Introduction
In August 2026, the Indian Patent Office released Guidelines on the Use of Artificial Intelligence in Patent Examination. The Indian Patent Office AI Guidelines adopt a clear position: Examiners and/or Controller may us AI as an assistive tool for various tasks, however, AI is not a substitute for the Examiner or Controller. Any AI-assisted output that influences examination must be reviewed, validated and adopted only after the concerned officer is personally satisfied with its correctness and relevance. This “human in the loop” approach is sensible, but it also raises a larger question—if AI in Patent Examination can influence classification, search strategy, claim analysis and preliminary reasoning, are the existing safeguards sufficient to protect applicants from an AI-influenced examination path?
Summary of the Guidelines
The AI Patent Examination Guidelines permit the use of AI as an assistive tool in patent examination, while making clear that AI cannot replace the Examiner or Controller. AI may support tasks such as identifying IPC and CPC classifications, generating search terms, extracting claim features, assisting with translation, improving the language and structure of office communications, locating possible legal or technical references and helping officers understand complex technical concepts.
At the same time, the Patent Office AI Guidelines place strict limits on such use. Any AI-generated output must be independently reviewed, verified against authentic sources and adopted only after the concerned officer is personally satisfied with its correctness and relevance. They specifically warn against errors such as irrelevant classifications, overly broad search terms, incomplete claim-feature mapping, inaccurate translations, hallucinated citations and technically unsound reasoning.
The Guidelines also include prohibited uses and administrative measures. Public AI tools must not be used for unpublished patent applications, confidential office records or internal deliberative material. AI also cannot be used as a substitute for the Examiner’s or Controller’s application of mind on substantive issues, and official communications cannot be issued solely on AI-generated output. The competent authority may require material AI use to be recorded, including the tool, nature and date of use, and may establish an AI governance committee, training, audits, impact assessments, feedback mechanisms and incident reporting. These provisions are an important part of the AI Use in Patent Examination framework.
Annexure-II contains a checklist and proposed declaration module. The checklist asks officers to confirm, among other things, whether the proposed use is permitted, whether no confidential information is being entered into unauthorised systems, whether AI output is manually reviewed, whether citations and legal statements are independently verified, whether critical documents and reasoning have been checked, and whether AI use has been recorded where required by office instructions. The checklist states that if any answer is “No”, AI output should not be relied upon without further verification or should not be used at all.
The declaration is framed as a “proposed module” requiring officers to state whether AI was used, identify the purpose of use, and declare that the use was permitted, confidential material was not entered into public or unauthorised tools, AI-assisted conclusions were independently reviewed, and the final reasoning is based on the officer’s own application of mind. On the wording of the Guidelines, this does not appear to be an automatic mandatory disclosure in every examination file. Rather, recording appears to be required where office instructions prescribe it, since the Guidelines say the competent authority “may” require material AI use to be recorded. Even so, the checklist and declaration are important because they convert broad principles of accountability into a practical compliance tool.
Critical Analysis of the Guidelines: The Good and the Bad
The strongest aspect of the Guidelines is their insistence on a “human in the loop” model. This is important because patent examination is not a mechanical exercise. Questions of novelty, inventive step, sufficiency, clarity and unity of invention require technical interpretation and legal judgment. By stating that AI is only assistive, the Guidelines preserve the principle that decisions affecting patent rights must be made by accountable human officers. This is particularly important when considering AI in Patent Examination India.
The Guidelines are also pragmatic. Patent office’s face growing volumes of filings, increasingly complex technologies, and vast prior-art datasets. AI can improve search efficiency, suggest technical synonyms, identify classification possibilities, and help organise claim features. Used properly, this can make examinations faster, more consistent and more comprehensive. For applicants, this may mean clearer office actions, earlier identification of relevant prior art and fewer avoidable delays.