Make your paper stand out from genAI: A section-by-section guide
In case no one’s told you yet, the hallmarks of a largely AI-generated research article are obvious, and they’re quickly becoming clear signals to desk-reject a paper. So say my clients and other contacts at business schools, the vast majority of whom are either reviewers or editors at top journals.
The bad news for everyone submitting to these journals is that AI has led to a vast increase in submissions. This means already swamped managing editors and reviewers are turning to your paper with even fewer resources—they’re swamped, exhausted, and sick of sifting through slop.
The good news? Non-slop stands out more than ever.
In other words, compelling, well-written research papers read like a breath of fresh air right now. This is something that I’ve seen borne out repeatedly over the past few years, as our percentage of edited papers that go on to be accepted has climbed steadily—recently reaching close to 85%.
While it obviously helps that we’re typically working with brilliant thinkers who ask great questions and produce robust findings, it’s my belief that our editorial focus on highly readable, story-driven papers and carefully polished details is a big driver of these results.
And for better or worse, another thing we now do (and encourage authors to do) is consciously avoid certain signals that AI wrote the paper. Which signals do I mean? Well, I spoke with a few of our clients who happen to be associate or senior editors at FT50 journals about the particularly egregious patterns they're seeing. I then turned our conversations into the points below, so that you can ensure your next submission avoids them.
Oh and it probably goes without saying, but don’t bother attempting to feed this info into an AI tool with a feeble prompt to avoid these things. I’m sharing this insider info openly because I know that simply won’t work—especially since most of these suggestions are creative tasks, so they’re inherently impossible for a statistical predictor to replicate.
Disclaimer:
To be clear, I’m not saying this is a comprehensive list of AI tells (which change constantly ) or that genAI use has no place in academic writing and publishing. But if you want to instantly boost your chances of acceptance at increasingly competitive journals, make sure your articles doesn't hit these AI-associated notes. Journal teams are now far more likely to engage with a paper (even a flawed one) if they believe they can trust its author.
Submissions that stand out from the slop: A section-by-section breakdown
Title & Abstract
Don’t
-Try to come up with a title that seems reductively catchy or “smart” (i.e., jargon-filled)
-Seek to emphasize the importance of your work
-(Over)use the current well-known AI tells (e.g., “delve”, “tapestry”, “quietly”, colons, etc.)
Instead
-Think about the questions that led you to conduct this research; consider using one in your title or abstract to stoke the same curiosity in readers
-Answer this question: What is the real and present danger if no one ever reads this research? Make sure that answer (i.e., the value of this work for readers) appears somewhere in your abstract in accessible language
-Intentionally avoid AI favorites, at least in your abstract. I know… it’s unfair and annoying. I wish I could still just say “Who cares if AI likes the em dash—so do I and I’m not giving it up!” But the fact is, I've actually had reviewers tell clients that some of these words and punctuation must be removed before the journal will publish their paper. So it’s worth staying abreast of which words to avoid, at least in this vital first impression of your submission.
Introduction
Don’t
-Write in jargon and technical language alone
-Write in similar-length, monotonous sentences and paragraphs
-Give a broad, abstract overview of the paper’s context and contributions
Instead
-Vary between the “Academese” that’s inevitably required to discuss your research and “normal”/non-academic language. The more important a sentence is, the plainer it should be. So push yourself to say important things simply.
-Write the way you’d speak, in real conversation to colleagues about this topic. Talk out loud as you write and work out what you want to say. It’s unusual, in human writing, to have your thoughts develop in sentences that just so happen to be almost exactly the same length and cadence. Let your points reflect real thinking and real speech so they’re varied—or "bursty."
-Dig into concrete details, even in the Introduction. What did you ask to kick off this research, and why did you want to know the answer? What did you do to get your answers? Where and how did you do it? What did you find and how did it surprise you? The answers to these questions reflect a human brain and body doing human activities; they can’t be captured by consolidating or describing other people’s work. And that’s why they stand out as compelling and trustworthy.
