Appearing in an AI Overview: 10 sectors measured

You want your company to appear in the block Google displays above the links. First question, rarely asked: does Google produce one at all, on the searches that concern you?

Since April 2026 we have been asking the same questions every day across ten sectors of the Swiss-French market. More than 2,300 measurements later, the average is 73.4 %. It says little once you look sector by sector.

First, check whether Google answers in your sector

Swiss Atlas, April to August 2026. Between 234 and 240 questions per sector, asked daily.

Forty-seven points separate the two ends. Google answers readily on what can be explained: a price, a procedure, an opening time, an insurance cover. It holds back on taste, experience and the high end.

A restaurateur and an online bank are therefore not playing the same game. The first still has one search in two passing without a generated block, and classic search optimisation keeps its place there. The second is described by Google in almost nine searches out of ten.

The format that works is not the same everywhere

The editorial article dominates in nine sectors out of ten. Restaurants are the exception.

Swiss Atlas. Share of the leading page type in each sector's citations.

In restaurants, directory listings come first, fr.tripadvisor.ch at the top, ahead of the article at 17.8 %. A restaurant investing in a blog is working the second format, not the first. Its listing, its reviews and its opening hours weigh more.

Two other sectors have a notable second format: retail with the comparison at 14.0 %, telecommunications with the comparison at 17.8 %. Elsewhere, the product page comes second.

You cannot cite yourself

Across the whole corpus, 97.1 % of the citations in an AI Overview point to a third-party site. By sector, the share going to the company's own site runs from 0.1 % to 7.2 %.

Swiss Atlas, AI Overview citations, April to August 2026. Company subdomains included.

The best placed tops out at 7.2 %. At the other end, in luxury watchmaking, of 798 citations collected, a single one pointed to the brand's own site. The other 797 came from resellers, forums, videos and the specialist press.

The gap is not accidental. The sectors at the top of the table publish information nobody else holds: train timetables, product ranges, course programmes. Those at the bottom are described by an ecosystem that gets along without them.

That does not mean your site is useless. It remains what Google reads, it converts, and it keeps its place on purchase searches where the block barely appears. But if your goal is the generated answer, the work happens mostly elsewhere: comparison sites, newspapers, trade directories, forums.

How many doors there are to push

The number of sites to work, and their relative weight, change a great deal.

Swiss Atlas. Number of distinct sites cited and concentration on the top five.

Two opposite work plans read here. In retail or restaurants, a few targets weigh heavily and the list is short: five sites capture nearly a quarter of citations. In travel and tourism, the top five make only 9.4 % across 269 sites. There is no main door, the effort spreads.

The sites that matter, sector by sector

Here are the five most cited sites in each sector, excluding general platforms and the measured brands' own sites. This is the starting list for press relations work.

Swiss Atlas, AI Overview citations, April to August 2026.

Three useful observations. Swiss-French comparison sites come first in regulated sectors, insurance and telecoms in particular. Reddit appears in six sectors out of ten, often in the top five, while almost no Swiss SME writes there. And in watchmaking as in restaurants, the selection is made by highly specialised sites, some run by a single person.

These are newsrooms and teams, not forms.

How long before judging

Across the whole corpus, 59.3 % of cited pages are cited on a single day. Here too the sector gap matters.

Swiss Atlas. Share of URLs cited on a single distinct day.

In health insurance, one cited page in two returns on the following days. In luxury watchmaking, seven in ten disappear after a day.

The consequence is practical. The more volatile your citations, the more weeks before a gain can be told apart from a fluctuation. A single test says nothing anywhere, and even less in the sectors at the top of this table.

The work plan that follows

Four questions in order, with the answer for your sector in the tables above.

Does Google answer in your sector? If the rate is low, part of your traffic still goes through the classic links, and search optimisation keeps priority there.

Which format is retained? Editorial article almost everywhere, directory listing in restaurants, comparison second in retail and telecoms.

Which sites to target, and how many? A short heavy list in retail and restaurants, a long flat one in travel.

Over how many weeks to measure? The higher your sector's volatility, the longer the window.

These figures describe one vertical per sector, measured daily. They give orders of magnitude and a method, not the detail of your market. For how the block itself works, see AI Overview and when it triggers. If you hand this work to a provider, the six questions to ask before signing.

Method and limits(5)

1. Corpus. Swiss Atlas, Repliq's observatory across ten sectors of the French-speaking Swiss market, questions asked in French, 20 April to 12 August 2026.
2. One vertical per sector. Each sector is observed through a real company of the Swiss-French market, tracked daily, with 234 to 240 questions. It serves as a probe and is not named: what we publish is the source landscape around its questions, not an assessment of that company. None is a Repliq client.
3. Trigger. A measurement counts as triggered when the search returns an AI Overview with non-empty text.
4. Volatility. A page counts as cited on a given day if it appears in at least one answer that day. Days are distinct, not consecutive.
5. What these figures do not say. They describe French-speaking Switzerland and do not transpose unchanged.

See the sources cited in French-speaking Switzerland→