Cycling through all six. Tap any point to stop.

Measured on six things

Plume rewrites AI drafts into natural, human-sounding writing, fact-checks every sentence against your original, and flags any inaccuracies so you canhttps://www.plume-ai.org

Speed

Desktop Lighthouse 94/100, but mobile drops to 70/100 with a 5.4s LCP and 4.1s FCP. CLS 0 and TBT 0ms show the layout is stable; the mobile first paint is the weak link.

Screenshots

1 / 2

Plume

About this app

Most AI rewriting tools have two problems: the text still sounds robotic, or it sounds human but quietly changes your facts. Plume fixes both. Paste an AI-generated or rough draft, pick a voice, and Plume rewrites it into natural prose that reads like a person wrote it. Every passage is checked against your original: names, numbers, dates and claims have to survive, and you watch a live fact-check confirm it as the text is written. Plume doesn't just rewrite — it reviews its own work. An AI editor scans the final text and flags anything worth a second look: facts that drifted, details that were added or dropped, ambiguous wording and grammar slips. Each mark comes with a short note explaining the issue, so you can spot inaccuracies instantly and rewrite them yourself before you publish.

Key features

1Passage-by-passage fact check against your original: names, numbers and claims must survive the rewrite.
2Six distinct voice registers (casual, blog, professional, storyteller, reddit, academic) plus English levels and optional natural typos.
3Editor's marks on the Unlimited plan flag meaning drift, changed figures, ambiguous wording and grammar slips with notes.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Content marketers and bloggers

Pain points

AI drafts sound robotic and get flagged by detectors.

Primary needs

Produce human-like content quickly while maintaining factual accuracy.

Demographic

Students and academics

Pain points

AI-written essays are easily detected and risk academic penalties.

Primary needs

Humanize AI-assisted writing to sound natural and avoid detection.

Demographic

Professional writers and journalists

Pain points

Paraphrasing tools alter facts and produce unnatural text.

Primary needs

Rewrite drafts in a professional voice with fact-checking assurance.

Scout Review

Written by AI from measured evidence, scored out of 100.

Plume is a fact-checking rewriter aimed at people who need AI drafts to sound human without quietly inventing details. Its strongest asset is the passage-by-passage audit against the original, which none of its listed alternatives offer. The landing page backs the claim with dated detector tests and a live editor mockup. Pricing is sharp: free at 1,000 words, Premium at €5 for 100k words, Unlimited at €15 for editor's marks and unlimited words. The weaknesses are mobile first paint (5.4s LCP), a missing CSP and DMARC, two accessibility failures, and the decision to paywall the editor's marks that the entire pitch depends on. It is a competent, clearly positioned early-stage tool with one genuinely differentiated feature.

Measure by measure

Usability

Paste → voice → live fact-check is a clean three-step flow, and 1,000 free words with no card is a real trial. Editor's marks sit behind the €15 Unlimited tier, so the headline feature is paywalled.

Design

Bold editorial landing page with a real before/after detector demo and a live fact-check mockup. Voice cards are distinct and well-labeled. No visual audit captured, so polish is judged from structure and copy alone.

Speed

Desktop Lighthouse 94/100, but mobile drops to 70/100 with a 5.4s LCP and 4.1s FCP. CLS 0 and TBT 0ms show the layout is stable; the mobile first paint is the weak link.

Security

HTTPS enforced with HSTS and most hardening headers present; 9/12 on the probe. Missing CSP and DMARC, and no SOC 2, ISO or bug-bounty claim found. Stripe handles payments and texts are deleted after seven days.

Accessibility

Lighthouse accessibility 89/100 with two real failures: insufficient color contrast and no main landmark. English levels and six voices add reach, but no WCAG statement or keyboard/screen-reader testing could be verified.

Growth

A real, specific problem — AI rewrites that quietly change your facts — with a passage-level fact check QuillBot, Scribbr and ZeroGPT don't offer. Detector-evasion is a moving target, which caps how durable the wedge is.

Improvement areas

Editor's marks are gated to the €15 Unlimited plan, so the feature the pitch depends on is unavailable to free and Premium users.
Mobile Lighthouse performance is 70/100 with a 5.4s LCP and 4.1s FCP; desktop is 94/100.
No Content-Security-Policy header and no DMARC record; no SOC 2, ISO or bug-bounty claim found.

