PRISMA 2026: What the Updated Systematic Review Standard Means for Medical Authors

PRISMA has been the dominant reporting standard for systematic reviews and meta-analyses since the original 2009 statement. The 2020 revision added abstract checklists, automation tools, and certainty-of-evidence items, and it tightened the flow diagram conventions that many journals now enforce at submission. Studies published through 2025 found that compliance with even the 2020 version averaged around 42 percent across published systematic reviews, meaning the majority of authors still submit without fully satisfying the checklist they signed off on.

PRISMA 2026, published simultaneously in the BMJ, PLOS Medicine, the Journal of Clinical Epidemiology, and Systematic Reviews, carries forward the 27-item core structure of the 2020 statement while extending three existing items and adding four new ones. The new items are the ones that matter most for 2026, because they address practices that have become widespread without any agreed reporting standard: using a large language model to screen thousands of abstracts, continuously updating a review as new trials are published, and depositing structured review data so that machines can read it as well as humans.

Why the 2020 Version Needed Updating

The PRISMA 2020 statement was drafted through 2019 and early 2020 and published in the BMJ and PLOS Medicine in March 2021. At that moment, large language models were not a tool medical researchers reached for when screening ten thousand citations. Living systematic reviews existed but occupied a small niche, mostly inside the Cochrane Library. And the idea that a review's search results and study flow could be deposited in a machine-readable structured format was more of an aspiration than a practical requirement.

By 2024 those assumptions had collapsed. A growing fraction of systematic reviews in clinical medicine, public health, and health technology assessment were using AI or machine-learning tools at the title-and-abstract screening stage. Tools like Rayyan, Covidence, Abstrackr, and newer LLM-based services were compressing what used to take weeks of dual-reviewer screening into hours. Yet PRISMA 2020 contained no specific item telling authors what they needed to report about that process. Readers could not tell from a methods section whether the AI tool had been validated on the review's PICO question, what the recall rate was at full-text retrieval, or how human checking was integrated.

The living-review gap was equally visible. BMJ Evidence Synthesis launched a living review stream. The Cochrane Library expanded living reviews across multiple clinical areas. JAMA and other flagship journals began accepting them on a case-by-case basis. Authors running these reviews adapted PRISMA 2020 by hand, because there was no standardized way to report update frequency, the trigger for re-running the search, or what happened when new evidence changed a pooled estimate. Peer reviewers had no shared vocabulary for evaluating whether a living review was being maintained responsibly.

Machine readability was a third gap. PRISMA 2020 specified what to include in the flow diagram, but not how to deposit the underlying data. A reader examining a published systematic review could count the included studies from a figure but could not automatically verify the diagram's numbers against a structured record, or aggregate flow data across reviews without reading each one individually.

What Stayed the Same and What Changed

The PRISMA 2026 working group was deliberate about not breaking backwards compatibility. The 27-item core checklist from PRISMA 2020 remains the foundation. Three items have been extended with additional reporting expectations, but any review that already satisfied those items under the 2020 standard satisfies the unchanged portion of the 2026 version as well. The four new items are clearly flagged in the published checklist, so journals and authors can identify exactly what is new without re-reading the entire statement.

This layered structure was a deliberate choice. The working group recognized that hundreds of journals already incorporate PRISMA 2020 into their submission requirements, and that requiring a wholesale switch would create confusion without proportionate benefit. The extended items are backward compatible. The new items are additions. In practice, a checklist produced under PRISMA 2020 rules needs to be supplemented with answers to four new questions, not discarded and rebuilt.

The AI Screening Item: What It Requires and Why It Matters

Item 8b is the one that will affect the largest number of authors immediately. It requires systematic reviewers who used AI or machine-learning tools in study identification, title-and-abstract screening, or data extraction to report specific details about that use. The item was designed to close a reproducibility gap that editors and methodologists had been flagging since at least 2023: a reader examining a published review could not determine whether an AI tool had been used, which tool it was, what its operating parameters were, or how human verification was integrated into the process.

What 8b asks for is not a simple disclosure sentence. The working group guidance specifies that authors should name the tool and version, describe the role it played in the screening workflow, report the proportion of records that received human verification, and indicate whether the tool was validated before use on the review question. That last point is important and often skipped. Many AI screening tools are validated on general biomedical literature but have not been tested on the specific PICO question driving a given review. A tool that performs well on cardiovascular trials may screen differently on pediatric surgery studies. Authors who used a general-purpose large language model to filter citations without any domain validation will need to say so clearly.

The practical implication is that the methods section of any systematic review using AI screening now needs a dedicated sub-paragraph. Something like: "Title and abstract screening was performed using [tool name, version], a machine-learning classifier trained on [corpus]. The tool was set to include any record where the predicted probability of eligibility exceeded 0.3. A random sample of 10 percent of records classified as ineligible was re-screened independently by two reviewers (recall rate 97.8%). All full-text records were screened by two human reviewers." That kind of specificity was optional under PRISMA 2020. It is required under 8b.

