The decline in organic click-through rates on Google, accelerated by the rollout of AI Overviews, changes the game for any website that relies on natural referencing. Analyzing and improving your website’s SEO is no longer just a classic optimization approach: it is a technical response to a measurable erosion of organic traffic that we have observed over the past few months across most sectors.
Crawl budget and JavaScript rendering: the blind spots of technical SEO
The majority of SEO analyses focus on title tags, meta descriptions, and loading speed. These elements matter, but they mask more structural issues.
The crawl budget, which refers to the number of pages Googlebot explores during each visit, remains underutilized in current audits. A site that generates thousands of parameterized URLs (catalog filters, duplicated paginations, sorting versions) wastes this budget on pages with no SEO value. The result: strategic pages are crawled less frequently, and their updates take longer to be indexed.
JavaScript rendering poses a comparable problem. Google executes JS in a separate queue, with variable delays. A site whose main content relies on a client-side framework (React, Vue, Angular without SSR) risks having its pages indexed partially, or not at all. We recommend checking the rendering via the URL inspection tool in Search Console, comparing the raw HTML and the rendered DOM.
Platforms like virank.fr allow for cross-referencing this technical data with ranking tracking, which speeds up the identification of under-crawled or poorly rendered pages.
AI Overviews and the decline of organic CTR: adapting content to generative search
A Similarweb study cited by In Data Veritas indicates that the top 500 American media outlets lost 27% of Google organic visits between February 2024 and February 2025. This drop coincides with the rise of AI-generated responses directly in search results.
Google has confirmed the rollout of AI Overviews in France for summer 2026. The direct consequence: an increasing share of informational queries is resolved without a click to a third-party site. The engine synthesizes the answer, and the user does not scroll down to the organic links.
Queries to prioritize for protection
Not all queries are threatened in the same way. Transactional searches (purchase, quote, booking) maintain a more stable CTR because the user needs to act on a site. Short informational queries (“what is SEO”, “definition of natural referencing”) are the first to be cannibalized by AI summaries.
We observe that the content that resists best shares three characteristics:
- It contains proprietary data or numerical comparisons that AI cannot rephrase without losing their value.
- It offers interactive tools (calculators, simulators, configurators) that require a visit to the site.
- It targets long-tail queries with mixed intent, where the answer cannot be summarized in three sentences.
GEO and SEO: two complementary analysis frameworks for search engines
GEO (Generative Engine Optimization) refers to optimizing visibility in responses generated by AI engines (Gemini, ChatGPT, Perplexity). This field does not replace SEO: it adds to it.
The difference lies in the selection mechanism. In classic SEO, Google ranks pages based on relevance and authority signals. In GEO, the language model selects content excerpts to integrate into a synthetic response. The selection criteria differ: the semantic structuring of content, the presence of verifiable factual data, and the citation of primary sources weigh more heavily than the link profile.
Content structuring for generative SEO
Content optimized for GEO relies on self-contained response blocks. Each section should be extractable and understandable in isolation by a language model. Concretely, this involves:
- Subheadings that formulate a specific question or statement, not a vague thematic label.
- A first paragraph under each H2/H3 that directly answers the subheading before elaborating.
- Fact-based data sourced in the body of the text, not just in footnotes.
- A Schema.org markup (FAQ, HowTo, Article) that facilitates extraction by generative engines.
SEO ranking tracking: measuring the real impact on organic traffic
Analyzing your website without measuring results over time is like driving blind. Ranking for a given query is no longer sufficient as a single indicator. CTR by position and by type of SERP becomes the priority metric.
A page in third position for a query without AI Overview generates very different traffic than the same position for a query where Google displays an AI summary. We recommend segmenting tracking into three categories: queries with AI Overview, queries with featured snippets, classic queries. This segmentation allows for directing content efforts toward queries where organic clicks remain accessible.
The other point of vigilance concerns internal cannibalization. Several pages on the same site targeting the same search intent neutralize each other in the results. A regular audit of positions cross-referenced with Search Console data (impressions, clicks, pages) identifies these conflicts and allows for consolidating content on a single URL.
Natural referencing in 2026 is no longer a channel that you “set up” and then leave running. The arrival of AI responses in Google search results requires continuous, technical, and editorial analysis to preserve every point of organic traffic.