How do you successfully use AI for visual recognition in intellectual property?

Increasingly powerful, Artificial Intelligence (AI) is attractive for its promises of saving time, boosting efficiency, and improving productivity. Today, AI finds a wide range of applications in the professional world — especially in the field of intellectual property (IP). Whether textual (like ChatGPT) or visual (Google Lens, Yandex, etc.), AI is impressive in its ability to process massive volumes of information in a short time. AI-powered visual recognition comes in many forms, often through accessible tools that provide fast — and often impressive — results. But staying up to date and mastering these tools is a real challenge, as progress is constant. Today, AI has found a legitimate role in IP research. So how can you best use AI visual recognition when researching a design? Does its use in IP research have limitations? Let’s take a closer look.

1. Diversify the sources
for more comprehensive visual analysis in IP

Partial coverage

When researching a design, the goal is to get the most complete overview possible to assess its originality and anticipate any potential infringement risks. AI may seem like a valuable ally for this. Two types of tools are available:

Mainstream tools (like Google Lens) that use data available online.
Specialized IP platforms that integrate AI to interrogate registers of filings or other data (like Fovea IP, Questel, or Corsearch).

However, these tools only cover part of the existing design landscape. Many designs still fall through the cracks — ancient objects, designs preserved in museums, poorly indexed publications… These are all sources that AI cannot see.

Not all designs are registered, digitized, or even available online.

Expand your sources

Online resources don’t reflect the full extent of design creations. That’s why exploring additional sources can enrich your search with dated, reliable, and historical results.

Taking the example of a piece of jewellery, here are a few avenues to explore beyond AI tools:

Sales catalogs and brochures: Maty, Cartier, Ti Sento, Agatha, Gas Bijoux, Swarovski, Histoire d’Or…
Specialist magazines: Dreams, C+ Accessoires, HKTDC Jewellery, l’Officiel Horlogerie et Bijouterie…
Museum collections: Musée des Arts Décoratifs, Victoria & Albert Museum…

Also consider using documentary databases, local and national press archives, specialist and economic press, auction catalogs…

This avenues of research can help uncover forgotten or unregistered designs that aren’t available on the web. Moreover, in the context of prior art searches, relying on a dated and indisputable source — like a respected magazine — adds legal value to your documentation.

2. Go deeper in AI analysis with human expertise

Broadly recognition

Designs are often defined by subtle details that make them unique: a particular seam, reinforcement, shoe sole shape, material combination… However, today’s AI visual recognition mainly focuses on general shapes. It struggles to detect the nuances that make a design stand out.

In addition, some online images and older deposits do not show the design clearly enough for the AI to identify it correctly (in profile, with an artistic blur, a poor quality photo, a partial zoom…). This makes detection even more uncertain.

An expert and human perspective

Despite constant improvements, human expertise remains essential for spotting the finer elements of a design. That said, there are ways to help AI along while minimizing errors or oversights: use multiple angles, include similar designs, zoom in on a detail…

Nevertheless, if AI requires too much human input to check the accuracy of its results or feed it relevant data, it can become counterproductive and end up wasting time.

In such cases, it’s often more efficient to rely on a documentalist specialized in design research. They are more experienced and can cross-check sources, detect what algorithms miss, and skillfully guide the use of AI tools.

3. Protect the confidentiality of new designs

Be cautious with unpublished designs

If your design is still confidential and has not yet been revealed, it is not advisable to submit a visual to an online recognition tool. According to the service’s terms of use, your image might be stored, reused to generate other content, or remain online before your official launch. Before using any tool, it is crucial to review its privacy policy carefully.

Many paid tools are aware of these concerns and use secure, reputable hosting providers — be sure to check.

A hybrid method for reliable results

Artificial intelligence-based visual recognition is a valuable approach that facilitates the search for designs by providing rapid results. However, for a rigorous research, AI alone is not enough. It should be used as a tool — guided by an thorough human eye — with a clear understanding of its approximations and potential limitations.

At Paperz IP, we combine technology and human expertise to offer the highest levels of quality and reliability.

Our documentary collection is itself integrated with a text recognition technology called ICR (Intelligent Character Recognition) in our digitised publications. This tool makes it possible to recognise a keyword or an exact phrase in our catalogues, magazines, old and new books… in just a few clicks.

By continuously monitoring AI developments, our team adapts its methods to stay ahead and provide the best results in a secure, well-documented process.

At Paperz IP, AI remains a tool — always guided by our documentalists.

Have questions about using AI in a design search?