Authors who did not use AI screening tools do not need to state that explicitly, but many journal submission systems are likely to add a checkbox confirming compliance with item 8b. The honest answer for most teams right now is that they used some automation at some stage of the search or screening process without calling it AI, and those teams will need to evaluate whether what they did falls within scope of the new item. If a tool was making eligibility decisions or filtering the citation pool, it almost certainly does.

Living Systematic Reviews: A Parallel Checklist Finally Exists

Living systematic reviews are reviews that are continuously updated as new evidence emerges, rather than published once and left static. The format has expanded considerably since it was first formally described around 2014. The Cochrane Library runs living reviews on high-priority clinical questions. The BMJ, JAMA, and the Lancet have each published living reviews on topics where evidence was accumulating quickly, particularly in oncology, infectious disease, and therapeutic areas with multiple active trials.

Under PRISMA 2020, authors running living reviews had to adapt the checklist by hand. There was no agreed way to report how often the search was re-run, what would trigger an update to the pooled estimate, or how numbered versions were being tracked for citation purposes. A reader citing a living review could not always tell which version of the evidence synthesis they had read, or whether a more recent update had changed the conclusions.

PRISMA 2026 adds a parallel living-review checklist that operates alongside the core items rather than replacing them. Authors of living reviews now need to report the planned update interval (for example, quarterly or triggered by accumulation of new trials), the specific conditions under which the search will be re-run and the meta-analysis updated, the approach to incorporating new evidence that changes the pooled estimate direction or size, and the version numbering convention being used for the published review. For journals that accept living reviews, these items give editors and reviewers the vocabulary they need to evaluate whether the review is being maintained to a consistent standard over time.

If you are not writing a living review, these items do not apply to your submission. The checklist is designed to be used selectively, with the living-review additions clearly flagged as applicable only to continuously updated reviews. Authors writing conventional snapshot reviews should not attempt to fill in these fields. The guidance from the working group is that leaving them blank or marking "N/A" is the correct response when the review is not designed to be updated post-publication.

Machine-Readable Flow Diagrams: What the JSON Deposit Means in Practice

The PRISMA flow diagram is one of the most reproduced elements in all of medical literature. Every systematic review that reaches peer review at a major journal includes one, and the numbers on it (records identified, screened, assessed for eligibility, included) are fundamental to understanding whether a review was conducted comprehensively. Despite this, the flow diagram has historically been deposited as a static image, meaning the underlying numbers existed only as embedded text in a figure that machines could not easily read or aggregate.

PRISMA 2026 adds a requirement to deposit a machine-readable structured version of the flow diagram alongside the visual figure. The format endorsed by the working group is a JSON schema developed in coordination with the EQUATOR Network, designed so that the visual diagram can be regenerated from the underlying JSON. The BMJ has been running a pilot of this approach for approximately a year, and its implementation is described in the explanatory document that accompanies the 2026 statement. Several other journals in the Springer Nature and Wiley portfolios have confirmed they will require the deposit as part of their updated submission checklists.

The practical workflow for most authors involves using a tool that generates both outputs simultaneously. The PRISMA 2026 working group has released a web application that lets you enter the flow diagram numbers and automatically produces the JSON file and a formatted visual diagram together. Authors who have been using PRISMA Flow Diagram generators that output only images will need to update their workflow or use the new tool for submissions to journals requiring the structured deposit.

The PRISMA-ScR Update for Scoping Reviews

Alongside the core PRISMA 2026 statement, August 2026 also brought a formal update to the PRISMA Extension for Scoping Reviews, known as PRISMA-ScR. The original PRISMA-ScR was published in the Annals of Internal Medicine in 2018 with 20 essential reporting items and two optional ones. The August 2026 update, published in the Journal of Clinical Epidemiology and indexed in PubMed, was prepared after a scoping review of the literature identified 47 papers proposing a combined total of 37 new reporting items for consideration.

The update was motivated by two developments that had emerged since 2018. First, automation and AI-assisted data extraction had become common in scoping reviews, and the 2018 checklist contained no guidance on how to report that use. Second, the original PRISMA-ScR was developed without patient and public involvement in the guideline development process itself, a gap the update explicitly acknowledges and corrects. The 2026 version incorporated patients and public partners throughout the development phase.

The most commonly needed new items identified in the scoping review of the evidence were related to objectives (flagged by 67 percent of included studies as underspecified), eligibility criteria (63 percent), and search strategy (51 percent). The new PRISMA-ScR items address each of these, adding specificity requirements that were either absent from the 2018 version or too loosely worded to generate consistent compliance. Authors writing scoping reviews should obtain the updated checklist from the EQUATOR Network or the Journal of Clinical Epidemiology publication directly before their next submission, because several journals including those in the Elsevier and BMJ portfolios have already announced they will require the 2026 version at submission.

