SEO is no longer just about gaining positions on Google. Between responses generated by AI and automated advertising, SMEs must rethink their visibility. A question becomes central: how many customers can these different levers really acquire?
Google AI Insights: When search is done without clicking
Online search has changed in just a few months. In March 2026, 33 million French people used an AI service, or 57.1% of the population (Arcom). Since July 22, 2026, Google has also been deploying its AI Previews in France: on some searches, an AI-generated summary is now displayed above the links. This change weighs on clicks.
In the United States, where AI Previews have existed since May 2024, 68% of Google searches ended without any clicks at the start of 2026, compared to 60% two years earlier (SparkToro). In France, the effect was immediate: nine days after the launch, the sites most exposed to AI Insights had already lost 23.1% of their click-through rate (Ahrefs). Over one year, the sites of the 40 information brands monitored by the ACPM lost 22.9% of visits between September 2025 and September 2026 (Digital Century).
For these news sites, the link with AI Previews remains to be established. But the trend is enough to worry businesses: what good is a well-positioned site if fewer and fewer Internet users click on it? For an SME, whatever its sector, investment in online visibility must remain profitable.
It is this question that led Premiere.Page, an SEO agency based in Anglet and active since 2013, to become a Search Marketing agency. His bet: to bring together SEO, SEA and GEO (and soon advertising in AI) in the same team, with a single measurement for all, the cost per customer.
“An SME does not only ask the question of positions or clicks,” summarizes Jérôme Tellechea, director of Premiere.Page. “If she invests in SEO, she asks herself how many customers it brings her, and at what cost. »And this, in a particular context: AI is shaking up the way French people search and buy online.
SEO, SEA, GEO and soon GEA: SEO becomes global
When it comes to online visibility, many companies still hesitate between natural and paid referencing. A third subject now concerns managers: how to be visible in Claude or ChatGPT? In other words, how does a brand manage to be mentioned in generative AI responses?
This natural referencing strategy in AI is called GEO, for Generative Engine Optimization. It now has its paid version, GEA (Generative Engine Advertising), like SEO has SEA: since August 24, 2026, ChatGPT has displayed advertising in France and in 30 other European countries (OpenAI). And AI is starting to sell: in the United States, Google has announced that you will soon be able to book a hotel directly from its AI Mode.
This is why these levers must be managed together, then adjusted according to the results. Each responds at a different moment in the search:
-
SEA (Search Engine Advertising) makes the company visible as soon as the ads are validated. He then needs a few weeks of learning before judging what he brings back. This is the lever to activate for a one-off campaign or a peak season.
-
SEO (Search Engine Optimization, natural referencing) builds visibility over time. It takes longer to establish, but it persists even when the advertising budget is cut.
-
GEO (Generative Engine Optimization) makes the company visible in AI responses. It relies largely on SEO assets: in France, 60% of pages cited by AI Previews come from the first page of Google, according to Publicis Connected Media (The Media Leader).
Taken separately, these levers compete for the budget. Piloted together, they form a virtuous circle. For an SEA agency that also manages natural referencing, the question is not “natural or paid referencing?” “, but “what leverage for what request, and until when? “.
SEA and SEO: accelerate rather than compete
Between SEA and SEO, it is not a question of competition, but of acceleration. As long as a keyword is not gained in naturalness, the advertisement occupies it. Once the site ranks on the query, the budget can shift to searches that the brand has not yet conquered. “Once we are placed in SEO, we withdraw and look for other positions,” explains Kaïs Faure-Brac, SEA Team Leader at Premiere.Page.
SEA also feeds SEO and GEO strategies. Google Ads campaigns reveal the keywords buyers use in just a few weeks. A heat pump installer may discover that “old house boiler replacement” brings more qualified quotes than “heating”.
These searches that convert also designate the questions to follow in AI, the “prompts”. Premiere.Page monitors them with Otrak, its citation tracking tool in ChatGPT, Gemini and Perplexity.
What’s changing for SMEs: measure before spending
Without conversion tracking, a campaign is judged by the number of visits, which says nothing about sales. We then fly blind. At Premiere.Page, monitoring therefore occurs before ads: GA4, Google Tag Manager, primary and secondary conversions, connection with the CRM, Merchant Center for e-commerce, etc. For an SME, effective SEO begins with an audit and by calculating the cost per customer.
Before spending a euro, an SEA strategy is based on local market figures: search volume, cost per click, forecast conversion rate, then margin generated by a customer. A query can only be purchased under its ceiling price, and each campaign responds to a single intention. The projection is made in the low scenario, taking into account the SME sector.
AI Max and ads in ChatGPT: test before adopting
Advertising is becoming automated at the same pace as search. Starting September 2026, Google is migrating Search campaigns that use campaign-level broad match or auto-created elements to AI Max. According to the timetable announced by Google, Dynamic Search Ads will follow in February 2027. These functions can broaden targeted queries and adapt texts or landing pages to the search context.
These tools learn from what is declared to them as success. If a simple started form counts as a “conversion”, Google will look for more people who start a form, not customers. You must therefore only report to them what is really profitable (a qualified request, a reservation, a sale), then check each week on which searches and on which pages they are spending the budget.
The same caution applies to new formats. Ads in ChatGPT, opened in France at the end of August 2026, were first tested by Premiere.Page on its own budget, and are not yet offered to its clients: if the format does not keep its promises, it is the agency that loses budget, not its clients.
The right indicator: cost per customer, not cost per prospect
A flattering cost per contact can hide a loss-making acquisition. Let’s imagine an SME which spends 1,800 euros on advertisements, receives 60 requests and signs 12 contracts. Its cost per contact is 30 euros, but its advertising cost per customer reaches 150 euros, before internal costs. If the margin on a contract is less than this amount, the lead volume masks a profitability problem.
The reasoning also applies to SEO, whose content and maintenance have a cost, and to GEO, whose commercial effect remains difficult to isolate. The measure will also remain partial: non-essential advertising trackers are in principle subject to prior consent, recalls the CNIL. Linking CRM, audience analysis and campaigns makes it possible to compare the conversions reported by Google Ads with the sales observed, without confusing attributed sales and additional sales.
The winning combination in 2027: a single piloting
For an SME, winning in 2027 will not come through an additional lever, but through a working order. The one that Premiere.Page follows is in five steps:
-
Measure the measurement before any expenditure: conversions, CRM, account in the name of the company.
-
Launch SEA on commercial requests, under a ceiling price, one campaign per intention.
-
Build SEO on keywords that have proven to sell, and remove the ad when natural takes over.
-
Publish verifiable facts to be cited by AIs, and track those citations.
-
Read everything by cost per customer, not by the number of clicks, positions or leads.
As Google automates its bidding and answers questions directly, the quality of business data could matter as much as a site’s position. This is the whole point of a 360° approach: reading ads, SEO and visibility in AI from a single number.

