QuillBot

By Minel Gunesoglu, founder of Vouch. This page is built from documents rather than from a session with the product: four Internet Archive copies of QuillBot's pricing page and two rendered readings of it, an archived copy of the humanizer page and a rendered one, QuillBot's own terms and both of Learneo's contracts read in a browser, the vendor's help article on refunds, the Apple, Google Play and Chrome Web Store listings, 247 archived Trustpilot reviews and the 50 most recent US App Store reviews. No text was put through the Humanizer for this page and no detector was tested. Read 11 and 15 September 2026, published 15 September 2026, re-checked monthly.

Four Internet Archive copies of QuillBot's pricing page are held here, taken between 22 July and 31 August 2026, and all four carry the same price list inside their own data: Monthly $19.95, Annual $99.95 at $8.33 a month. What the Premium card above that list displayed was not the same in all four. The July copies showed $8.33 a month billed annually. The copy served with a United States country code on 10 August showed $12.50. The copy of 31 August showed a one-time first-term discount in front of that same $8.33 rate. A rendered read from Turkey on 11 September 2026 met a different table again: an annual plan at $49.95, and no monthly plan in it at all.

QuillBot's AI Humanizer is the feature this page measures, and it is one tool inside a writing suite that also sells a paraphraser, a grammar checker, a plagiarism checker and an AI detector, published under the footer line "Quillbot, a Learneo, Inc. business". Two limits belong before the first heading. This page does not put text through the tool, so nothing here describes what a rewrite reads like. And it does not test what any detector does with a rewrite, so no sentence below claims an outcome of that kind in either direction; where a reviewer describes one, that reviewer is quoted, named and dated, and the claim stays theirs.

What QuillBot's AI Humanizer Is, and Where It Sits in the QuillBot Suite

QuillBot's AI Humanizer takes text you already have and rewrites it to read more naturally. The vendor's page, archived 21 August 2026, lists it fifth in a row of ten tools: "Paraphraser Grammar Checker AI Detector Plagiarism Checker AI Humanizer AI Chat AI Image Generator Translate Summarizer Citation Generator". Nobody buys the Humanizer. They buy QuillBot, and the Humanizer is one of the things a subscription switches on.

The vendor publishes what it considers the tool to be for, on that same page: "Appropriate use cases for Quillbot’s AI Humanizer include a writer making their project sound less technical or a non-native English speaker making their email flow more conversationally. Just be sure to follow your institution or organization’s AI rules and recommendations." That is the seller's wording, dated, quoted rather than summarised, because a seller's statement of purpose is a fact about the seller.

One language fact sits on the humanizer page and nowhere in the plan table: "Humanize AI text in Spanish, German, French, Portuguese, and 4 dialects of English (US, UK, AU, and CA)." A writer working in Turkish, Arabic, Hindi or Mandarin falls outside that list as the vendor states it.

QuillBot's AI Humanizer and QuillBot's Paraphraser: What the Plan Table Says Each One Does

QuillBot's plan table, in the copy archived 31 August 2026, treats the two as separate products with separate meters, which is the firmest evidence available on a question the rest of this market answers by assertion. On the free tier the paraphrasing row reads "Up to 125 words" with no daily cap printed beside it, while the humanizing row reads "Up to 125 words, 6 uses/day". Paraphrasing has its own modes, "Standard and Fluency" on free against "9 unique modes" on Premium, and its own reach, printed as "Paraphrase in 23 languages". Humanizing has neither row. What it has instead is "Get Humanizer insights & history", reading "No history & limited insights" on free and "Full history & insights" on Premium.

So QuillBot meters, documents and limits the Humanizer and the Paraphraser differently, on one page, in one archived copy of it. Whether the text each hands back differs is a separate question, and no row on this card measures it.

Where QuillBot's AI Humanizer Runs: Chrome Extension, Word Add-In and the Mobile Apps

QuillBot's plan table, in that same copy, lists the surfaces in a single row: "Google Chrome extension Microsoft Edge extension macOS app Word add-in Android keyboard and app iOS keyboard and app". The Chrome Web Store listing, read 11 September 2026, is titled "Quillbot: AI Writing Assistant with Humanizer, Paraphraser & AI Detector" and shows 4.7 out of 5 on a rounded "5.9K ratings" beside 6,000,000 users, a figure covering the whole extension rather than the Humanizer inside it. Apple's listing for "Quillbot-AI Writing & Keyboard", read the same day, says the feature reaches the keyboard too: "You can also humanize AI text using the Humanizer mode in our AI Keyboard". One more surface sits outside that row and is named on the humanizer page archived 21 August 2026, where a Premium line offers to "Work faster by humanizing text directly in ChatGPT and Gemini". What no captured page shows is a programmatic route in or out. Every surface named, that one included, is worked by a person.

