Skip to content
MakerGain
← Back to cases

Making AI writing “sound human”: Natural Write reported US$100,000 in three months, but its website stops short of promising a pass

Edited by MakerGainPublished: Article updated:

Natural Write humanizes text generated by tools such as ChatGPT, with detection, polishing, and citation checking sharing one word allowance. Seeing humanizers trend on TikTok, Nikita and Yevgeny decided to build a product in two weeks, with Alexander handling the technology. Paid influencers converted poorly; they then created their own accounts, copied viral formats, and posted daily. Their first self-made video brought the first sale. The pricing page advertises bypassing Turnitin, while the FAQ and terms stop short of guaranteeing a pass. Subscriptions emphasize annual billing collected upfront.

Every plan’s feature list on Natural Write’s pricing page includes “Bypass Turnitin.” Basic names Turnitin, GPTZero, and Quillbot; higher tiers list more (pricing page[1], as of November 18, 2026). Yet the same website’s FAQ says no output is guaranteed to bypass third-party AI detectors, and that it is not affiliated with Turnitin (official FAQ[2]). One page sells the promise; another hands it back to the user. That contrast falls squarely on the boundary of what the product delivers.

The product itself is simple. The FAQ describes a short workflow: paste text from ChatGPT, Claude, Grok, or another AI tool, click a button, and get improved output within seconds. Natural Write sells a humanizer—the act of making writing sound more human.

Its main workflow comes after generation

But there is more than one button. The FAQ lists AI Detector, AI Humanizer, Paraphraser, Grammar Checker, and Citation Checker. Detection, rewriting, polishing, and citation checking all circle around the same text.

According to the About page, these tools run on the web, iOS, Android, macOS, and an API. One account and one word balance span every tool; desktop and mobile apps share the web balance, and Doc Editor saves automatically as users type (About Natural Write[3]). The practical implication is simple: users do not pay for each feature separately. Words are the common currency across tools.

The main workflow handles existing text. The About page states that plainly: it takes “words that already exist”—a rough draft or a paragraph that is not yet flowing—and makes them clearer. But it does not entirely avoid the writing step. Doc Editor has autocomplete: pause while writing, and it suggests the next sentence. Enter accepts it as written; Tab accepts a rewritten version. Ignored suggestions do not consume words (official FAQ[2]). Its main workflow sits after generation, but it keeps a hook reaching forward.

Image: From the official product demonstration[2], obtained on 2026-11-18; official presentation retained as shown, for product illustration only.

They saw others go viral and decided to build it in two weeks

On September 6, 2025, Starter Story published an interview with Nikita and Yevgeny, who said they were in Serbia and Poland, respectively. A third member, Alexander, handled the technology (original interview[4]).

Nikita described an incidental origin: during research, they noticed many humanizer videos on TikTok and concluded, “We can do the same thing.” It would be easy to promote on TikTok because it was trending. His timeline was two weeks to build the product, then start promotion. Host Pat Walls summarized the approach on the spot as thinking about growth before the product.

Those “two weeks” explain why the product first took the form of a single-purpose tool better than any technical description. They first identified a type of humanizer video already taking off on TikTok, then worked backward from “we can do the same thing” to build a product. The detection, rewriting, polishing, and citation-checking tools that later grew around the humanizer, along with Doc Editor, all share one word balance. Web and mobile share an account, too.

Nikita’s description matches the website’s workflow: users paste generated text, click humanize, and receive “100 human written” text. He said it could usually bypass most AI detectors and that users were mainly school and university students, with marketers too. “Usually bypasses” was a statement made on camera, not a product term.

The terms are more cautious. The FAQ promises only that models are continually updated to improve performance, while saying results vary by detector. The Responsible Use page draws a broader boundary: it does not represent or warrant that processed content will be deemed original, human-authored, or compliant with third-party policies and standards (Responsible Use[5]). The pricing-page feature, the FAQ’s “no guarantee,” and the disclaimer’s “no guarantee of being deemed human-authored” coexist. Features for hesitant buyers; terms for potential disputes. This looks more like three statements for three situations in the same business than a typo.

They described a clear detour in acquisition. Initially, they paid influencers to make videos, getting some views but almost no conversions. They then made their own TikTok accounts, filmed their own videos, and posted daily. The first video brought the first sale.

Their TikTok method was specific: find videos already viral in the niche, copy them one for one, post one a day, and test original formats after going viral. They also described warming up accounts: a VPN, a target-country phone number, a new email address, and a new Apple ID, then 30–50 minutes a day for 3 to 5 days browsing niche content, liking, commenting, and following. The barrier lay in repetition rather than technology.

Two interview details merit a pause. Their first viral example copied an existing format—a comic scene about an essay due in five minutes. They made it in two hours and had not expected 4 million views. Another video demonstrated “100% AI-generated text in, 100% human-written text out.” The original had about 150,000 views; Yevgeny’s copy reached 1.5 million. The same actions, ten times the result. The sequence described in the interview was to copy proven formats first, then test original formats after videos began going viral.

