Frequently asked questions

Three steps. You show Döppelscript examples of your writing, by pasting posts, uploading a file, connecting a platform, or picking traits on sliders. Döppelscript analyzes the samples and turns them into plain-English rules about how you write. You can read those rules. You can edit them. Then you type a topic and Döppelscript generates a post that follows the rules, in your voice, not in ChatGPT's.

Those tools have one voice: the voice of whatever an LLM sounds like by default. That voice has a shape. It leans on “delve, ” “leverage,” and “in today's digital landscape.” It hedges. It writes in bullet lists when you ask for prose. It sounds like it was written by a committee that has never met you. Döppelscript is built around the opposite premise. Every generation is filtered through rules extracted from your own writing, and the output is post-processed to strip the AI tells at the character level before you ever see the post. The brand is “Your voice, amplified,” not “our voice, branded as yours.”

It depends on how you build it. The trait sliders take about thirty seconds. Pasting three or four samples takes under a minute of reading time and about thirty seconds of analysis. Uploading a file takes a little longer because the analysis runs against more content. Connecting a platform (LinkedIn, Medium, Substack) imports historical posts automatically and is the most accurate option. There's no wait-for-approval step. As soon as the profile says ready, you can generate.

No. Never. If you choose to connect your LinkedIn account for importing past posts, that goes through LinkedIn's official OAuth flow, which means LinkedIn controls the permission and you can revoke it at any time from your LinkedIn settings. Döppelscript never sees or stores your password. If you'd rather not connect anything, pasting posts or uploading a file works just as well.

Döppelscript is built around LinkedIn-sized, LinkedIn-shaped posts as the primary output, because that's where the everyone-sounds-the-same problem is the loudest. The voice profile itself is platform-agnostic, and the rewrites work fine for Twitter, Substack intros, newsletter openings, and similar short-form content. Long-form blog posts and essays are not what Döppelscript is optimized for, and you will get better results from a different tool.

Exactly one thing: extract the patterns that make up your voice profile, and then delete the file.

In more detail: the upload is written to Cloudflare R2 temporarily, a background worker downloads it once, parses the posts out, sends the text to Claude to extract your voice patterns and rules, writes those patterns to your voice profile in the database, and then immediately deletes the original file from R2. The raw text is never written to the database at any point. The only thing that persists after analysis is the voice profile itself, which is the abstract rules about how you write, not the writing samples you uploaded.

We don't keep it. After the voice profile is built from your uploaded file (usually within a minute or two), the raw file is deleted from storage and nothing about the original text remains on Döppelscript's servers. The voice profile that gets built from it stays in your account (that's the output you paid for), but the source material does not. If you want to upload the same file again tomorrow because you want to rebuild the profile, you have to re-upload the file because Döppelscript no longer has it.

No. Döppelscript does not train any AI models. Every generation uses Claude via Anthropic's API, and Anthropic's API terms state that content submitted through the API is not used to train their models. That is an Anthropic policy, not a Döppelscript promise, but it's the policy that applies to every request we send on your behalf. Your writing goes to Anthropic, Anthropic answers the request, nobody trains on it.

Yes. Voice profile settings has a “Delete Profile” button with a confirmation dialog. Deletion cascades to the uploaded file (if any is still present), the extracted rules, the pattern records, any generations made from the profile, and every associated piece of data in the database. The operation is irreversible and takes effect immediately. There is no thirty-day “recycle bin.” If you click delete, the profile is gone.

Self-service deletion lives in Settings with two modes.

Soft delete (the default). Your account is marked deleted, you're logged out everywhere, and you can't log in. Your voice profiles, generations, and uploaded files stay in the database indefinitely so support can restore the account if you change your mind. Email support@doppelscript.com if you want to come back.

Hard delete. Your account, voice profiles, generations, uploads, and all related data in the database and storage are permanently erased. Not recoverable by support, not recoverable by anyone. The one exception: Stripe retains transaction records (invoices, charges, payment method metadata) on their side for as long as their billing-compliance rules require, and Döppelscript cannot delete those. If you want that record minimized, contact Stripe directly.

For reference on the corpus itself: your uploaded writing samples aren't in this discussion because they were already deleted at the moment the voice profile was built.

You get thirty days' notice by email. You get a one-click export of your voice profiles, generation history, and uploaded files in a machine-readable format. At the end of the thirty days, every database row and every file in storage associated with your account is permanently deleted. This is a commitment made in writing in the Terms of Service, not a marketing promise.

Each generated post costs ten credits. Regenerating a post costs another ten credits, because “regenerate” is literally running the whole thing from scratch with the same topic. Refining a post using the “make it less AI-sounding ” button is free. You already paid for the original generation, and if it came out sounding generic enough that the quality judge flagged it, asking you to pay again to fix a draft you didn't love would be a tax on Döppelscript's own failure. Refines stay free.

Credits are $0.05 each, so ten credits is fifty cents per post. Packs start at $10 for 200 credits. Bigger packs get bigger discounts, but the unit price never goes below the actual Anthropic API cost, because that would mean losing money on every generation.

No. Credits never expire. Buy them now, use them in five years. The subscription trap where your credits reset every month exists to pressure you into using features you don't need on a schedule that serves the vendor. Döppelscript doesn't do that.

