Por
Justin Wong
—
Jenni AI vs. Consensus: Qual verifica e cita melhor um parágrafo de tese?

Consensus shows you what published studies say. A strict supervisor will check the paragraph you write next for claims your sources support and complete references.
TLDR: Jenni or Consensus?
Jenni AI is an AI research and writing assistant: it finds sources and cites them inside your draft. Consensus is an AI academic search engine: it answers questions using published studies.
Choose Jenni if - you want weak sentences flagged and fixed inside your draft, plus complete references.
Choose Consensus if - you want to see what published studies say about a question before you write.
Jenni | Consensus | |
Checking the argument | Comments on the weak sentences, in the draft | Ratings in a list, your sentences left as written |
Citing the claim | Complete APA reference, with volume, issue and pages | APA reference missing volume, issue and pages |
Fixing a clunky sentence | Rewrite in your draft, every change marked | Rewrite in a chat answer, to copy back by hand |
Getting the file out | Word file, citations still linked to the reference list | Copied text or a PDF, papers exportable to Zotero |
Seeing how studies split | Jenni does not do this | Counts papers that say yes or no to a question |
Price | Free plan; Plus $30 a month or $144 a year; Pro $73 a month or $348 a year | Free plan; Pro $20 a month or $144 a year; Deep $65 a month or $540 a year |
More on Consensus - the Jenni review of Consensus is a full post on Consensus alone. The Elicit vs Consensus post compares it with Elicit, another research tool.
All 4 writing rows in the table come from one test thesis paragraph, with 3 problems we added on purpose. One sentence stacked 3 hedge words in a row. Hedge words make a claim sound less certain.
Consensus adds sources but does not change your words, so it put a citation on that sentence and left the hedge words in. Its rating, in a list under the paragraph, said to remove them. But to a reader, a sentence with a citation looks checked.
Jenni put a comment on the sentence itself, saying what to change. The paragraph also had a figure from a clinical trial with no source. Both tools found the trial, and Jenni's reference for it was the complete one.
The rest of this post covers:
How we tested - the paragraph and its 3 problems.
4 writing jobs - how each tool did.
1 research job - one that Consensus does and Jenni does not.
Strengths, prices and advice - where each tool is ahead, what each costs, and which to use.
<CTA title="Get comments on the sentences that need them" description="Peer Review in Jenni scores your draft like a reviewer and puts each comment on the passage it is about" buttonLabel="Try Jenni Free" link="https://app.jenni.ai/register" />
How the test worked
The test used one 201-word thesis paragraph about lipid nanoparticles carrying a gene-editing tool to the liver. We added 3 problems on purpose:
A clunky sentence - 43 words long, opening with 2 grammar mistakes.
3 hedge words in a row - "could perhaps maybe", in the sentence about reaching organs other than the liver.
A claim with no source - an 87% drop in a blood protein, from a 2021 trial in the New England Journal of Medicine (NEJM). The draft never cites the trial.
The topic is biology, but problems like these turn up in drafts in any field.
The limits - this is one paragraph, not a whole thesis. 4 of the 5 jobs are about writing, which suits Jenni. The 5th is about research, which suits Consensus.
Here is the full paragraph, with the 3 problems in bold.
"Lipid nanoparticles have become the most clinically advanced non-viral vehicle for delivering CRISPR-Cas9 to the liver, and the reason is mostly pharmacological rather than genetic. The particle protects the messenger RNA and guide RNA in circulation, it is taken up by hepatocytes through apolipoprotein E mediated endocytosis, and the cargo is transient, so the nuclease is expressed for days rather than for the life of the cell. The first in-human systemic trial of an LNP delivered CRISPR therapy reported a mean reduction in serum transthyretin of 87 percent at the higher dose after a single infusion. There is a lot of things that still needs to be worked out though, for example how the particles behave outside the liver, whether repeat dosing causes an immune response, and what the long term editing outcomes look like at off target sites. It is possible that ionizable lipid chemistry could perhaps maybe be the main lever for extrahepatic targeting, although this has not been conclusively shown. Taken together, the evidence suggests that LNP delivery is a mature platform for hepatic gene editing and an immature one for everything else, which is the gap that most of the current preclinical work is trying to close."
Checking the argument: comments on the sentence or ratings in a list
Winner - Jenni, because its comment sits on the sentence with the 3 hedge words and says what to change. Neither tool rewrote that sentence for you, and Consensus kept its advice in a list under the paragraph.
What Jenni did
We ran 2 of Jenni's checks:
Claim confidence - checks each claim against its sources and suggests edits in the draft.
Peer Review - reads the draft like a reviewer, gives it a score and leaves comments.
Claim confidence found the 87% figure was right, but the sentence never said the result was interim, meaning an early result, not the final one.
The sentence already tied the 87% to the higher dose. The edit added that it was measured at day 28, in an interim analysis of 6 patients. It also attached the trial paper.
The edit also dropped a detail the trial abstract supports: the result came after a single infusion. So read each suggestion before accepting it.
Then Peer Review scored the edited paragraph 7/10 and put each of its 4 comments on the passage it is about.

