What should a research paper Discussion section contain?
Before choosing a tool, it helps to understand the actual job of the Discussion section. A strong Discussion should clearly state the most important finding without simply repeating the entire Results section, and then explain what that finding may mean in a broader context.
It should also compare the results with previous research, including studies that reached different conclusions, and offer possible explanations for both similarities and differences. This helps position your work within the existing body of knowledge and shows critical engagement with the literature.
Finally, a strong Discussion should describe the study’s limitations, explain the practical or theoretical implications of the findings, and suggest specific directions for future research.
The 5 best free AI tools for writing a research paper Discussion section
1. Consensus: Best for finding peer-reviewed evidence
The most dangerous mistake in a Discussion section is not awkward writing. It is using evidence that does not exist or citing a real paper for a claim it never made.
Consensus is an academic search engine designed to retrieve and summarize findings from peer-reviewed research. Its database contains more than 220 million papers, and its answers are connected to research sources that you can open and inspect.
Where Consensus fits in your workflow
Use Consensus when you need clear answers about how your findings connect with existing research. It helps you see whether previous studies support your results and whether similar patterns appear across different populations. You can also uncover opposing conclusions and identify variables that explain differences between findings.
Rather than searching broadly, it is far more effective to ask a focused research question. A vague query like “Social media and student anxiety” often leads to scattered insights, while a precise question yields more relevant evidence. For example, asking about daily social-media use and anxiety among university students produces clearer, comparable results.
By refining your search in this way, you can better align external research with your own findings. This strengthens your discussion and makes your conclusions more grounded and credible.
How to use Consensus for a Discussion section
Turn your main result into a clear research question, then search for studies that explore the same relationship. As you review the results, refine your search by filtering for publication date, study design, population, or sample size to focus on the most relevant and reliable evidence. Open the most pertinent papers and verify that the full text supports any summaries you encounter.
Carefully examine each study’s sample, methodology, findings, and limitations, and always cite the original research rather than relying on AI-generated summaries to ensure your work remains credible, precise, and academically sound.
Example
Suppose your study found that students who slept fewer than six hours reported lower test performance.
You could search:
Is sleeping fewer than six hours associated with lower academic performance among university students?
You could then separate the evidence into:
Studies supporting your finding
Studies reporting no meaningful relationship
Studies showing that the effect depends on sleep quality, stress or another variable
That structure gives you something useful to discuss. It is stronger than simply writing, “The result agrees with previous studies.”
Free-plan reality
Consensus offers free access, but advanced research and deeper analysis features have usage limits. It is most useful when you reserve those features for your most important research questions rather than using them for every minor claim.
Pros | Cons |
Best for building the evidence foundation for your Discussion section | Not best for writing your final interpretation |
Helps gather and organize relevant research findings | Does not replace critical thinking or analysis |
Supports identifying key themes and patterns in data | May require manual refinement for clarity |
Saves time during the research phase | Limited in generating original insights |
Useful for structuring supporting arguments | Cannot fully capture nuanced academic reasonin |
Our take: Consensus is one of the strongest starting points for a Discussion section because it anchors your argument in real, peer-reviewed evidence. However, it should be treated as a discovery tool, not a shortcut for interpretation. Its value comes from helping you find and verify studies quickly, but the responsibility for connecting those findings to your own results still rests with you.
2. Elicit: Best for comparing findings across papers
Finding ten relevant studies is only the beginning. The harder task is understanding how they differ. One paper may study 60 university students for two weeks, while another may study 2,000 employees for three years.
A third study may use interviews rather than numerical measurements. If you treat those studies as interchangeable, your Discussion section will become vague or misleading.
Elicit helps organize research by searching papers, summarizing them and extracting comparable information into structured tables.
Where Elicit fits in your workflow
Use Elicit when you need to compare key aspects across multiple studies, such as sample sizes, participant characteristics, study designs, measurement methods, and interventions.
It is also helpful for examining main findings, identifying confounding variables, and understanding the limitations reported in different papers.
