QuillBot Review (2026): The Paraphraser on Academic Text

Does QuillBot keep "87 percent at the higher dose" intact when it rewrites a clinical results sentence? In Fluency mode its Paraphraser did, and the Citation Generator built an APA 7th entry for the NEJM trial from the title alone.

Two other tools were less careful with that sentence. A Grammar Checker card stripped the main verb from it, and the Summarizer cut the dose that the 87 percent belongs to. QuillBot suits rewording and reference lists, as long as each figure is checked against the source afterwards.

<CTA title="See what an edit changed before you accept it" description="Improve fluency rewrites the sentence, marks every change and says why, before anything replaces your text" buttonLabel="Try Jenni Free" link="https://app.jenni.ai/register" />

TLDR

What QuillBot does for a postgraduate writer

QuillBot, a Learneo, Inc. business, is a set of separate writing tools, each on its own page and listed in a sidebar menu. The ones built for writing a paper:

• Paraphraser - rewrites pasted text in a chosen mode, with a Synonyms slider that sets how far the vocabulary moves.

• Grammar Checker - now a document editor with a Writing assistant panel, four writing scores and buttons to Improve, Rewrite, Continue writing or Ask AI.

• Summarizer - Paragraph, Bullet Points and Custom modes, with a length slider from Short to Long.

• Citation Generator - searches by title, URL, DOI, ISBN or keywords and formats in "over 1,000 different styles".

• AI Chat - a chat marked Beta, with tools for web search, paraphrasing, summarizing, translating and images.

Outside the browser, QuillBot has a Word add-in, and its site lists browser extensions and desktop and mobile apps as well.

Who should use QuillBot?

For a Masters or PhD student, QuillBot is a rewording tool and a reference formatter. It suits background paragraphs where the wording has to move and nothing numeric is at stake.

It also suits reference lists where the title of each source is already known. It is not the place to draft results or methods: three of the five tools tested here changed the trial sentence or sourced it badly.

How the test ran

One paragraph went through five QuillBot tools, each left on its default settings unless noted.

• The paragraph - 201 words on lipid nanoparticles as a CRISPR delivery vehicle, written for these comparisons and used whole or in part in each tool.

• The planted faults - an agreement error, "There is a lot of things that still needs", and a stacked hedge, "could perhaps maybe". Four compound adjectives also lack their hyphens: "E mediated", "LNP delivered", "long term" and "off target".

• The fact to protect - serum transthyretin down a mean 87 percent at the higher dose (Gillmore et al., NEJM 2021). The PubMed abstract gives 52 percent at 0.1 mg per kilogram and 87 percent at 0.3.

• The record - counts, timings and every quoted output were copied from the screen during the session.

The five tools, one paragraph

Paraphraser: Fluency kept the result whole

Paraphraser - the test used the paragraph's three middle sentences, 96 words: the trial result, the clunky sentence and the hedge. The test ran Fluency and Standard, the default mode. Each run returned its output 9 to 12 seconds after the click.

Fluency kept the trial sentence whole: "The first in-human systemic trial of LNP-delivered CRISPR therapy found an 87 percent mean reduction in serum transthyretin at the higher dose after a single infusion." It also added the hyphen in "LNP-delivered".

Paraphraser in Fluency mode: the rewrite of the three sentences, with mean, 87 percent and the higher dose kept.

Fluency fixed the clunky sentence in 35 words, opening "However, many questions remain to be answered". The hedge kept "It is possible that" and turned "main" into "primary": "It is possible that ionizable lipid chemistry is the primary mechanism for extrahepatic targeting, but this has yet to be demonstrated conclusively."

Standard took the clunky sentence from 43 words to 34:

"However, there are still many unanswered questions, such as how the particles behave outside the liver, whether repeated dosing triggers an immune response, and what the long-term editing results look like at off-target sites."

Its hedge lost "perhaps maybe" and gained a courtroom phrase: "Although it hasn't been proven beyond a reasonable doubt, ionizable lipid chemistry may be the primary lever for extrahepatic targeting." Beyond reasonable doubt is a criminal standard of proof, and "hasn't" is a contraction in a formal sentence.

The mode bar and its More menu hold eight more modes, from Formal and Academic to Custom, and the test ran two.

<ProTip title="⚖️ Read the hedges too:" description="Match every figure and every hedge in a rewrite against the original before keeping it. Standard turned not conclusively shown into a courtroom standard of proof" />

Grammar Checker: 12 corrections, one that breaks the sentence

Grammar Checker - the full paragraph went into the document editor, which counted 201 words and detected the language as English (AU). Between 8 and 27 seconds after typing stopped, the Writing assistant listed 12 grammar corrections and an Accept all button.

The scores read Grammar 85, Fluency 49, Clarity 57 and Engagement 93, labelled Average, with a note suggesting a score of at least 80. Most of the 12 cards were right:

• Agreement - "There is" to "There are", and "needs" to "need".

• Hyphens - E-mediated, long-term and off-target.

• Punctuation - the comma splice after "circulation" became a semicolon, and "though" and "for example" gained their commas.

• The hedge - "maybe" removed, which leaves "could perhaps be", still two hedges in a row.

Card 3 of 12 went wrong. It offered to add a word, turning "LNP delivered CRISPR therapy reported" into "therapy and reported", and the sentence that results has no main verb.