Is QuillBot's AI Humanizer Free? The 125-Word Ceiling and the Six Runs a Day

Yes, up to a ceiling the vendor states plainly and twice. The humanizer page's free panel reads "Humanize texts of 125 words or fewer", against Premium's "No word or usage limit". The pricing page, in the copy archived 31 August 2026, puts the same limit in its own unit: "Humanize up to 125 words, 6 uses/day". The Turkey-served render of 11 September 2026 shows the identical free line, so the ceiling is not a regional quirk.

The first two rows come from the tier-by-tier table on the pricing page and the third from the plan cards above it, all in the copy archived 31 August 2026. Six runs of 125 words is around 750 words a day if every run is filled, under two pages, and no page read for this review names a reset hour for that allowance. Two smaller answers for anyone about to sign up. The free panel carries a "Create account" button, and whether the tool runs at all without an account was not established here. And the vendor answers the price question itself, dated 21 August 2026: "Yes, we keep our AI humanizer free because everyone should be able to turn stiff writing into direct, understandable language."

What QuillBot's Free Account Licence Allows: Personal, Non-Commercial Use Only

QuillBot's service terms, read in a browser on 11 September 2026, draw a line no copy of the pricing page held here mentions. Section 1 is unambiguous about the free tier: "Free Accounts. If you register for or use a free account, you may use the Quillbot Service only for your personal, non-commercial purposes."

Paying does not automatically move anyone across that line, and the document names the screen that does: "If the applicable subscription or sign-up process does not state that your account or subscription may be used for personal, commercial purposes and/or internal business purposes, then your right to use the Quillbot Service is limited to personal, non-commercial use only." The Premium pricing page states no commercial scope in any copy held here, and the sign-up screen itself was not read for this review, so what Premium grants is a question these captures leave open. One tier settles it inside the document: "Team Plan users may use the Quillbot Service for their personal or commercial purposes, including both internal and external business use." What that tier costs, and how many seats it takes, is not in any capture held for this page, so no price for it appears here.

Ownership runs the other way. Learneo publishes two contracts, and the one this card was read from is headed "Learneo Terms of Service for Non-European Users"; its section II says "we assign to you all of our right, title, and interest, if any, to the AI Outputs", inside a licence granted "for your personal, non-commercial purposes" unless a tier's own terms say otherwise. Section IV keeps the input yours and licenses it back: "You retain all ownership rights over your User Content", followed by a grant to Learneo, its service providers and other users of "a worldwide, non-exclusive, perpetual, royalty-free, irrevocable, sublicensable and transferable license to use, modify, aggregate, reproduce, distribute, prepare derivative works of, and otherwise fully exploit the User Content in connection with any Learneo Services". A humanizer receives a whole draft rather than a prompt, which makes that clause sharper here than for a tool that only ever sees an instruction. Those are the words on the page on the day they were read, and nothing here says how any of them would be applied.

What QuillBot Premium Costs, and Why the Number Depends on the Visit

Every figure below was read out of the bytes of the copy named in its own row, five of the six rows on 11 September 2026, four of them archived copies and one a rendered reading, and the sixth rendered on 15 September 2026. QuillBot's servers refuse this desk's tools outside a rendered browser, and the rendered browser available here answers from Turkey, which gets its own table. That is the honest boundary around this section.

The last two rows are the same page, from the same country, four days apart, and they are not the same table: on 15 September a monthly plan at $9.95 had appeared where the 11 September reading found none, while the other three entries held to the cent. The scored price, $19.95 a month, is the Monthly entry in the archived copies' own price data rather than a figure read off a rendered monthly view: the monthly state of the billing switch was never rendered in any copy held here. The $12.50 on the 10 August copy is not a transcription slip on our side, it is a page displaying one table while its own data carried another, and that copy's payload holds alternative tables with monthly rates of $21.95, $23.95 and $24.95 and annual totals of $119.95 and $149.95 sitting beside the default.