Theory/Literature Review section
Don’t
-Focus on “synthesizing” anything here; this is now considered a clear AI tell among many reviewers (if you must synthesize, make sure it’s clear why anyone would do so—i.e., what problem does it fix?)
-Fail to check that every citation is (a) real and (b) included in your reference list
-Shoehorn in references to the journal editors’ work
Instead
-Argue. Use past studies as support and/or counterfoil for your own original argument. Lists of past publications and findings are completely pointless without a unifying reason for sharing them.
-Manage all citations and references using a reference manager like Zotero. Then cross-reference citations and references with a tool like ReciteWorks, if you want to be extra sure they’re all there and formatted correctly. Citations are a big trust signal now that genAI’s gained a reputation for hallucinating them, so they’re worth polishing carefully.
-If your research does in fact engage with work by one of the journal’s editors, great. But if it doesn’t, these nods don’t just look like attempts at flattery anymore—they look like evidence of heavy AI use, as many AI tools "know" to advise authors that this tactic can increase their chances of acceptance.
Methods
Don’t
-Fail to engage with and detail an actual methodology, so this section feels like a “missing middle” (i.e., one that looks structurally complete but says nothing)
Instead
-Detail your methodological approach and why you chose it (i.e., why it was the best approach, setting, etc. over other available options to address your specific research question). Bonus points if you have empirical findings and preregistered studies since those are clearly harder (or at least more nerve-wracking) to fake with AI.
Findings/Results
Don’t
-Shy away from sharing idiosyncratic, checkable details (if they matter, that is)
-Use this section to simply report what you found, laundry list style
Instead
-Include relevant details that only someone who actually conducted this study would know (e.g., a footnote or parenthetical comment about something that happened during fieldwork, a dispute between coauthors about how to interpret a finding and how it was resolved, etc.).
-Present your findings strategically, to illustrate your points and support the contributions you’ll expand on in the Discussion. Your findings won’t all have equal importance, and that should be reflected in the way you write about them (e.g., some warrant only a brief mention in a footnote while others get grouped and compared in detail).
Discussion & Conclusion
Don’t
-Use this section to simply restate/summarize what the paper’s already stated
-Offer a list of propositions
-Under-edit your Discussion so it reads as “verbose, punctuation-light writing” (or, conversely, overuses the same punctuation or structures)
Instead
-Interpret the findings section—what do your findings mean for research and practice?
-Challenge yourself to connect your thoughts and takeaways in a flowing, easy-to-follow chain of reasoning (AI sucks at this, so it usually offers a skimmable list of claims as a substitute).
-Extend the implications of what you found to offer other promising directions and be honest about the limitations of your work.
-Proofread your Introduction and Discussion with particular care since they’re the points where readers enter and exit your paper. Boredom (due to repetitive language and structures), frustration (due to confusion around what you’re saying and why it matters), and other negative reader experiences hit extra hard here, affecting the perceived value and rigor of your work. Edit by reading out loud (again) to better gauge how these sections, in particular, will “sound” to a reviewer.
Final tip: If your paper does make it to R&R, DON’T rush to resubmit within a matter of days. In the past, this might have signaled a motivated, amenable attitude and gratitude for the opportunity. Today, it signals that you outsourced the revisions to AI, rather than engaging with reviewer feedback in any meaningful way.
After 16 years of editing submissions to top business journals, I'll admit AI is drastically changing many things—but not the formula for success in this challenging field (at least not yet). Problem-led research, honestly conducted and shared with a strong narrative, is still ultimately coming out on top, as far as I can see.
Want to know more about stacking up career-changing publications? The CS Phares Article Accelerator is a 12-week program that takes you through the process of writing and submitting a top-tier article, giving you a replicable publishing system you can use for the rest of your career. Join the January 2027 waitlist to secure early access when spots open.
Comments