Monthly traffic to generative artificial intelligence services increased by 251% between February 2024 and March 2025, according to Cloudflare’s network data. This acceleration affects information retrieval, document management, visual content creation, and everyday programming. At the same time, the European regulatory framework is tightening with the gradual implementation of the AI Act, which imposes new obligations on providers and users of AI systems.
Autonomous AI Agents and the End of Traditional Web Search
The conversational assistants of 2023 were limited to answering isolated questions. In 2024, a new generation of tools takes a leap forward: AI agents capable of chaining multiple tasks without supervision. Booking a ticket, analyzing a spreadsheet, operating a SaaS tool, drafting and then sending a report, all in a single sequence of instructions.
Further reading : The Latest Trends and Innovations in the Automotive World in 2024
This change alters the way users interact with the web. Google rolled out its AI Overviews in the United States in May 2024. The measured observation is clear: the click-through rate drops by 61% when an AI summary is displayed at the top of search results.
For website publishers, the consequence is a drastic decline in incoming traffic, raising questions about the viability of business models based on organic search.
See also : Everything You Need to Know About the Definition and Different Uses of Email in Daily Life
Field reports vary on this point: some sectors (recipe websites, simple definitions) experience a collapse in traffic, while complex or transactional queries seem less affected. The available data does not yet allow for a conclusion on the definitive extent of this shift, which remains in gradual deployment, including in Europe. Several analyses published on the jdmag.net tech site document these developments over the months.
European AI Act: What the Regulation Changes Practically
The AI Act, adopted by the European Union, is the first comprehensive legal framework dedicated to artificial intelligence. Its implementation occurs in stages, with obligations varying according to the risk level of the concerned system.
High-risk systems (credit scoring, CV screening, biometric surveillance in public spaces) must meet requirements for transparency, traceability of training data, and human oversight. For companies developing or deploying these tools, this entails technical audits, enhanced documentation, and, in some cases, compliance assessments by third-party organizations.
Obligations for Generative AI Providers
Foundation models (like GPT, Claude, Gemini) are subject to specific transparency obligations. Providers must indicate when content has been generated by AI and document the datasets used for training. Synthetic content (images, videos, audio) will need to carry detectable labeling.
In contrast, uses considered to have limited risk (customer service chatbots, anti-spam filters) are subject to lighter obligations, primarily focused on user information. Full compliance will unfold over several years, with different deadlines depending on system categories.
Massive Adoption of Generative AI: Beyond the Media Phenomenon
The 251% increase in traffic to generative AI platforms in one year, measured by Cloudflare, reflects a large-scale change in habits. ChatGPT, Claude, Midjourney, and their competitors are used daily for:
- Writing professional documents, administrative letters, or meeting summaries, with significant time savings on repetitive tasks
- Generating images, mockups, or visual materials without prior graphic skills, redistributing certain tasks within teams
- Coding prototypes, correcting scripts, or automating data processing, including by non-developer profiles
- Searching for information conversationally, partially replacing traditional search engines
This adoption raises concrete questions about the reliability of generated responses. Hallucinations (false information presented confidently) remain frequent, and non-expert users do not always have the means to verify outputs. In the health or legal sectors, the consequences of erroneous information can be severe.
Energy and Digital Infrastructure: The Hidden Cost of AI
The explosion in the use of generative AI models leads to a considerable increase in energy consumption in data centers. A query sent to a language model consumes significantly more resources than a traditional web search. This reality directly conflicts with the climate commitments of major tech companies.
The demand for computing capacity exceeds the available renewable energy supply in several regions where data centers are concentrated. Projects for modular nuclear power plants dedicated to powering data centers have been announced in the United States, indicating that the sector anticipates a sustainable structural need.
Challenges for French Companies
For companies integrating AI into their processes, the cost of accessing APIs from major models remains volatile. Prices have decreased throughout 2024, but reliance on a few American providers (OpenAI, Anthropic, Google) raises questions of digital sovereignty and service continuity. European initiatives exist, but their maturity lags behind dominant solutions.
The year 2024 marks a measurable turning point in the integration of AI into daily life, but many uncertainties remain: regulatory framework still being deployed, unreliability of outputs, increasing energy footprint. The coming months will determine whether this adoption stabilizes around productive uses or if technical and legal limits hinder the current momentum.