Buying inside Apple's app is a separate channel again: the App Store listing, read 11 September 2026, prints "In-App Purchases Yes Monthly Plan $19.99 Annual Plan $99.99 Semi-Annual Plan $79.99". Those are Apple's prices and they are never mixed with the web ones here. The practical instruction is short. Open the pricing page yourself, in your own country, in the billing state you intend to buy, and read the number in front of you before committing to it.

Can You Cancel QuillBot or Get a Refund? What Learneo's Terms and the Help Article Say

Cancelling and being refunded are different questions here, and the documents answer them differently. QuillBot's help article on refunds, archived 27 April 2026, describes cancellation as self-service: "You can cancel your subscription at any time by turning off auto renewal. You will continue to have Premium access through the end of your billing period."

On money already taken, that same non-European contract’s section I sets the floor: "Except as otherwise provided in writing or as required by law, subscription fees for paid services are charged upfront and nonrefundable." Section VII adds that "We do not offer refunds for any unused portion of a subscription after a User Account has been deleted", so deleting the account is not a route to one. QuillBot's terms section 3 leaves local law standing: "These Terms do not override any mandatory local laws regarding your cancellation or refund rights." The help article names what follows. It says refunds are issued "only if required by law and in accordance with Section 5" of the terms, tells an EEA or UK reader that "you have the right to withdraw within 14 days of purchase of your QuillBot subscription", and routes app purchases away from the vendor: "If you purchased your subscription through the Apple App Store or Google Play Store, contact Apple or Google to request a refund."

Two limits sit on that refund reading. The article QuillBot points a cancelling customer at is a help page rather than the contract, and it was four and a half months old on the day it was read. One of the two clauses it points at was found on 15 September 2026, after this card was scored, and the article had pointed at it correctly. The withdrawal right it sends an EEA or UK reader to Section X for is in Section X, of a second Learneo contract this desk had not opened. Which section 5 the article means could not be settled: the capture of QuillBot’s own terms held here keeps sections 1, 2 and 3 and the Team Plan terms, and a live re-render on 15 September returned the page furniture without the contract body, so this page says nothing about what that section contains. Section X does exist, in a document this desk had not opened: Learneo publishes TWO contracts, and the one the card was read from is headed "Learneo Terms of Service for Non-European Users". The other covers the EEA, Switzerland and the United Kingdom, and there the nonrefundable sentence is written "subject to Section X (Consumer Right of Withdrawal)", with a fourteen-day right set out in full. Read against the rubric both contracts land on the same rung, which is why the 64 does not move: the non-European one makes fees nonrefundable except where law requires, and the European one says the same and then prints what the law requires, describing it in its own words as "a statutory right to withdraw your subscription to a Paid Service" rather than as something the seller is offering. The same section lists exceptions that can extinguish it once delivery of digital content has started, and which of those applies turns on a checkout screen this review could not open. The figure grades documents promising the legal minimum, on a date, and a document is not a record of conduct.

What QuillBot customers report about the statutory minimum is dated and quotable. An App Store reviewer on 8 May 2026, under the title "Refund Refused", wrote: "If you let your subscription renew, they don’t care how quickly you request a refund after the charge posts". Another, on 12 February 2026, wrote: "I was charged for a full year of QuillBot without any effective renewal notice." A Trustpilot reviewer on 8 January 2026 wrote: "It is impossible to cancel the subscription." Each is one named person's account of their own experience on a stated date, and none is about the Humanizer; they describe the account and payment flow under every tool in the suite. Read beside section I, the first is describing the policy working as written.

What Learneo's Indemnity Clause Puts on a QuillBot User

Learneo's terms section IX opens: "You agree to indemnify and hold harmless Learneo and its respective employees, directors, officers, agents, vendors and suppliers from and against any and all claims", and names what those claims may arise from, including your use of the services, the content you submit, and your breach of the terms. Indemnifying somebody means covering their losses and their legal costs when a claim lands on them. A search of Learneo’s Terms of Service for Non-European Users for "indemnif" and "hold harmless" on 11 September 2026 turned up nothing running the other direction. The same search was made in a browser against QuillBot’s own terms that day and its full text was not kept, so this page rests the finding on the Learneo document, which is where the clause lives, and the European contract read on 15 September carries the same clause running the same way, with the scope worded "the Learneo Services" where the other reads "the Services".