This approach is especially useful when the existing literature appears inconsistent, as it allows you to systematically evaluate and contrast the evidence.
How to use Elicit for a Discussion section
Begin with a focused research question, select the most relevant papers, and create columns for the details you need. This structured approach helps you organize information clearly and ensures that you are comparing studies on consistent criteria.
Useful extraction columns include population, sample size, study design, and measurement method. Population shows whether studies examined comparable participants, while sample size helps identify whether one result came from much stronger evidence. Study design separates experiments, surveys, interviews, and reviews, and measurement method reveals whether the studies measured the same concept differently.
You should also include columns for the main finding, authors’ explanation, and limitations. The main finding makes agreements and contradictions visible, authors’ explanation shows how previous researchers interpreted their results, and limitations help you avoid comparing findings without proper context.
Example
Imagine that your study found remote work improved employee productivity, but previous research appears mixed.
An Elicit table might reveal that:
Studies reporting improvement focused on individual knowledge work.
Studies reporting no difference examined mixed job roles.
Studies reporting lower productivity focused on new employees or highly collaborative work.
Some studies measured output, while others measured self-reported productivity.
You now have a defensible explanation for the conflicting evidence.
Instead of writing:
Previous studies showed mixed findings.
You could write:
The apparent inconsistency may reflect differences in job type and measurement. Studies focused on independent knowledge work tended to report productivity gains, whereas studies involving collaborative or newly onboarded employees reported smaller or negative effects.
You would still need to verify that interpretation against the original studies, but Elicit can make the pattern easier to see.
Free-plan reality
Elicit’s Basic plan provides broad paper search, summaries and chat with papers when full text is available. More advanced Research Agent, report and systematic-review workflows have limited free usage.
That makes the free plan useful for a normal research paper, although a large systematic review may require a paid plan.
Pros | Cons |
|---|
Turns a collection of papers into a clear comparison matrix | Does not help interpret what the differences mean for your specific study |
Organizes multiple studies in a structured format | Lacks deeper analytical insights |
Makes it easier to identify patterns across papers | Requires additional effort to draw conclusions |
Saves time when summarizing large volumes of research | May oversimplify complex findings |
Helps visualize similarities and differences quickly | Not tailored to your specific research context |
Our take: Elicit is extremely useful when your Discussion depends on comparing multiple studies rather than citing them individually. It helps you move from “mixed findings” to a structured explanation. However, it does not replace interpretation. You still need to decide what those differences mean for your research question.
3. SciSpace: Best for understanding difficult methods and terminology
You cannot compare your study with previous research if you do not understand how that research was conducted.
Methods sections often contain unfamiliar statistical models, technical measurements and field-specific terminology. Skipping those details can lead you to compare studies that are not actually comparable.
SciSpace provides tools for searching research, reading papers and asking questions about uploaded documents.
Where SciSpace fits in your workflow
Use it when you encounter an unfamiliar statistical method, a measurement scale you do not recognize, or a complex experimental design.
It is also helpful when dealing with confusing inclusion or exclusion criteria, a limitation written in highly technical language, or a result that appears to contradict the abstract.
Useful questions to ask
Instead of requesting a general summary, ask targeted questions that help you engage more deeply with the research. For example, you might ask the AI to explain a statistical method in plain language without removing its important assumptions, or to clarify why the researchers chose a particular method instead of a basic linear regression.
You can also probe the study’s limitations and implications by asking what constraints the sampling method introduces, whether the results demonstrate causation, correlation, or neither, and which participant groups are not represented in the sample. These types of questions push you to think critically about the study’s design and findings.
Finally, it is useful to ask what you should be careful not to claim based on the study. Questions like these help you understand both the strengths and boundaries of the research, ensuring that you use it appropriately in your Discussion section.
Example
Suppose a paper uses a multilevel model because participants are grouped within schools.
A simplified explanation may help you understand that students from the same school are not completely independent. That matters when comparing the study with your own research.
Your study may have:
Used participants from only one institution
Ignored group-level differences
Used a simpler model
Included a broader but less controlled sample
Those differences belong in your Discussion because they may explain why your results were different.