The hedge words are 1 of 4 highlighted passages, each with its own comment
The comment on the hedge words asked for a narrower claim. It began "“Could perhaps maybe” stacks hedges without clarifying the degree of uncertainty." It also pointed to Zhuo et al. (2021), a paper on another way to reach organs other than the liver.
What Consensus did
Consensus gave back 2 things:
A cited paragraph - its citation search finds studies for or against each sentence. Every sentence got a citation, and the wording stayed the same.
Ratings under the paragraph - it rated 7 of the 8 claims Partially supported, each with a suggested change.
Consensus confirmed the 87% figure and, like Jenni, suggested adding the number of patients and when the drop was measured.

The fix for the hedge words stays in a list under the paragraph
For the hedge words, it suggested "Replace “the main lever” with “one important lever” and remove the repeated hedging." The cited paragraph does not include that change.
It still has the 3 hedge words, now backed by Song 2024, a source Consensus added. Copy that paragraph as it is, and the hedge words go into your thesis with a citation.
<ProTip title="📌 Check each cited sentence against its rating:" description="In Consensus, citations sit inside the paragraph and ratings sit under it. Read the rating before you copy the sentence" />
Citing the claim: a complete reference or one missing 3 details
Winner - Jenni, because its APA reference had the volume, issue and pages. The Consensus reference left out all 3.
What Jenni did
Jenni's citation search runs inside your draft. One catch: when we searched the claim's key terms, the trial came 5th, after a 2022 reprint with the same title. Check that the year says 2021 before you cite.
Citing it took one click. (Gillmore et al., 2021) went in after the sentence, and Jenni started the reference list in APA (7th ed.). Jenni lists 10,000+ citation styles, and this test checked APA only.

The Jenni reference has the volume, issue and pages
The reference's volume, issue and pages, "385(6), 493–502", match the PubMed record. The DOI comes with them.
What Consensus did
Asked about the 87% sentence, Consensus confirmed it and listed the NEJM paper first. The problem was the reference it gave.
Its APA reference for the trial ends like this, exactly as Consensus printed it, double period included:
"(2021). CRISPR-Cas9 In Vivo Gene Editing for Transthyretin Amyloidosis.. The New England journal of medicine. https://doi.org/10.1056/nejmoa2107454"

The Consensus reference has no volume, issue or pages
APA Style asks for the volume, issue and page range in a journal reference. This one has none of the 3, so you would add them by hand before your supervisor sees it.
<ProTip title="📚 Match the reference to PubMed:" description="Compare any reference a tool gives you with the PubMed record of the paper before it goes into your reference list" />
<CTA title="Cite with the volume, issue and pages included" description="One click on Cite adds the in-text citation and an APA reference with volume, issue, pages and DOI under your draft" buttonLabel="Start Writing Free" link="https://app.jenni.ai/register" />
Fixing a clunky sentence: marked edits or a rewrite to copy
Winner - Jenni, because its rewrite appears in your draft with every change marked and a note on what changed. Both tools fixed the grammar.
Both rewrites started with the same 3 words. Jenni's had 28 words and Consensus's had 31, down from 43.
Jenni - its fluency edit, for smoother wording, returned "Several questions remain, including how the particles behave outside the liver, whether repeat dosing triggers an immune response, and what the long-term editing outcomes are at off-target sites."
Consensus - its rewrite began "Several questions remain unresolved, however" and listed the same 3 questions. It sat in a chat answer, with its changes listed, so you copy it back into your draft by hand.
Getting the file out: a Word file or copied text
Winner - Jenni, for a Word file you can keep editing, with each citation still linked to the reference list.
Exporting from Jenni
If your thesis is in Word, Jenni's export keeps each citation in Word's own citation format, linked to the reference list. For Google Docs, choose hyperlinked citations instead. Jenni also exports LaTeX.