A one-way indemnity of that shape is what most consumer software asks for, and this card scores it as ordinary rather than as a fault: the sub-component reads 68. What is unusual here is not the direction of the clause but where it lives. The indemnity is Learneo's, in Learneo's agreement, while the licence that decides what a free account may do with a rewrite is QuillBot's, in QuillBot's. One brand, two contracts, and a reader has to hold both to know what they agreed to.

What QuillBot Reviewers Actually Say About the Humanizer, and What This Page Does Not Test

The Humanizer is a small part of what people write about when they write about QuillBot. Five Internet Archive copies of the blocked Trustpilot profile are held here, spanning 22 April 2025 to 31 March 2026, and their page data carries 247 distinct reviews. Thirteen of the 247 mention the Humanizer by name. The 50 most recent US App Store reviews were read from Apple's public feed and are counted separately. That is a thin seam and it is presented as one.

One fact about the vendor belongs here first, dated, because a reader is entitled to know who is on both sides of this question. Learneo sells an AI detector on the same domain as the Humanizer, and the humanizer page sends readers to it. The copy archived 21 August 2026 reads: "Check how human your text sounds Make sure your text is free of unnatural-sounding language with this Premium feature, or click your score to try Quillbot’s AI Detector". Read live on 15 September 2026 the panel had been reworded and still carries the cross-sell. It now calls the score "a signal on stiff or awkward phrasing as you revise one input among several, not a verdict", and the same passage goes on to ask "Want a second opinion on a draft?" before naming QuillBot’s own detector. Both readings are disclosures, not arguments about how well either tool works.

The thirteen run eight at five stars, three at four, one at two and one at one. Every one that reports an effect on the writer's own text is printed below. Seven report none, and four of those seven are printed here as examples of the shape. A reviewer on 21 April 2025 marked it as new: "Just recently, they have added humanize mode." One on 11 January 2026 wrote: "It's so good this ai humanize". One on 14 January 2026 wrote: "Very good to humanize". One on 16 March 2026 listed it among the rest: "it has many functions such as paraphrasing and humanizing, so it is always useful to utilize it". None of them says what text was run or what changed in it.

The other six describe what happened to their own text, and they are printed one at a time, in the writer's own words, with the date and the star rating each carried. They are not added up, averaged or characterised here. This page runs no detector and measures no rewrite, so it has no finding of its own to set beside any of them, and what a checker does with a rewrite is not a question it answers in either direction.

On 25 July 2025, at two stars: "if you want to use quillbot for ai detection or ai humanizer, just don't, because it's awful. it has 5% success rate". The figure is that reviewer's own and nothing here tests it.

On 27 November 2025, at four stars: "i generated a text from chatGPT and i put it through the ai detector, it gave a score of 76% written by ai. I'm fine with that. So, just to see would would happen, i put the text through the humanizer, the paraphraser. Surprisingly, once i had the new-humanized text, i put it through the ai detector, but this time, it gave the text a 100% written by ai. What's the point of a humanizer that's caught by ai? Either way, this is truly a great app." The review names no checker beyond "the ai detector".

On 21 December 2025, at five stars: "The ai detector is really good and helpful and your humanized ai is really saved me".

On 14 January 2026, at one star: "How do you use paraphraser and humanizer function in premium and the AI detector result is still very high. Its a waste of time".

On 18 March 2026, at four stars: "However, the Humaniser appeared not to be very effective. Anytime I humanise AI-generated content, it is difficult to see any significant reduction in % of AI use."

And on 30 March 2026, at four stars: "The writing was excellent. However, my humanized text was flagged off by turnitin.."

One more voice sits on the other platform and meets the same rule, so it is printed by it. An App Store reviewer on 22 February 2026, at one star, wrote the whole of their review as "Humanizer doesn't work at all." It names no text and no checker, and it is printed rather than dropped because leaving out a review this desk read, on the grounds that it is thin, would be a choice about the record rather than a reading of it.

Only one of the six Trustpilot reports names the tool the text came from, the ChatGPT draft in the November review; one other says only that the text was AI-generated. Turnitin is named once, in March; three of the others point at the detector QuillBot sells itself; two name no checker at all. A reader deciding whether this feature suits their own writing has seven other people's single experiences to weigh, and no measurement from this desk to weigh them against.

