At Markify, AI is used selectively and purposefully to enhance trademark search, data processing, and classification workflows. Our approach is focused on delivering accuracy, transparency, and control, while ensuring that sensitive client information remains protected.
A key principle of our AI strategy is that we are model-agnostic. We continuously evaluate developments in AI technology and retain the flexibility to adopt, replace, or combine models where they provide measurable improvements for our clients. Our choice of technology is driven by performance, reliability, security, and business value rather than reliance on any specific provider.
Core Trademark Search Technology
Our core Word and Image matching capabilities are proprietary and built in-house.
The Word engine is a rule-based system that has been refined over many years using real trademark opposition data, including over 1.5 million opposition and citation conflict pairs. It is not a machine-learning model and does not train itself.
The Image engine is built on an open, pre-trained vision model that has been further fine-tuned in-house using Markify's own trademark data.
Neither the Word engine nor the Image engine is a large language model (LLM), and no generative AI is to this day involved in the core search and matching process.
Goods & Services Similarity
Markify's Goods & Services similarity capability uses self-hosted, open-source embedding models operated within our own infrastructure.
This functionality does not rely on external AI providers, and data processed through this feature remains within Markify-managed environments.
Generative AI
Markify uses generative AI in a limited number of targeted scenarios.
Today, generative AI supports:
An optional, client-enabled layer within Comprehensive Search.
Internal processing workflows that help structure public trademark-office data, such as extracting and organizing information from registry documents and PDFs across multiple jurisdictions.
In both cases, only public register data or search-term data is processed. Customer documents and confidential customer information are not sent for these purposes.
Data Privacy and Security
Protecting client data is a fundamental design principle.
When generative AI is used, it is applied only in carefully controlled workflows. Customer documents and confidential client information are not used in these processes. Markify's primary generative AI applications relate to public trademark-office data and optional client-enabled search features.
Ongoing Innovation
AI technology continues to evolve rapidly, and Markify actively evaluates new approaches that can improve trademark search, classification, data quality, and workflow automation.
As a model-agnostic organization, we are not tied to a single AI model or provider. We maintain the flexibility to adopt different technologies where appropriate, ensuring that our solutions continue to deliver the highest levels of performance, accuracy, and value for trademark professionals.
Download the full document to learn more.