A company's website may reveal where its technology is heading—but only if someone can separate real signals from polished marketing. This study turns those public pages into a moving map of industrial change. Monitoring technological trends matters, but traditional sources such as surveys, interviews, and patents are costly, resource intensive, and unable to capture technological change in real time.

Those limits create a need for flexible, scalable approaches that use changing sources of information to follow technology and emerging trends more effectively. The proposed approach uses corporate web content to identify and track technological trends, rather than relying only on traditional sources or fixed categories.

It first identifies firms whose websites provide evidence of artificial intelligence adoption, then pulls out words and short phrases and sorts them into technological or non-technological terms. Finally, equivalent terms are grouped, and the results are organized for each time point so changes can be compared.

The collected website material was cleaned and combined into one body of text for each company, then narrowed to passages most closely related to artificial intelligence. This narrowing acts like sorting a huge pile of mail before reading it: it reduces the amount of text needing deeper review while preserving different kinds of context.

A deeper reading then determines whether those selected passages provide actual evidence that a company uses artificial intelligence. Using this approach, three hundred fifty-six companies were classified as artificial intelligence adopters in 2023. By 2025, the number had risen to four hundred sixty-seven companies, indicating increased artificial-intelligence engagement across the sector.

The model achieved very high overall accuracy, but it was less complete when judging individual words than when judging clearer multi-word technology expressions. That tradeoff came from a cautious strategy: it prioritized avoiding false alarms, which preserved high overall accuracy but missed some less obvious technology references.

The comparison tracks how often the most referenced technologies appeared on company websites between 2023 and 2025. Most terms increased, with several more than doubling, while only a small number declined modestly. The spread is striking: predictive maintenance reaches one hundred percent, while mobile devices and neural networks decline, leaving a wide gap between the fastest risers and laggards.

The study also examined which technologies appeared together, revealing that firms often present them as complementary bundles rather than isolated tools. Automation frequently appeared alongside computer vision, robotics, real-time systems, machine learning, and Industry 4.0, framing automation as part of a broader digital production system.

Connected devices and data streams also appeared with real-time systems and machine learning, pointing to a close relationship between connectivity, automation, and intelligence in smart production. There is an important catch: company websites are a form of self-presentation, so firms may exaggerate fashionable technologies or hide capabilities for secrecy or compliance reasons.

The comparison between 2023 and 2025 may also be affected by the rapid spread of generative artificial intelligence and the growing use of artificial-intelligence-assisted writing on websites. The study focuses on Spanish manufacturing firms, which limits how broadly the findings can be applied, although the framework can be used in other settings and languages.

Corporate websites can provide a fast, detailed view of how manufacturing firms talk about artificial intelligence and the technologies around it. But the map reflects both real activity and what companies choose to say publicly.