The real fault is the missing hyphen in "LNP-delivered", and no card flagged it.

Grammar Checker, card 3 of 12: Add a word, turning therapy reported into therapy and reported.

Two more cards followed the detected English (AU): "percent" to "per cent" and "ionizable" to "ionisable". The language menu sits beside the word count, so the variety can be set before a check.

Accept all applied the 12 cards, the broken one included. The scores then read Grammar 100, Fluency 88, Clarity 82 and Engagement 94, labelled Excellent, above a sentence with no main verb.

<ProTip title="🛑 Card by card:" description="Skip Accept all on a technical paragraph. One wrong card applied in bulk removed the main verb from a results sentence, and the score went up" />

Summarizer: 104 words, and the dose went missing

Summarizer - in Paragraph mode, with the length slider where the page left it, the 201 words became 6 sentences and 104 words in about 2.5 seconds. An Export icon and a Paraphrase Summary button sit under the result.

The trial sentence became "The first systemic trial showed an 87% reduction in serum transthyretin after one infusion." "Mean" and "at the higher dose" are gone, so 87 percent reads as the whole trial's result. The abstract puts the lower dose at 52 percent.

Summarizer, Paragraph mode: the trial sentence in the input, and the summary sentence that drops the dose.

Two claims also changed. "Could perhaps maybe be the main lever" became "Further research into ionizable lipid chemistry may enhance extrahepatic targeting", and "a mature platform" grew into "proven for hepatic gene editing".

<ProTip title="✂️ Watch the qualifiers:" description="Summaries drop conditions first. If a figure applies to one dose, one group or one time point, write that condition back in" />

Citation Generator: the right paper from the title

Citation Generator - the page opens on APA 7th edition. The trial title returned 20 results, with the 2021 NEJM paper first and a 2022 reprint under the same title second.

Each result row shows the title, source type, year and surnames. There is no journal name or citation count, so the year and the surnames are the only clues that separate the paper from the reprint.

One click turned the first result into a reference list entry:

The author list follows the APA 7th form for 21 or more authors: 19 names, an ellipsis, then the last author. Journal, volume, issue, pages and DOI are right.

Two details are not. The title prints "CRISPR-CAS9" where the paper has "CRISPR-Cas9", and "Täubel" lost its umlaut.

Citation Generator: the entry for the NEJM trial, with 19 authors, an ellipsis and the last author.

The Suggestions tab for this source read "All good! We've identified no suggestions to improve or clarify your citation." The list menu offers Export to Word and Copy to clipboard, and the in-text form is (Gillmore et al., 2021).

<ProTip title="📚 Open the DOI:" description="Before pasting a generated reference, open the DOI and check author names, accents and gene names against the paper. Capitals and diacritics are where entries slip" />

AI Chat: a press release for a journal claim

AI Chat - with Web search switched on, the chat was asked to find the source for the trial sentence and give its APA 7th reference. It answered within 19 seconds.

It named a Regeneron press release dated 26 June 2021, not the NEJM paper. The figure and dose were right, 87 percent at 0.3 mg/kg, but the reference it wrote is for the press release and has no DOI.

Where QuillBot slips

The slips across the five tools:

• Paraphraser in Standard - turned "not conclusively shown" into "beyond a reasonable doubt", a criminal standard of proof.

• Grammar Checker - proposed "therapy and", and Accept all left a sentence with no main verb under a Grammar score of 100.

• Summarizer - dropped the dose that the 87 percent belongs to.

• AI Chat - cited a press release where a journal article exists.

• Citation Generator - the right paper and the right author format, two typing slips, and a Suggestions tab that found nothing.

None of this rules QuillBot out for academic work. It means every QuillBot output on a results or methods sentence needs a line-by-line check against the source.

What QuillBot costs

Standard prices from QuillBot's own pages, in US dollars, with each tier described in QuillBot's words.

October 2026 pricing, from quillbot.com/premium and QuillBot’s Premium price page.

Premium can also be billed every three months, at $39.95. Paid yearly, Premium works out to $8.33 a month, the figure the Premium page shows.

QuillBot against Jenni and Grammarly

Jenni - a writing editor where the citation goes in at the cursor and the reference list builds itself under the text. On the same clunky sentence, Jenni's Improve fluency gave a 28-word rewrite in four seconds and showed what changed and why before replacing anything.

Jenni also covers two jobs the QuillBot tools tested here do not: five Reviews, including Claim confidence for claims that lack a source, and export of the whole document to Word with live citation fields or to LaTeX with a .bib file.

Grammarly - on the same paragraph, Grammarly flagged "LNP delivered" as needing a hyphen, the fault QuillBot answered with "and". The full side by side is in QuillBot vs Grammarly.

The same test paragraph has also been through ChatGPT, Claude and Paperpal in earlier tool reviews.

Is QuillBot worth it for a thesis?

For rewording background paragraphs and turning known titles into reference entries, yes, with Fluency as the mode and every figure checked by hand. Keep the DOI of each source, because the generator's entries still need a look.

<CTA title="Cite the trial at the cursor, reference included" description="Search papers from inside the editor, insert the citation and let the reference list build itself as you write" buttonLabel="Start Writing Free" link="https://app.jenni.ai/register" />

For the chapters that carry results, draft the thesis in Jenni, where the citation, its reference and the claim check stay in one document. Bring QuillBot in for the occasional sentence that needs new wording.