Design & Usability

Plume's landing page is unusually honest for the category: it shows the AI draft scoring 100% on QuillBot, Scribbr and ZeroGPT detectors next to the rewrite scoring 0%, and it dates the test (6 Sep 2026, 330-word essays, three voices). The six voice cards — casual, blog, professional, storyteller, reddit, academic — are presented as distinct registers with sample lines, not synonym swaps, and the live fact-check mockup with its editor's marks is the clearest explanation of the product on the page. Usability follows the same logic: paste a draft, pick a voice, watch the passage-by-passage audit, then review the editor's marks. The free tier gives 1,000 words a month with no card, which is a genuine trial rather than a demo. The catch is that the editor's marks — the feature the whole pitch rests on — are gated to the €15 Unlimited plan, so free and €5 Premium users get the rewrite without the review layer that supposedly makes it trustworthy.

Speed & Security

Speed is split cleanly by device. Desktop Lighthouse performance is 94/100, but mobile lands at 70/100 with a 5.4s Largest Contentful Paint and 4.1s First Contentful Paint. Cumulative Layout Shift is 0 and Total Blocking Time is 0ms, so the page is stable and the main thread is idle — the problem is first paint on mobile, not jank. For a tool whose users paste long drafts and wait for a live fact-check log, a slow mobile first impression is a real friction point. Security is competent for the stage: HTTPS is enforced with HSTS, and frame protection, X-Content-Type-Options, Referrer-Policy and Permissions-Policy are all present, giving 9/12 on the header probe. The gaps are a missing Content-Security-Policy and no DMARC record, and there is no SOC 2, ISO, pen-test or bug-bounty claim to be found. Payments run through Stripe and the site states texts are deleted from servers after seven days, which is a reasonable data-handling posture for an early product.

Accessibility & Growth

Accessibility scores 89/100 on the Lighthouse audit, with two concrete failures: background and foreground colors lack sufficient contrast, and the document has no main landmark. Those are fixable in an afternoon, but they matter for a text-heavy product where reading is the entire job. On the plus side, Plume offers English levels from native down to basic and six voice registers, which is real reach for non-native writers. No WCAG statement, keyboard-navigation testing or screen-reader support claim could be verified. Growth is the more interesting question. The problem Plume names is real and specific: AI rewriting tools either sound robotic or silently change your facts. The passage-level fact check against the original is a genuine differentiator — QuillBot, Scribbr and ZeroGPT are paraphrasing and detection tools, not fidelity auditors. The target customer is named (content marketers, students, professional writers) and the pricing is aggressive at €5/€15. The limit is that detector-evasion is a moving target; the durable wedge is fact fidelity, and that is the part worth building the brand on.

Conclusion

Plume is worth trying if your actual pain is facts drifting during a rewrite — that is the problem it solves better than the alternatives, and the free tier is enough to test it on a real draft. If your pain is purely detector scores, know that the detector landscape shifts and Plume's own page admits it re-tests every release. The product would be stronger if editor's marks were available on Premium rather than only Unlimited, and if the mobile LCP came down. As it stands, it is a focused tool with a real wedge and a few fixable rough edges.

Suggestions for improvement

Move editor's marks to the €5 Premium tier or offer a limited monthly quota on free — it is the feature the pitch rests on.
Fix the mobile LCP (5.4s) by optimizing the hero and detector-comparison assets for first paint.
Add a Content-Security-Policy header and a DMARC record to close the two security-probe gaps.
Fix the two Lighthouse accessibility failures: color contrast and a missing main landmark.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Plume	QuillBot	Scribbr	ZeroGPT
Fact-checking against the original draft	Passage-by-passage audit with live log and editor's marks on Unlimited	No dedicated fact-checking; paraphrases without fidelity audit	Citation and plagiarism checking, not rewrite fidelity auditing	AI detection focus; paraphrasing without fact-fidelity checks
Voice / register options	Six voices: casual, blog, professional, storyteller, reddit, academic	Multiple paraphrasing modes (standard, fluency, formal, etc.)	Academic-focused rewriting and citation tools	Paraphrasing feature alongside AI detection
Free tier	1,000 words/month, casual and academic voices, live fact check, no card	Free paraphrasing with word limits and mode restrictions	Free tools available; premium for full citation and plagiarism features	Free AI detection with limited paraphrasing
Paid pricing	Premium €5/mo (100k words); Unlimited €15/mo (unlimited words, editor's marks)	Premium subscription tiers for unlimited paraphrasing and modes	Premium plans for plagiarism and citation features	Free core detection; paid tiers for extended use
AI detector positioning	Claims 0% AI on QuillBot, Scribbr, ZeroGPT; re-tests every release	Offers its own AI detector alongside paraphrasing	AI detector aimed at academic integrity	Core product is AI detection