Why QuillBot's App Store and Google Play Ratings Are Not Ratings of the Humanizer

91.2 is the biggest number on this card and the easiest to misread. It pools two store records by sample size: Google Play's 4.54 out of 5 across 22,077 ratings for "Quillbot: AI Writing Assistant", and Apple's 4.69 across 2,972 ratings for "Quillbot-AI Writing & Keyboard", both read on 11 September 2026, for a pooled sample of 25,049. Every one of those people rated the whole app. A score for a product bundling a paraphraser, a grammar checker, a translator, an AI detector and a humanizer is not a score for the humanizer, and the usability row should be read as what it is: a large audience broadly satisfied with a writing suite.

The Chrome Web Store's 4.7 is a third record of the same kind and sits outside that cell, because its count is published as "5.9K ratings", a rounded display, and the pooling arithmetic needs an exact one. Neither Google Play nor Apple publishes whether its reviewers were prompted or arrived unasked, and nothing here can tell you which.

QuillBot on Reddit and G2: What Could Not Be Read for This Review

Reddit was closed to every method tried on 11 September 2026: the JSON endpoint returned 403 under two user agents, the same path under a Googlebot agent returned a "Blocked" page, and old.reddit.com answered with a logged-out consent page instead of the thread. Product Hunt met every guessed address with a Cloudflare challenge. Capterra was blocked live, and the Internet Archive settled nothing: one query returned no rows and a second, broader one timed out on the Archive’s side and was not retried, so this page makes no claim either way about whether a Capterra listing exists. G2 was readable only through archived copies, and it renders review dates in the browser rather than in the served page, so nothing from it carries a date this page could stand behind; of the ten readable review pairs in the copy archived 4 June 2026, none mentions humanizing. So the customer record behind this page is two platforms wide, Trustpilot and the App Store, and that is its whole extent.

How QuillBot's AI Humanizer Scores 73.4, and Which Row Is Not a Reading of Rewrite Quality

One reading inside that row is worth spelling out, because the obvious number would have been the wrong one. QuillBot advertises 23 languages, and that figure belongs to its paraphraser. The humanizer page names five. This card scores the humanizer, so the language reading takes the middle position rather than the top, and carrying the suite's larger number across would have lifted the row on evidence about a different product. The card was rescored on 14 September 2026, before publication, when this row and the surfaces row were read against the criterion for the first time.

Delivery envelope is not the name of a dimension this site set out to score. It is the name of what went into that slot, and the card discloses that by naming the row Delivery envelope where an ordinary card reads Capability, which is the whole of what it says on the point. The criterion is registered as delivery_envelope_v1, and it asks something narrow and answerable: what does the paid tier let you walk away with, and through which surfaces? Four components carry a value on this tool and all four are worth naming, because the row is the one a reader is most likely to misread. Two were read when the card was first scored. The first is what the paid tier lets you leave with: a paste comes back as editable text and lands inside the buyer's own files through the Word add-in and the browser extensions, which scores 92. That 92 is worth reading for what it is. The rung is met by the deliverable being text, editable by nature wherever it lands, and the add-in and the extensions are where it lands rather than why it scores. A sibling shows the difference: Wordtune’s own product list, read 12 September 2026, publishes no Word add-in, and its card reaches the same rung anyway. A text rewriter hands back text, so inside this product class the rung separates nobody. It is not automatic across the corpus: where a tool publishes a finished file and no editable one, the same component scores 74. The second is how the work is driven, which scores 50: nothing the vendor publishes for this feature runs without a person, and the nearest thing to a shortcut is a Premium row on the humanizer page archived 21 August 2026 offering to "Work faster by humanizing text directly in ChatGPT and Gemini", which moves where the clicking happens rather than removing it. The other two were read on 14 September 2026, before this page went up, because they had never been put to this card at all: the surfaces the paid tier runs on, which scores 92 and is read off that same Integrations row; and the languages it rewrites in, which scores 74 for the reason given above. The concentration is worth stating plainly, because it is the row's real limit. Three of these four components are read off one archived plans page and carry about four fifths of the row between them. Narrowing that to a single line on the page narrows nothing, since the same three components are the ones that line is quoted for; and on the reading this page argues for two paragraphs above, where the surfaces and the driving of the work turn on that line but the editable-text rung does not, the share resting on it is nearer a half. Five further components carry no value, and the rubric files all five under one label that flattens four different reasons. Two are genuine silences: nothing published about marking output, nothing about a revision allowance. One does not apply to a text tool at all, since a rewrite has no resolution or sample-rate ceiling. One is a unit mismatch rather than a silence, because the ladder counts items and what the vendor publishes is "No word or usage limit". And one is decided somewhere else and scored there: the commercial scope of the paid tier is a contract question, and it sits in the bottom row of this same card. The four that remain are not averaged evenly. The row is a weighted mean, and the weights are worth printing so the figure can be checked: 0.144 on the editable-text component, 0.12 on the surfaces, 0.08 on the languages and 0.072 on the one scoring 50, divided through by their own sum of 0.416 because the other five components carry no value. That is why four numbers whose plain average is 77.0 come out at 81.3. None of that touches the sentence a buyer actually wants, which is whether the rewrite is any good. Two cards on this site can print a figure in this row and be reading two different kinds of thing, so it is worth checking before it is trusted across a category.