Free-plan reality
SciSpace currently provides a Basic plan with a monthly allowance of Agent credits. Different AI activities consume different amounts of credit, so the exact number of tasks you can complete depends on their complexity.
Use those credits for papers or sections that genuinely block your understanding.
Pros | Cons |
|---|
Removes comprehension barriers when reading difficult papers | Not suitable for making final methodological judgments |
Helps simplify complex academic language | May oversimplify nuanced arguments |
Speeds up understanding of dense material | Can miss subtle context or assumptions |
Useful for early-stage literature review | Not reliable for critical evaluation |
Supports non-native English readers | Should not replace expert interpretation |
Our take: SciSpace is most valuable when you are stuck, not when you are writing. It helps you understand methods and terminology that would otherwise slow you down, but it should not be used to interpret results or draw conclusions. Think of it as a support tool for comprehension, not analysis.
4. Paperpal: Best for improving academic clarity and tone
After researching and drafting your Discussion, you may have another problem: the ideas are valid, but the language is repetitive, informal or unclear.
Paperpal is an academic-focused writing assistant. It provides grammar, language, rewriting and research-related features intended for students and researchers.
Its most useful role in this workflow is editing, not inventing your interpretation.
Where Paperpal fits in your workflow
Use Paperpal after you have already written a complete rough draft. It can help identify informal wording, repetitive sentences, unclear transitions, and wordy explanations.
It also helps catch grammar problems, inconsistent terminology, and phrases that do not fit an academic register.
Example
A rough sentence might say:
Our results are different from the other study because their participants were older.
A clearer academic revision could be:
The difference between the findings may partly reflect the older participant population examined in the previous study.
The revision is more cautious and precise. It does not introduce a new fact or pretend that age is definitely the cause.
That distinction matters. Academic editing should improve your claim without making it stronger than the evidence allows.
How to use Paperpal safely
Write your argument before opening the editor. Review one paragraph at a time and accept only changes that preserve your intended meaning.
Reject unnecessarily complicated vocabulary and recheck citations after rewriting. Read the paragraph aloud to ensure it still sounds like you.
Confirm that hedging words such as “may,” “suggests” and “could” were not removed.
Free-plan reality
Paperpal’s free plan currently includes a limited number of monthly language-editing suggestions and a limited number of daily uses for writing features.
This can be enough for targeted editing, but you may exhaust the free allowance if you repeatedly rewrite an entire thesis or dissertation.
Pros | Cons |
|---|
Helps refine and polish an already developed Discussion section | Not ideal for generating arguments from scratch |
Improves clarity and coherence of existing ideas | Requires prior reasoning and structured content |
Enhances academic tone and readability | Limited usefulness for early-stage writing |
Assists in tightening arguments and removing redundancy | May not provide deep conceptual insights |
Useful for final revisions before submission | Depends on the quality of the initial draft |
Our take: Paperpal is best used at the final stage, when your argument is already complete. It can significantly improve clarity and tone, but it cannot fix weak reasoning. If your interpretation is unclear, editing alone will not solve the problem.
5. Perplexity: Best for recent context and practical implications
Peer-reviewed literature should form the core of most academic Discussions, providing a solid and credible foundation for analysis. However, some topics also require recent information that may not yet appear in journal articles. This is especially true when research needs to reflect rapidly evolving developments.
Such situations are common in fast-moving fields like artificial intelligence, education policy, cybersecurity, public health guidance, technology adoption, government regulation, and labour-market trends. In these areas, relying solely on traditional academic sources may leave important gaps. Incorporating up-to-date information helps ensure that your discussion remains relevant and comprehensive.
Perplexity searches the web and presents answers with links to sources, allowing you to trace information back to its origin. Its value lies not in assuming every answer is reliable, but in enabling you to verify and evaluate the sources yourself.
Where Perplexity fits in your workflow
Use it when you need to find a recent government report or an updated industry survey. These sources can provide reliable, up-to-date information that supports your research and helps you stay aligned with current developments. It is also useful for locating current institutional guidance or a newly introduced policy.
