Does the FDA use AI to review drug labels?

The FDA uses AI as a helper. When the FDA launched its agency-wide internal generative AI tool in June 2025, it said the tool can perform faster label comparisons, among other tasks. In May 2026 the FDA said its staff are involved at every stage, with human experts verifying inputs and outputs.

A comparison shows what changed between two versions. It doesn't show whether each word means what it should. A correctly spelled wrong word that sits in both versions won't show up as a difference, so a sponsor's own read for meaning is still the check that keeps one off a published label.

What the FDA has said

From the FDA's own press announcements:

• June 2025: the FDA launched an internal generative AI tool for employees, from scientific reviewers to investigators. The FDA listed uses including summarizing adverse events, faster label comparisons and help with clinical protocol reviews. The FDA said the models don't train on data submitted by regulated industry.
• May 2026: the FDA announced an upgraded version with features including document generation, searchable text from scanned documents, and better search across large document sets, and said it consolidated submission data systems into one platform the tool can work on top of. The FDA said human subject matter experts verify all inputs, analytic processes and output implementation.

We haven't seen the FDA describe how reviewers use the tool on any specific application, and we don't speculate.

What a label comparison can and can't show

A comparison is good at finding what changed. It isn't built to judge whether an unchanged word is the right one. Compare "hypotension" and "hypertension": both are real words, and if the same one is in both versions, nothing is flagged. More in compare-to-master vs read-for-meaning.

Should we AI-check before submitting?

Our view: it can be a useful extra pass, but it isn't a substitute for a person reading the high-stakes lines. See is a tool enough for labeling QC, can an AI proofreader introduce wrong words and using AI as a second reader.

What regulators have said about AI in drug development

The FDA's January 2025 draft guidance on AI to support regulatory decision-making does not address AI used for operational efficiencies, such as drafting or writing a regulatory submission, when that use doesn't affect patient safety, drug quality or the reliability of study results. It isn't a labeling rule.

On product information specifically, EMA's reflection paper on AI in the medicinal product lifecycle, section 2.3.5, says:

AI/ML applications used for drafting, compiling, editing, translating, tailoring, or reviewing medicinal product information documents should be used under close human supervision. Given that generative language models are prone to include plausible but erroneous or incomplete output, quality review mechanisms need to be in place to ensure that all model-generated text is both factually and syntactically correct before submission for regulatory review.

Who's the last check?

Our reading: the firms behind the label, not the reviewer or the tool. See who is responsible for AI-drafted label errors. That page is general information, not legal advice.

Sources

• FDA Launches Agency-Wide AI Tool to Optimize Performance for the American People
• FDA Expands AI Capabilities and Completes Data Platform Consolidation
• FDA: Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products
• EMA: Reflection paper on the use of AI in the medicinal product lifecycle

Last reviewed: October 14, 2026

More on contextual errors in FDA DailyMed labels.

Questions about wording on your own labels? Talk to us.

We scan freshly published FDA DailyMed labels every day and pull out these high-stakes wrong words.

Every Tuesday and Friday, a new case like this lands in your inbox if you subscribe.

Does the FDA use AI to review drug labels?

The FDA says its internal generative AI tool can help staff with tasks including faster label comparisons, and that human experts verify its inputs and outputs.

Will the FDA's AI catch a wrong word in our label?

Don't count on it. A comparison shows differences, and a correctly spelled wrong word can sit in both versions. Our reading: the firms behind the label are responsible for its wording.

Should we AI-check our labels before submitting?

In our view it can be a useful extra pass, but keep a person's read for meaning as the final check.

Does the FDA's AI tool train on our submission data?

The FDA says its models don't train on data submitted by regulated industry.