The bottom row, 59.1, is the weakest, built from the four clause positions quoted higher up: the one-way indemnity at 68, the free tier's personal-use licence at 50, the unresolved commercial scope of Premium at 50, and the statutory-minimum refund posture at 64. Four clauses, read on 11 September 2026, with two of them, the indemnity direction and the refund position, re-read against the second contract on 15 September and neither moved by it. Nothing in that row watches what happens when a customer actually asks. The rubric behind each row, and the list of things this site does not do, are on the methodology page.

Why No Row Here Rates How Well QuillBot's AI Humanizer Rewrites

This desk holds captures of five studies naming QuillBot. Four are independent and the fifth was authored at a detector vendor. All five measure QuillBot's paraphraser, and all five were published before 12 February 2025, the earliest date the Internet Archive has a request on file for the humanizer's own page. A study of a different feature is not a measurement of this one, and stretching it into one would be the easiest way for this page to claim a measurement it does not have.

No independent measurement of the Humanizer's rewriting leaves a gap on this card, and the gap is named rather than filled. Sellers in this market publish a great deal of their own arithmetic about their own output, and a seller’s arithmetic about the quality of its own output is the one source a score here may never rest on, so none of that enters this card in any form. What a seller publishes about its prices, its contract and its surfaces is a different thing and does enter: three of the four rows rest on it. Nothing on this page will tell you whether QuillBot’s rewriting reads well to you. The free tier will show you the free rewrite, at 125 words a run, and the vendor’s own panel says that is not the paid one: free is listed as "Basic humanizing" against Premium’s "Advanced humanizing".

Who QuillBot's AI Humanizer Suits, and Who Should Not Pay for It

Buy it if you already want QuillBot. The Humanizer is one switch on a suite rather than a product sold on its own, so the real question is whether the paraphraser, grammar checker, summarizer and the rest are worth $19.95 a month to you, with the humanizing included. The strongest thing on this card speaks to that: two store records covering 25,049 ratings of the whole app, which is what the 91.2 converts.

Think twice before paying if the rewriting is client work, marketing copy or anything else done for money on a personal subscription, at least until you have read what the sign-up screen says about commercial scope, because the terms hand that decision to a screen this review could not open and set the default to personal use only. Think twice if you write in a language outside the five the vendor names for this feature. If the rewriting has to happen inside something automated, this is not the product, because everything QuillBot publishes for it is driven by a person clicking. And whatever you decide, put the renewal date in your own calendar: Learneo's terms make fees nonrefundable except where law requires otherwise, and two of the customers quoted higher up describe meeting exactly that.

Is QuillBot's AI Humanizer Good for Students and Academic Writing?

Two facts decide most of this and neither is a judgement. The free ceiling is 125 words a run and six runs a day, so a 2,000-word essay does not fit through the free tier in one sitting. And the free account is licensed for personal, non-commercial use, which is what coursework is. On the second question every student actually asks, this page has nothing to offer: it does not test what any checker does with a rewrite, it does not advise on it, and it treats the accounts quoted higher up as what they are, individual dated reports rather than a measurement.

The vendor moved on this between the reading and the publication of this card. On 15 September 2026 the humanizer page carried a line absent from the only earlier copy of that page held here, the Internet Archive copy of 21 August 2026, in which the string "evad" does not occur at all: "The Humanizer improves how your writing reads — it isn’t a tool for evading AI detection. If you used AI, disclose it per your institution’s or employer’s policies." That is the seller's position on the question this page declines to answer, and it is carried here as the seller's. The older instruction on the same point still stands, in its own words: "Just be sure to follow your institution or organization’s AI rules and recommendations." Those rules, not this card and not the seller's marketing, govern a submitted piece of work. What each row on a card in this category rests on is on the AI humanizers ledger.