These types of documents are essential when you need authoritative direction or want to understand recent changes in regulations or standards. Additionally, you can use it to find recent adoption data or a primary-source announcement, offering direct insights and strengthening the credibility of your work.
How to use it responsibly
Ask Perplexity to prioritize primary and authoritative sources.
For example:
Find the most recent government or university reports on AI-literacy requirements in higher education. Exclude blogs and marketing websites. Provide the publication date and original source for every claim.
Then evaluate each result.
Check:
Who published it?
Is it the original source?
When was it published?
How was the data collected?
Does the source actually support the claim?
Is the information appropriate for an academic paper?
Do not cite “Perplexity” as the evidence. Cite the government report, institutional document or original study it helped you locate.
Free-plan reality
Perplexity provides general search access without requiring a paid subscription. Advanced research, model selection and higher-volume features have additional limits or require an upgrade.
For one paper, its free search can be useful for discovering recent sources.
Pros | Cons |
|---|
Helps find current primary sources outside the academic publishing cycle | Not suitable for establishing main scholarly evidence |
Provides access to up-to-date information | May lack peer review and academic validation |
Useful for exploring emerging topics | Sources may be less reliable or verified |
Can offer diverse perspectives and real-time data | Information may be incomplete or inconsistent |
Supports early-stage research and idea generation | Not ideal for supporting final academic arguments |
Our take: Perplexity is useful when your Discussion needs current context that academic papers cannot yet provide. However, it should only support your argument, not replace peer-reviewed evidence. Always trace its answers back to original, credible sources before using them.
Find Your Tool in 10 Seconds
Which AI Research Tool Should You Use?
If you're stuck on... | Use... | Because it... |
|---|
Finding evidence that supports or contradicts your finding | Consensus | Searches with natural-language questions across peer-reviewed studies |
Comparing samples, methods, and results across several papers | Elicit | Extracts structured comparisons when literature looks inconsistent |
A paper whose methodology you can't follow | SciSpace | Turns dense methods sections into plain language |
A finished argument that reads repetitive or unclear | Paperpal | Sharpens tone for a final academic-language pass |
A fast-moving topic peer-reviewed research hasn't caught up to | Perplexity | Pulls recent policy, institutional, and industry context |
You almost certainly need two or three of these, not all five.
From Results to Discussion: A Working Map
A Discussion section becomes much easier once you stop treating it like a blank page. You are not starting from nothing. You already have a research question, a set of results, previous studies, and your own understanding of the subject. The task now is to connect these elements in a clear and meaningful way. The following map can guide you as you draft.
Start here: What did you actually discover?
Begin by writing your main finding in one simple sentence:
In this study, we found that…
Do not worry about formal academic language at this stage. Write as if you are explaining your result to someone familiar with your field but not your specific project. Once you have your sentence, test it by reflecting on a few key questions: Did the result answer your original research question? Was it expected? How strong or uncertain was the effect? Was it statistically significant? Does it matter beyond your dataset? And importantly, what should not be concluded from it?
That last question is essential. A strong Discussion does not only explain what your result shows; it also clarifies what it cannot prove.
Build an evidence board
Rather than searching broadly for papers on your topic, turn your finding into a set of focused questions. For example, if your study found that remote work improved productivity, you might ask whether other studies have found the same effect, whether the effect varies by job type, whether self-reported productivity leads to different conclusions, or whether experience level changes the outcome. You might also look for explanations that researchers have proposed.
Use tools like Consensus or other academic search engines to explore these questions individually. As you gather studies, mentally group them into those that support your finding, those that challenge it, and those that provide additional context. Avoid collecting papers simply because they mention your topic; instead, focus on those that help explain your specific result.
Make the studies comparable
Once you have selected the most relevant studies, organize them in a table so you can compare them directly.