How Journals Are Implementing PRISMA 2026

The PRISMA 2026 working group has issued model wording that journals can use to update their instructions for authors, and has provided templates for updating submission system checklists. How quickly individual journals adopt the new standard will vary. The BMJ piloted elements of PRISMA 2026 before publication and is implementing the full requirements immediately. Cochrane has been involved in the development process and is updating its systematic review guidelines in parallel. Most Springer Nature and Wiley titles that currently require PRISMA 2020 are expected to publish updated guidance through late 2026 and early 2027.

In the short term, expect a split landscape. Some journals will continue accepting PRISMA 2020 checklists without requesting the new items. Others will require full PRISMA 2026 compliance. A few, like the BMJ, will require the machine-readable flow diagram deposit as a standalone deliverable alongside the PDF submission. The safest approach for authors preparing manuscripts now is to complete the full PRISMA 2026 checklist and have the machine-readable flow diagram ready, so that the submission can satisfy any version the target journal is using.

Authors should also be aware that the EQUATOR Network database, which maintains the canonical versions of all major reporting guidelines, has already been updated to list PRISMA 2026 as the current version. Journal editors who check compliance against the EQUATOR database will see the 2026 standard. If a submission checklist was prepared from the 2020 version, the mismatch may not trigger an automatic rejection, but it will likely prompt a revision request asking for the new items.

The Compliance Problem That Preceded This Update

Before accepting any reassurance that PRISMA guidelines are widely followed, it is worth revisiting the compliance evidence. A meta-epidemiological study published in 2025 and covering published systematic reviews found an average PRISMA 2020 adherence score of roughly 42 percent across included papers. That is after journals had four years to implement the 2020 standard, after the pandemic accelerated the volume of published systematic reviews in clinical medicine, and after multiple editorials in major journals called for stricter enforcement.

The consistently poorest-performing checklist items in compliance studies are the ones requiring authors to explain their reasoning rather than just describe their actions. Risk of bias assessment, certainty of evidence, and sensitivity analyses are routinely underreported even when authors claim to have completed the checklist. The same pattern will almost certainly emerge with item 8b. Reporting that an AI tool was used is straightforward. Documenting validation on the review question, recall rates at human verification, and disagreement resolution procedures takes more effort and is easier to omit without any automated detection catching it.

PRISMA 2026 does not solve the compliance problem. It clarifies what compliance requires. The actual improvement in reporting will come from journals that enforce the checklist at desk review rather than accepting authors' attestations without checking. Evidence from journals with dedicated statistical or methodological reviewers suggests that enforcement at the desk review stage is the only intervention that produces measurable compliance gains. Authors who prepare complete, specific PRISMA 2026 checklists before submission are less likely to encounter revision requests later, and more likely to avoid the resubmission cycle that delays publication of legitimate evidence syntheses.

A Pre-Submission Checklist for Systematic Review Authors in 2026

Before submitting a systematic review or meta-analysis to any journal in 2026, the following questions reflect the minimum required under PRISMA 2026. They do not substitute for reading the full statement and its explanatory document, but they cover the areas where new requirements are most likely to generate revision requests or desk-level queries from editorial staff.

What This Means for Authors Preparing Manuscripts Now

If you completed your systematic review before PRISMA 2026 was published and are now preparing for submission, you do not need to restart the review. The 2026 standard is backward compatible with 2020. What you do need to do is go through the four new items and the three extended items and determine which apply to your review. If you used AI tools in any part of the search or screening process, item 8b requires detailed documentation that you may not have systematically collected during the review. Gather that information from your co-reviewers now, before submission, when the process is still fresh. Trying to reconstruct the parameters of a tool used six months ago at the revision stage is harder.

The machine-readable flow diagram can be completed at any point after the review is finished, because it contains only numbers that should already be documented in your study flow records. The PRISMA 2026 working group tool for generating the JSON deposit is free and takes under five minutes to complete once you have your flow diagram numbers. There is no reason to delay this step.

Authors planning a systematic review that has not yet been started have a cleaner path: register the review at PROSPERO with a protocol that already specifies whether AI screening tools will be used and how they will be validated, document the AI use prospectively as part of the review workflow rather than retrospectively, and prepare for machine-readable deposit as a routine step alongside flow diagram drafting. Reviewers and editors examining systematic reviews in 2027 will increasingly expect to see AI screening documented with the specificity PRISMA 2026 requires, and reviews that omit it will look incomplete in ways they did not before.

There is a broader principle worth noting. Reporting standards like PRISMA exist because systematic reviews are supposed to be reproducible. A reader examining your review should be able to understand not just what you found, but how you searched, how you screened, and what you would have done differently if you had been wrong in a methodological choice. AI tools introduced a new variable that the previous checklist left unreported. PRISMA 2026 closes that gap. Authors who treat item 8b as a nuisance requirement rather than a reproducibility obligation are missing why it was added.

Further Reading