Journals Are Using AI Submission Screening: What Authors Should Expect in 2026
AI screening is no longer hypothetical. Major publishers are using it before peer review, which changes what authors need to catch before they submit.
Readiness scan
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan before you submit. Catch the issues editors reject on first read.
Quick answer: Yes. Springer Nature tested an AI quality-check tool on more than 100 open-access journals and 100,000 submissions, with 14 pre-review suitability checks. JMIR Publications uses automated manuscript evaluations across its portfolio, and Wiley reports more than 1,000 journals using Research Exchange screening. These systems flag issues for human review; the current publisher documentation does not describe a robot making acceptance decisions alone.
Use the free readiness scan for an author-side check across overlapping risk categories, not as a replica of any publisher's proprietary screen.
Who This Is For
This guide is for authors deciding what to clean up before they upload to a journal that runs automated screening. It is not for authors trying to beat an AI detector, and it is not a language-editing resource. For line-level English, use a copyediting service. For the privacy and reliability of author-facing AI review tools, use the safe AI manuscript review guide; this page owns publisher-side screening and author preparation.
Evidence basis and source limitations
We reviewed current public documentation from Springer Nature, JMIR Publications, and Wiley on July 10, 2026. In our analysis of those publisher materials, "AI submission screening" is an umbrella term, not one standardized test: the documented checks, timing, outputs, and escalation rules differ by platform.
Source limitations: We did not access confidential screening models or submit test manuscripts to these publisher systems. Public publisher pages may describe portfolio-level capabilities that are not active at every journal. This page therefore does not claim detection rates, unpublished thresholds, or journal-by-journal coverage beyond what the named sources state.
Which publishers document AI submission screening?
The strongest public evidence is publisher-specific. The table separates what each organization actually documents from what authors should infer.
Publisher or platform | Publicly documented scale and checks | Human-decision boundary |
|---|---|---|
Springer Nature | Tested on 100+ open-access journals and 100,000+ submissions; 14 suitability checks include data-availability statements, human and animal ethics, clinical trials, and misuse threats | A human expert double-checks results before a final decision |
JMIR Publications with Signals | Used across the JMIR portfolio for automated, transparent research-integrity evaluations and AI-supported investigation through Sleuth AI | The announcement describes support for integrity checks, not autonomous editorial decisions |
Wiley Research Exchange | 1,000+ journals, 10,000+ manuscripts screened monthly, and 30+ screening checks; current guidance lists scope match, reference analysis, paper-mill patterns, problematic phrases, and machine-generated-content detection | Wiley says AI detection alone is not sufficient for a decision and must be considered with other checks and expert assessment |
Springer Nature's January 7, 2025 announcement is unusually concrete. The tool alerts editors to potentially unsuitable manuscripts before peer review, but Springer Nature explicitly says a human expert double-checks the result. The named checks show that screening extends beyond prose similarity into structured ethics, trial, data-availability, and misuse-risk questions.
JMIR Publications announced on November 12, 2025 that it had begun evaluating submissions with Signals Manuscript Checks across its portfolio. Scientific Editorial Director Tiffany Leung described the integration as a research-integrity measure. Signals combines network analysis, expert knowledge, and AI; the public announcement does not enumerate every check or claim that the system makes final decisions.
Wiley's current Research Exchange material is the clearest evidence that screening can combine many different tools. Wiley's editor FAQ states explicitly that its machine-generated-content detector should not be a primary decision-making tool because false positives are possible. Editors are told to consider the result alongside reference analysis, plagiarism detection, tortured-phrase detection, other integrity flags, and expert judgment.
The practical conclusion is narrower than "journals let AI reject papers." Publishers are using automated evidence to prioritize checks and surface concerns earlier, while their own public guidance retains human review.
What these systems are usually screening for
Authors often assume "AI screening" means a single detector looking for ChatGPT-like prose. The current publisher documentation shows a broader stack.
Screening area | Documented examples | What an author can verify |
|---|---|---|
Submission and policy completeness | Data-availability, ethics, clinical-trial, disclosure, and required-statement checks | Required files and statements match the journal's current instructions |
Reference and text integrity | DOI or reference problems, retracted references, similarity, problematic phrases, and paper-mill patterns | References exist, support the attached claims, and are consistently formatted |
Scope and content signals | Journal scope match and machine-generated-content probability | Abstract and cover letter state the fit accurately; AI use is disclosed under the journal's policy |
Escalation support | Combined flags sent to editors or integrity staff for review | Provenance notes and underlying records are available if the journal asks questions |
Not every journal runs every check. Wiley's published Research Exchange list should not be projected onto Springer Nature, JMIR, or an individual journal without evidence.
What authors can control before submission
The biggest change is timing. Structured integrity, policy, reference, and scope signals can now appear before peer review, so an editor may ask for clarification or hold a manuscript before a reviewer sees it. That makes consistency across the manuscript, submission form, disclosures, figures, and references an earlier requirement.
Named failure patterns in our pre-submission review work
In our pre-submission review work, the manuscripts most likely to run into screening friction are rarely the ones with the worst writing. We observe a pattern: the package is internally inconsistent, and a structured check surfaces that inconsistency before a human editor would. Each named failure pattern below maps to a manuscript component an author can inspect. These are Manusights review patterns, not claims about the hidden rules of any publisher model.
- Citation hygiene: references verified against live databases, not written from memory, with each cited paper checked for existence and support for the attached claim.