The numbers themselves need qualification. The interview’s totals were about 9 million TikTok views across all accounts, about 300,000 visits, 250,000 signups, and roughly 80% visit-to-signup conversion. Yet the opening said “single TikTok account,” while the metrics section said “all our accounts.” The opening’s 250,000 downloads also differs from the later 250k signups. These remain founder-reported figures, without attribution by account or video. The website’s own figures do not align either: the homepage says “1M+ users,” while About and pricing say “2M+.”

Sold by the word, charged by the year

Words are the pricing unit. All four displayed tiers are annual plans shown as monthly equivalents: Basic has 5,000 words/month at an equivalent $4.99/month (with $9.99 in the original-price position), billed $59.88 annually. Pro has 15,000 words at $9.99/month equivalent (originally $19.99), billed $119.88 annually. Ultra’s 30,000 words and Unlimited’s unlimited allowance cost $239.88 and $479.88 annually, respectively (pricing page[1]). The Terms say paid subscriptions automatically renew by default until canceled, and annual plans generally charge the full year at once even when displayed as equal monthly prices (Terms of Service[6]).

The free allowance is 300 words, with no credit card required. Visitors get one free citation check; after signup, each citation check deducts 250 words from the balance. Monthly words reset on the subscription start date, while separately purchased one-time word packs remain until used.

The refund threshold is clear: a refund may be requested within 14 days of purchase if fewer than 500 words have been used since that payment. Beyond that, the service is considered delivered. The Terms add that website refunds are generally limited to one per account; used or partially used subscription fees are otherwise nonrefundable. 500 words is a low threshold—enough to get a feel, not enough to process a long essay and then return it.

US$100,000 in three months, with profit unclear

In the same interview, the founders reported revenue of US$41,000, US$32,000, and US$27,000 in their first three months, totaling US$100,000. These are historical figures from the 2025 interview, falling across the three months. They are neither current MRR nor profit.

On costs, the interview listed only servers at 35 a month, Resend at 20 a month, a VPN at 5 per month per five devices, and US phone numbers at 5 each per month. They described running costs as low apart from time. The list omits labor, outsourcing, UGC payments, payment-processing fees, and taxes, so profit still cannot be calculated. The website now also has a UGC recruitment page advertising that creators can earn “$4,000 and more” a month (Influencer Program[7]). Such spending is also absent from that list.

A low barrier—where did the money come from?

For the same humanizing task, competitor NexBot puts a free trial and low annual price at the entrance in its own comparison page, checked on October 14, 2026, not an independent review. Its humanizer offers a 350-word trial without signup and without daily replenishment; Pro is $7.99/week or $39.99/year (NexBot comparison page[8]).

Two things visible in the public information are not model capabilities: a replicable TikTok distribution routine—find viral videos, copy one for one, post daily, warm up accounts—and a subscription structure emphasizing annual payment collected upfront. Every plan on the pricing page is shown as a monthly equivalent, but the whole payment is taken at the start. The material does not say which contributed more to that US$100,000.

What this business reliably delivers is the act of rewriting. The outcome of “passing” is something its own website will not promise. The pricing page prints “Bypass Turnitin” in the feature list; the FAQ takes it back. Each version has a clearly defined audience.

Continue reading

AI writing
14 questions, one business plan: before PlanArmory can replace consultants, people must find it

Built by a solo founder for first-time entrepreneurs, PlanArmory lists submitting a plan to a bank for an SBA loan as one use case. Users do not write the plan themselves. They answer 14 questions. The website claims a plan of 30 pages or more in about 60 seconds, with PDF downloads in paid Pro. In his post, the founder reported 83 signups and 5 paying customers three months after launch. He said being found was the biggest challenge. A month before posting, he had launched a marketing-plan generator as a retention feature.

AI writing
Where a post goes comes before how it is written: RightBlogger's blog automation

Ryan Robinson and Andy Feliciotti founded RightBlogger in 2023, with Ryan handling marketing and Andy product. It starts by establishing where articles go—an existing blog or drafts saved in RightBlogger—then publishes on schedule. After publication, Site Agent checks internal links, broken links, and meta descriptions, with an option to wait for human approval before making changes. Ryan, a blogger for over 15 years, sees fully automated blog publishing as its product-market-fit milestone and describes it as the tools they wanted, though the price was sometimes beyond their budget, when they were content creators.

AI writing
Writing the proposal took longer than the strategy; in 2025, he said the business brought in about $5,000 a month

Writegenic AI handles the stretch from a blank project document to a first draft: users enter project information, generate a technical specification, project plan, or business-document draft, then review and judge it themselves. Founder Umair Quarishi first tested an outline-generation prototype in a small network before expanding to full documents. A quiet launch used a simple landing page, community demos, and trials exchanged for feedback. In June 2025, he reported about $5,000 in monthly company revenue. Profit and whether that revenue continued remain unknown.

AI publishingAI writingAI imagesAI video
Do Not Start From a Blank Page: CaptureFlow Catches What You Say; the Signature Is Still Yours

CaptureFlow catches expression from voice, links, video, or files and hands it to Content Agent to draft; only after the user revises and explicitly approves does it publish to LinkedIn, and responsibility remains with the person. Chris Koronowski runs the company in Poznań, Poland, charging in three tiers—Pro, Business, and Agency—with the latter two priced by seat; a case study says InStream has 100 paid seats, while TrustMRR separately shows 8 subscriptions, and the two cannot be combined into a customer count.