110 websites run by 71 businesses across eight sectors, covering 1.43 billion Google appearances, 67.3 million clicks and 9 million distinct search terms. The window runs from 19 May 2025 to 28 September 2026. The primary comparison is 19 May to 30 June 2026 against the same weeks of 2025.

One definition governs how to read every percentage in the report. Changes are the typical change within each business: each business’s sites are combined, and the stated figure is the median of those per-business changes. They are not calculated from the medians shown beside them.

What it found

For the typical business, comparing those weeks:

Clicks per 100 appearances, all devices: 4.77 fell to 3.03, a stated fall of 29%. Calculated directly from those two numbers it would be about 36%, which is the definition at work, not an error.
Appearances and clicks, all devices: appearances rose 38% while clicks fell 9%.
Mobile: clicks per 100 appearances went from 3.98 to 3.10, a stated 21% fall. Mobile appearances rose 22% and mobile clicks fell 8%.
Desktop: a stated 49% fall, but the report says desktop appearances look inflated by automated traffic since September 2025, so it leads with mobile.
Branded searches: clicks per 100 appearances went from 19.04 to 15.59, a stated 16% fall.
Non-branded page-one terms: 1.96 fell to 1.27, a stated 31% fall.

Branded searches are 13% of appearances and 75% of clicks, based on 60 businesses between June and August 2026. In this sample, then, most clicks come from people who already know the name, although the report notes the branded share of clicks is probably slightly overstated.

What it says about AI Overviews

Algorithm checked 5,850 non-branded page-one terms and found an AI Overview on 4,100 of them, 70%. That is a single snapshot per term, not a history. By funnel stage the share was 75.4% for learning searches, 74.5% for buying, 77.0% for comparing and 34.9% for navigating.

For 32 businesses with the same page-one terms in both years, clicks per 100 appearances on terms that now show an AI Overview fell from 1.56 to 1.35, a stated 22%. On terms without one the stated fall was 9%. At position one, between June and September 2026, the combined figure was 6.5 clicks per 100 appearances with an AI Overview and 20.4 without. The within-business ratio was 0.73, based on only 12 businesses, and the report says to read the gap as a range. We would too.

Sectors, with a warning

The stated change in clicks per appearance ran from 15% to 50%: leisure, education and property fell 50%, financial services 37%, FMCG and beauty 33%, industrial and B2B services 29%, home and DIY retail 29%, fashion and health retail 15%, and automotive 15%. The report says its sector groups “show direction rather than precise benchmarks”, with between 5 and 25 businesses in each. The industrial and B2B group is 25 businesses, which is the closest to our own market, and it is still a small sample.

The part we found most useful: three breaks in Search Console data

The report found three points where the data itself shifted, and it avoids comparing across them:

1–2 July 2025: average position moved from about 7.5 to about 20, and clicks per 100 fell from about 4.6 to about 3. The report says this points to automated tools loading results pages.
10–15 September 2025: reversed when Google stopped serving 100 results per page to rank-tracking tools.
Mid-December 2025 to late April 2026: the share of non-branded appearances at position one rose from about 10% to 24%, then unwound between 16 and 28 April 2026. The report says it does not know the cause.

If you have been reading a Search Console chart as a clean record of demand, this is the argument for annotating it. Our companion piece on altered results shown to rank trackers covers a separate, current disruption to third-party data.

What it cannot show

The report states its own limits, and they are real:

The sites are Algorithm’s clients and lean toward the sectors it works in, so this is not all South African search.
Appearances are not searches. Two of the sample’s sites showing on one results page count as two appearances.
Search Console does not show what searchers did instead of clicking.
The AI Overview check is one snapshot per term.
Search Console names only some terms: for the typical business about 65% of appearances carry a named term.

We read the findings as correlations: the study shows clicks per appearance falling alongside AI Overviews appearing, and it does not show cause.

Our own numbers, as a small illustration only

Between 1 August and 31 October 2026, doublebarril.com recorded 33 clicks and 12.1 thousand impressions in Search Console, at an average position of 42. For 4 to 31 October it was 7 clicks and 1.11 thousand impressions, at an average position of 25.2.

That is a very small sample at deep positions, mostly for definitional queries, and it should not be compared with Algorithm’s typical business. It illustrates something different: impressions at deep positions rarely turn into clicks. We have not checked how many of those queries show an AI Overview, so we draw no conclusion about zero-click behaviour from our own numbers.

What to change in your reporting

Report visibility and clicks as separate lines. A rising impressions line beside a flat clicks line is a different problem from a falling rank.
Split branded from non-branded. In Algorithm’s sample about 75% of clicks were branded, and the non-branded fall was larger (31% against 16%). A blended click rate hides what happens to the searches where you are being discovered.
Mark known breaks on the chart, including the three above and, for third-party position data, 13 and 20 September 2026.
Give SEO, answer engine optimisation and generative engine optimisation one owner, and report them together. That is the recommendation Algorithm makes, as Bizcommunity reported it, and we agree: when they sit in three plans, the impressions go to one team and the explanations to another.