- Figure provenance: figures with a clear creation and editing history, no duplicated panels, and no last-minute ambiguity about how they were generated.
- AI-use disclosure: any real AI assistance disclosed in the exact way the target journal's policy asks for, not omitted because it felt minor.
- Ethics and completeness: ethics statements, competing-interest declarations, and reporting items present and consistent across the methods and the submission system.
- Claim-to-evidence calibration: the abstract and conclusion earned by the data, not inflated novelty or a vague gap statement that an editor will filter quickly.
- Journal fit made obvious: the framing names why the paper belongs at this journal, so an editorial-triage screen does not read it as a scope mismatch.
We trace each pattern back to a specific part of the manuscript, the abstract, figures, references, methods, or disclosure statements, so "AI screening" becomes earlier manuscript due diligence rather than a detector hunting for one forbidden sentence.
The goal is not to promise that a manuscript will "clear" an unknown publisher system. It is to remove preventable inconsistencies and make the underlying evidence easier to defend when an editor investigates a flag. Manusights does not use submitted manuscripts to train models, and its findings are tied to passages in the uploaded text.
What authors should do differently now
The answer is not "avoid AI entirely." The answer is to assume your paper will face more structured scrutiny before review. The practical move is to treat the pre-submission pass as the moment to catch what a screening tool would, so you are fixing problems on your own schedule rather than reacting to a flag mid-submission.
A practical pre-submission checklist
- verify references against live databases, not memory
- make sure the claims in the abstract and conclusion are earned by the data
- check the target journal's AI policy and disclosure requirements
- verify image provenance and figure-generation history
- clean up generic filler that makes the manuscript sound fluent but empty
- make the journal fit obvious in the framing
For the privacy and model-safety side of that decision, use what safe AI manuscript review requires.
How It Works: An Author-Side Readiness Check
An author-side screen is not copyediting and cannot reproduce a publisher's confidential model. It can inspect overlapping, observable risk categories before upload:
- citation metadata and claim-to-reference consistency
- figure provenance and manuscript-to-figure consistency
- ethics, competing-interest, data-availability, and AI-use statements present in the manuscript
- journal fit, claim calibration, and reviewer-facing weaknesses
The output should tell the author what to verify or revise, with the relevant manuscript passage attached. It cannot guarantee that a publisher tool will not flag the paper, reveal an unpublished screening threshold, or replace an editor's judgment. Use the free readiness scan to identify author-controlled risks, then check every requirement against the target journal's current instructions.
Paid Manusights reviews include a 60-day money-back guarantee, and we do not train models on submitted manuscripts. If the only gap is a missing statement, file, or checkbox already named in the journal instructions, a paid review is not worth it; fix that requirement directly.
What not to do
Do not respond to this trend by trying to outsmart journal screening with more polished AI prose. That is the wrong game.
The stronger response is:
- cleaner evidence
- tighter references
- better disclosure
- more realistic framing
- and a manuscript that still looks defensible when someone reads past the first paragraph
Submit If
- the references, ethics language, and AI disclosures have been checked against the journal's actual policy
- the figures have clear provenance and no last-minute ambiguity about how they were created or edited
- the manuscript package is consistent enough that an early screening tool will not surface preventable compliance noise
Readiness check
Run the scan to see how your manuscript scores on these criteria.
See score, top issues, and what to fix before you submit.
Think Twice If
- you are hoping polished prose will hide weak evidence, vague methods, or incomplete disclosure
- the paper still contains unresolved citation or figure questions that a reviewer could verify in minutes
- you have not checked whether the target publisher now screens for integrity and policy issues before review
Bottom line
AI submission screening is already here at major publishers. The systems are uneven, and they are not replacing editors, but they are moving the first quality filter earlier in the pipeline.
That should change what authors do before submission. The manuscript that survives best is not the one that "beats the detector." It is the one that is cleaner, better sourced, better disclosed, and easier for both machines and humans to trust.
If you want to see where your manuscript is exposed before the journal does, run the AI manuscript integrity check.
Before submitting, a manuscript readiness and journal-fit check can catch the fit, framing, and methodology gaps that editors screen for on first read.
Frequently asked questions
Springer Nature says its editorial quality-check tool was tested on more than 100 open-access journals and 100,000 submissions. JMIR Publications uses Signals Manuscript Checks across its portfolio. Wiley says Research Exchange Screening serves more than 1,000 journals and processes more than 10,000 manuscripts per month.
The checks vary by platform. Documented examples include submission completeness, ethics and integrity statements, clinical-trial and data-availability information, scope match, reference problems, paper-mill patterns, unusual phrases, text similarity, and potential machine-generated content.
The publisher documentation reviewed here describes decision support rather than autonomous rejection. Springer Nature says a human expert double-checks results before a final decision, and Wiley says machine-generated-content detection should not be a primary decision-making tool.
Check the target journal's current submission requirements, verify references and disclosures, confirm figure provenance, and make the manuscript package internally consistent. Do not try to optimize for one detector because publishers use different checks and human judgment remains part of the process.
Sources
- Springer Nature launches an AI-driven editorial quality-check tool
- JMIR Publications adopts Signals Manuscript Checks
- Wiley Research Exchange screening
- Wiley AI in research publishing FAQ
- Publisher claims and platform scale were checked on 2026-07-10. Screening tools, journal coverage, and editorial policies can change; verify the target journal's current instructions before submission.
Final step
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan. See score, top issues, and journal-fit signals before you submit.
Anthropic Privacy Partner. Your manuscript is never used to train any model.