Every drug, device and therapy that reaches patients starts as a study, and for more than two decades the public record of those studies in the United States has lived in one place: ClinicalTrials.gov. Run by the National Library of Medicine (NLM) at the National Institutes of Health, the registry has quietly undergone one of the largest technology overhauls in federal health data. In May 2026, NLM announced that its multi-year modernization of both the public website and the submission system behind it was complete, and it has now begun switching off the old tools for good.
The result is less a redesign than a new foundation: a rebuilt public site, a modern data API, a new submission portal for sponsors, and a dataset that is increasingly read by software and AI models as well as by people.
A registry at national scale
The scale explains why the rebuild mattered. A query to the registry’s own public statistics endpoint on November 14, 2026 returned 611,771 study records, with data refreshed on November 13. According to a February 2026 notice in the Federal Register, NIH estimates the work of registering and updating studies and submitting results adds up to roughly 1.4 million burden hours a year across sponsors and investigators.
The same notice restates the registry’s legal backbone. ClinicalTrials.gov was created under the FDA Modernization Act of 1997, gained a results databank under the FDA Amendments Act of 2007 (FDAAA), and operates under the regulations at 42 CFR Part 11, which require certain trials of drugs, biologics and devices to register and report summary results whether or not the product is ever approved. Its stated goals are to help patients find studies and to track the progress of clinical research for public benefit.
From classic to modernized: the PRS transition
Most of the modernization happened out of public view, in the Protocol Registration and Results System (PRS), the portal sponsors use to submit records. NLM’s Technical Bulletin lays out the timeline: the Modernized PRS became the primary system for protocol registration in August 2024 and for study documents and results in September 2025. As of April 2026, more than 80% of users were registering new records in the modern system.
From July 2026, new records could no longer be created in the Classic PRS, and in fall 2026 all protocol submissions move to the modern system. The classic results modules and XML upload are targeted for retirement in 2027 as adoption thresholds are met, while NLM says API features will remain in place. An NCBI Insights post summarizes the practical upgrades: a redesigned Record Summary page that shows where a record stands and what comes next, a Status Dashboard that counts validation issues and reviewer comments per module with one-click filtering, a rebuilt Delayed Results module, and XML upload of a single record containing both protocol and results data.
APIs instead of screen scraping
The public site itself is now a single-page web application, a change with a direct consequence for anyone who used to scrape it. In a July 2025 bulletin, NLM explained that automated scraping of the modern site returns mostly JavaScript bootstrap code rather than study data, and pointed users to the open REST API at clinicaltrials.gov/api/v2, which returns full study records in JSON, supports search by condition, and pages through large result sets with tokens. The live version endpoint currently reports API version 2.0.5.
On the submission side, the PRS API lets large organizations push records directly from their own trial management systems. NLM’s September 2026 bulletin reports a monthly average of 393,000 XML uploads between September 2025 and January 2026, and says 219 organizations used the upload API between March and June 2025. The same bulletin asks those organizations to drop thirteen deprecated data elements, some dating to early FDAAA rules, that the system still accepts but no longer processes or displays. It is a classic modernization move: keep backward compatibility while nudging users toward a cleaner schema.
Fuel for AI and research analytics
Clean, machine-readable trial data is also what makes AI tools practical. NIH’s own example is TrialGPT, a large language model framework built by researchers at NLM and the National Cancer Institute to match patient summaries against trials listed on ClinicalTrials.gov. According to the NIH announcement, clinicians who used it spent 40% less time screening while keeping the same accuracy. The peer-reviewed study in Nature Communications reports that its retrieval step recalled over 90% of relevant trials using under 6% of the collection, and its criterion-level matching reached 87.3% accuracy, close to expert performance. NIH describes the tool as a research effort that still relies on human review, and the team received a Director’s Challenge Innovation Award to test it in real clinical settings.
Researchers use the same data for oversight. A 2025 analysis in the Journal of the Academy of Public Health pulled every historical version of study records through the registry’s data to measure how reporting changed after the 2017 Final Rule.
The results reporting gap
That study shows why the modernization is about more than convenience. Among trials likely subject to FDAAA, the share reporting results within 12 months rose from 8.3% for trials completed in 2010 to 2014 to 23.2% for those completed in 2017 to 2021, and reporting within 36 months rose from 40.8% to 49.2%. NIH-funded and industry trials now report at similar rates. Yet most trials in the later window still missed the one-year deadline. The authors argue that enforcement alone will not close the gap and that better tools to cut the administrative burden of submission are needed, which is exactly the problem the Modernized PRS, its dashboards and its upload API are meant to address.
What it means
For research organizations, the message is operational: the classic portal is going away, so submission workflows, staff training and any in-house XML generators need to target the modern system and current schema. Institutions that integrate their trial management platforms through the PRS API can reduce manual entry and keep public status information current.
For data teams and developers, the API is now the supported way in. Building on it rather than on scraped pages makes analytics, dashboards and AI pipelines more stable and easier to audit.
For patients and the public, the payoff is indirect but real. A registry that is easier to submit to and easier to query should, over time, mean more complete records, more timely results, and better tools for finding relevant studies. Decisions about joining a trial remain a conversation between a person and their care team.
Conclusion
ClinicalTrials.gov’s modernization is a reminder that some of the most consequential health innovation is infrastructure. By rebuilding its website, submission system and APIs, NLM has turned a compliance registry into a modern data platform. Whether that translates into faster, fuller results reporting is the next test, and the registry’s own open data will be how everyone measures it.
Sources
- ClinicalTrials.gov Modernization (ClinicalTrials.gov)
- Phasing Out the Classic PRS: Timeline, Features, and User Support (NLM Technical Bulletin, May 2026)
- Protocol Registration and Results System (PRS) Modernization (NCBI Insights, May 2026)
- Using the PRS API for XML Upload: Cleaning Up Your Data (NLM Technical Bulletin, September 2026)
- Does Screen Scraping ClinicalTrials.gov Work? (NLM Technical Bulletin, July 2025)
- ClinicalTrials.gov API v2 study statistics endpoint
- Information Program on Clinical Trials: Maintaining a Registry and Results Databank (Federal Register, February 2026)
- NIH-developed AI algorithm matches potential volunteers to clinical trials (NIH, November 2024)
- Matching patients to clinical trials with large language models (Nature Communications, 2024)
- Compliance with Results Reporting at ClinicalTrials.gov Before and After the 2017 FDAAA Final Rule (Journal of the Academy of Public Health, 2025)