Third party tools receive limited information – they will not receive any information about users or institutions – queries are agnostic of users' attributes and completely decoupled. Any data used by a third-party tool is broken down, filtered, blended with our prompt/engine design, then combined with the relevant data from our corpus. No data is stored in or used to train third-party tools, as these tools have a zero retention policy.

IEEE may occasionally review the searches and results for the purposes of assessing the quality of responses, but the queries themselves and the associated user data that entered the query will not be used for AI training

IEEE Xplore AI was developed in-house by IEEE. It leverages a variety of external software, including pre-built Large Language Models (LLMs) that are then fine-tuned by IEEE, along with proprietary software and algorithms that have been appended to these components.

Our dedicated data science team trains and fine-tunes our own instance of an LLM, optimizing the engine for IEEE and other scientific content and domain-specific terms, and continually tests and makes adjustments to ensure relevance and quality of results.

It also uses AI Inference through Retrieval Augmented Generation to make the tool authoritative, citing sources and grounding the outputs in up-to-date information.

The IEEE Xplore AI has been designed with extremely strict instructions to limit the search retrieval process to only query the IEEE peer-reviewed full text or the expertly curated set of STEM-related metadata and abstracts from OpenAlex. Following this stringent criteria, the AI engine’s response must match the intent of your query and will abstain from providing a response if the research results do not retrieve any STEM-related abstracts that are relevant to the user’s query. 

If the IEEE Xplore AI is unable to find relevant scientific content from these strictly defined databases, it will inform the user. In addition, the AI Overviews reference and link to the source documents to ensure users can quickly and easily verify and trust the answers provided

Finally, our dedicated data science team continually tests and fine-tunes the IEEE Xplore AI to ensure relevance and quality of results.

Together, these factors greatly reduce the risk of any hallucinations.

IEEE Xplore AI was developed with several safeguards to prevent risk that bias will impact any answers provided:

• IEEE Xplore AI has been designed with extremely strict instructions to limit the search retrieval process to only query the IEEE peer-reviewed full text or the expertly curated set of STEM-related metadata and abstracts. As a result, any biases that may be present in other sources on the open web cannot impact the results provided.  
• In addition, it provides responses that match the purpose and intent of a user query to provide relevant results. As a result, other criteria such as the number of citations an article has received, the popularity of a journal or the reputation of an author do not impact the set of results retrieved. If the user does have a specific need to search based on factors such as article citation rate, article popularity, or works from a specific author, IEEE Xplore has other search tools and filters that the user can utilize that are not AI-powered.  
• IEEE also has several feedback mechanisms in place that enable users to report any issues with responses. 
• IEEE Xplore AI is equipped with guardrails to filter out ‘unsafe’ answers; these are typically responses that exacerbate prejudice, harm, or stereotypes against specific individuals or groups.

IEEE conducted extensive alpha and beta tests of the IEEE Xplore AI Research Suite from April 2025 through May 2026. During this time, over 200 customer accounts and more than 3,000 users tested the features and provided valuable feedback. We also collaborated with our own network of IEEE volunteer engineers and scientists, as well as the library and information professional community to ensure the features and results would be relevant and useful to researchers globally. All of this research and feedback helped inform decisions, helped shape the product and its features and functionality and optimize the accuracy and relevance of the result delivered.

Unlike general-purpose AI chatbots, the IEEE Xplore AI Research Suite was custom-built for researchers and optimized for IEEE content and other scientific publications to deliver the best possible results based on the highest quality data in the field of engineering and computing.

While we can’t speak to detailed specs of the foundational technology used in other AI tools, the research tools provided in the IEEE Xplore AI Research Suite were developed following industry best practices to prioritize data and security, ethical AI principles (fairness, transparency), and utilize high-quality data, continuous monitoring, and integrate human oversight to ensure reliability and relevancy of results.

In addition, it is the only AI solution available in the industry that leverages the full-text of quality, peer-reviewed journals and conferences from IEEE to fine-tune the LLM, providing an AI solution that is optimized for IEEE content and other scientific publications to deliver the best possible results based on the highest quality data in the field of engineering and computing. It is also the only solution that is fully integrated with IEEE Xplore, seamlessly aligning with the researcher’s current workflow and research process.