Study | Population | Method | Measurement | Main finding | Main limitation | Connection to your result |
Study 1 | | | | | | |
Study 2 | | | | | | |
Study 3 | | | | | | |
Tools like Elicit can help extract this information, but you should always verify important details in the original papers. The purpose of this table is not just organization; it helps you see that studies which appear to disagree may actually differ in important ways. Differences in population, sample size, observation period, outcome definitions, cultural context, or statistical methods can all influence results. These differences often explain more than the headline findings themselves.
Investigate the disagreements
The most valuable study in your Discussion may be the one that contradicts your findings. When you encounter a difference, avoid simply stating that your result contradicts previous research. Instead, explore why the difference might exist.
Consider whether the participants differed in age, experience, location, health, education, or socioeconomic background. Reflect on whether the outcome was measured differently, such as through self-reports versus direct observation. Think about whether the study design varied, since experiments, surveys, interviews, and longitudinal studies produce different types of evidence. Context also matters; institutional, cultural, technological, or policy differences can shape outcomes. Finally, consider whether one study was more precise due to larger samples, better controls, or stronger statistical methods.
If the methodology is difficult to understand, tools like SciSpace can help clarify it, but always return to the original study before making claims about it.
Turn the evidence into an argument
At this stage, stop collecting information and begin shaping your argument. A Discussion paragraph should not be a list of studies but a structured explanation.
Start by stating your result, then compare it with previous research, noting both similarities and differences. Next, offer an explanation for those similarities or differences, and finally, explain what the combined evidence suggests.
For example:
The present study found that [finding]. A similar result was reported by [study], which found [relevant evidence]. However, [other study] reported [contrasting result]. This difference may be related to [population, measurement, method or context]. Taken together, the evidence suggests [careful interpretation].
This structure is a guide rather than a formula. Use it flexibly to build clear and logical paragraphs.
Add the boundaries of your conclusion
Before finalizing your Discussion, reflect on the limits of your interpretation. Consider whether you are describing a relationship as if it were causal, whether you are generalizing beyond your sample, or whether you are ignoring studies that disagree with your findings. Also check whether you are presenting a possible explanation as a confirmed fact or simply repeating your Results section instead of interpreting it.
A strong Discussion is confident about what the evidence supports while remaining careful about what lies beyond it. You do not need to exaggerate the importance of your findings. Instead, focus on showing what happened, how it compares with previous research, why similarities or differences may exist, what your result contributes, and where uncertainty remains. That is the real work of the Discussion section.
Conclusion
A strong Discussion section isn't about finding more tools, it's about following a clear sequence. Write your finding before you touch anything else, turn the real gaps into sharp questions, organize what you find instead of drowning in it, understand why studies disagree instead of just noting that they do, then write the interpretation in your own words. The tools in this workflow exist to remove friction from steps two through four. Steps one and five are yours alone, and that's exactly where your paper actually gets written.
Frequently asked questions (FAQs)
Do I have to follow these five steps in order?
Yes, mostly. Step one matters most because everything after it depends on you actually knowing what your finding means before you go looking for context. Steps two through four can loop a bit as you refine your evidence, but skipping straight to step five without doing the comparison work first is usually where weak Discussion sections come from.
What if a paper doesn't fit cleanly into any of the five buckets in step four?
That's normal, and it's usually a sign the study measured something adjacent to your variable rather than the same thing. Put it in "different context" or "not comparable" rather than forcing it into "supports" or "contradicts." A forced comparison is worse than an honest exclusion.
Can I use Consensus and Elicit for the same paper?
Yes. Use Consensus to find it, then Elicit if you need to pull its data into a comparison table alongside several other papers. They're not competing tools, they solve different parts of the same problem.
How many papers do I actually need in my comparison table?
There's no fixed number. Stop adding papers once new sources stop changing the pattern you're seeing. Five well-chosen papers that clearly agree or disagree are more useful than fifteen that all say roughly the same thing.
What if my finding doesn't match any previous research at all?
That's still a valid Discussion section, it just needs an explanation instead of a comparison. State plainly that this appears to be a new finding, then reason through what in your methodology, sample, or context might explain why. Don't force a false comparison just to have something to cite.