Citations stay linked in Word, or hyperlinked in Google Docs
Exporting from Consensus
From Consensus you can copy the cited paragraph as text, or save it as a PDF. There is no Word export.
<ProTip title="📎 Keep the papers Consensus cited:" description="Before you move on, save them as an RIS file, a reference file that Zotero, EndNote and Mendeley can import" />
Seeing how studies split: a job Jenni does not do
Winner - Consensus, because it shows how many papers say yes and how many say no to a question.
The test paragraph treats one point as an open question: does repeat dosing cause an immune response? We asked Consensus that question. It answered "Yes, repeat dosing triggers immune responses that can cause accelerated clearance and systemic inflammation."

The Consensus Meter: 92% yes, from 13 papers
Its Consensus Meter sorted 13 papers: 12 said yes and 1 said no. That is the 92% in the screenshot. Consensus adds that the meter "may not be a perfect reflection of the current science".
The draft calls this point open. Read those 13 papers before you keep that wording.
<ProTip title="📊 Check how many papers are behind the meter:" description="A split from 13 papers is a reading list, not a verdict" />
Where Jenni is ahead
Every job in one draft - checking, citing, rewriting and exporting all happen in your document.
Complete references - the volume, issue and pages arrive with the citation.
Edits you can accept - accept a Claim confidence suggestion, and Jenni rewrites the sentence and attaches its source.
Citations that stay linked in Word - the export uses Word's own citation format, not pasted text.
What Consensus has that Jenni does not
A meter for yes or no questions - it counts how many papers say yes and how many say no. Jenni has no equivalent, and checks the claims in your own draft instead.
A graph of linked papers - Consensus can draw a citation graph from the first 5 papers of a search. It shows which papers cite which. Jenni instead finds papers one sentence at a time, from inside your draft.
What each tool costs
Plan | Jenni | Consensus |
Free plan | Free - $0; 10 AI Autocompletes (sentence completions) a day; 3 AI Edits (rewrites); 5 AI Chat messages; 3 Reviews (draft checks); unlimited citations | Free - $0; basic paper search; 10 Pro messages a month; up to 3 Deep reviews a month |
First paid plan | Plus - $30 a month, or $144 a year | Pro - $20 a month, or $144 a year |
Top plan | Pro - $73 a month, or $348 a year | Deep - $65 a month, or $540 a year |
Prices from jenni.ai/pricing and consensus.app/pricing.
Paying yearly - Jenni Plus and Consensus Pro both cost $144 a year.
Paying monthly - Consensus Pro is $10 a month cheaper than Jenni Plus.
What Jenni Plus covers - 500 AI Edits a month, the rewrites used for the clunky sentence. It also covers 10 Reviews a month. Peer Review is one of the Reviews.
What Jenni Pro adds - no caps on AI Edits or Reviews.
Which one should you use?

If your supervisor checks every reference
Use Jenni - the reference arrives with its volume, issue and pages. Weak wording, like the hedge words, gets flagged before your supervisor reads the draft.
<CTA title="Hand your supervisor a Word file with linked citations" description="Jenni exports your draft to Word with each citation still linked to the reference list" buttonLabel="Start Writing Free" link="https://app.jenni.ai/register" />
If you are exploring a question before you write
Use both - ask Consensus how the studies split, then export its papers as an RIS file for Zotero. Write and cite the chapter in Jenni. Jenni can import from Zotero and Mendeley.
If your paragraph already has citations
Use Jenni - run Claim confidence, Jenni's claim check, to fix the claims inside the paragraph. Then run Peer Review, its reviewer-style read, for comments on what still needs work.
Start your next chapter in Jenni, so each citation, its full reference and the comments on your argument are in one document.