No. Sign up with just an email and get a hundred free credits, which is ten posts. No credit card, no trial window, no forced onboarding sequence. If you run out, you can buy more. If you don't, the free credits stay in your account until you do.

They stay exactly where they are. Voice profiles are free to build, free to keep, free to edit. You only pay credits for generating new posts. If you run out of credits, the worst that happens is the “Generate” button becomes a “Buy more credits” button. Nothing disappears.

Yes, on unused credits, within fourteen days of purchase, no questions asked. Email support@doppelscript.com from the address on your account and the money goes back to the original payment method. Credits you've already spent on generations are not refundable, because the API cost to generate those posts has already been paid to Anthropic and the money is gone. If a specific generation came out so badly that you want that credit back, send the generation ID in the same email and we'll evaluate on a case-by-case basis. The default is yes.

You need a LinkedIn account with followers. You submit your LinkedIn profile URL and a screenshot of the profile showing your current follower count. Every submission is reviewed manually by a human (currently just Ian) within one business day. Once approved, you get one credit for every thousand followers, deposited into your account. No OAuth required, no application form, no minimum follower threshold, no waiting list.

The manual review is deliberate. Automated verification via LinkedIn's API is not straightforward (follower counts are not exposed through standard OAuth scopes), and scraping LinkedIn is against their terms of service. Human review is the honest path until the volume of submissions justifies something more elaborate.

Yes. You can submit a new verification once every thirty days. Each time, Döppelscript awards credits based on your current follower count minus whatever was awarded in your last verification, still at one credit per thousand followers. If your follower count grew since last month, you get credits for the growth. If it stayed flat, you still get a small baseline grant for being part of the program. The intent is to reward ongoing audience-building, not to reward a single upload from years ago.

No. Same policy as purchased credits. Earned credits are credits, full stop.

No, but it's a fair question and I'll answer it seriously.

Cheating implies passing off as your own something that isn't yours. Döppelscript doesn't create anything from scratch. It learns the patterns in your existing writing (your sentence rhythms, your word choices, your stance on the topics you write about) and generates new text that follows those patterns. The voice is yours. The assembly is assisted. Using Döppelscript is closer to using spellcheck or Grammarly than to hiring a ghostwriter, because the voice being produced is mechanically derived from your own writing rather than being imported from a stranger.

That said: if you'd feel uncomfortable telling a colleague you used Döppelscript to write a specific post, that discomfort is information worth listening to. Different contexts have different honesty norms. A LinkedIn thought-leadership post is not the same as a peer-reviewed journal article. You are the only person who can decide whether using a tool feels right for a specific piece of writing, and Döppelscript will never try to talk you out of saying no.

Ghostwriting is when somebody else's ideas, phrasing, and judgment become the work you publish under your name. Döppelscript flips every piece of that. Your ideas, your topic, your stance, your edits. The tool handles the sentence assembly. The “ghostwriter” in this analogy is not a person with their own opinions; it's a mechanical process that takes the rules you wrote and turns them into sentences.

There's one case where using Döppelscript would be dishonest, and it's important enough to say out loud: if you feed Döppelscript somebody else's writing and try to pass the output off as your own voice, that's dishonest and it's also a terms-of-service violation. Voice profiles must be built from your own writing.

If Döppelscript is doing its job correctly, no. That is the entire product premise. The whole anti-slop pipeline exists because the answer to “can people tell” for generic AI writing tools is almost always yes: the em dashes, the “In today's landscape” openers, the bullet lists where prose was asked for, the word “delve,” the hedging, the lifeless rhythm. Döppelscript strips those character-level tells before you see the post, and the voice-profile layer fights the bigger tell, which is that every generic-AI post sounds like every other generic-AI post.

The honest caveat: very sophisticated readers who know what to look for in AI-generated text may still flag suspicious patterns in some generations. Nothing is foolproof. The goal isn't to trick AI detectors; the goal is to produce writing that sounds like the person it claims to represent. If you want to avoid the “did AI write this” question entirely, your best move is to use Döppelscript's output as a draft and edit it with your own hands before publishing. That's what the editor step is for.

No. LinkedIn's Terms of Service prohibit automated posting, scraping, and behavior that looks like a bot. Döppelscript does none of those things. It helps you draft content that you then manually review, edit, and post yourself through LinkedIn's normal interface. Tools that help a human write are explicitly allowed. Tools that post on the human's behalf, or scrape other people's posts at scale, are not. Döppelscript is the first kind, not the second.

When Döppelscript imports your existing posts to build a voice profile, that import happens through LinkedIn's own API with your explicit OAuth consent, which is exactly the interaction LinkedIn designed the API for. No scraping, no bypassing rate limits, no hidden browser automation.

You are. Every Döppelscript generation goes through an editor step where you review the draft before it's finalized, because the product is built on the premise that you are in the loop. If a post contains a factual error, or it phrases something in a way that causes harm, the responsibility lies with the person who hit “Publish” on LinkedIn, which is you. This is the same as Grammarly, spellcheck, or a human editor: the tool can help, but the accountability stays with the author.

That said, Döppelscript has a responsibility too, and it's to do the quality work honestly. The anti-slop rules exist so that the default output isn't embarrassingly generic. The inline judge (a second LLM that scores every post before you see it) exists so that flat or hedged output is flagged and offered a rewrite. If a generation comes out badly enough that you believe the tool failed, email support@doppelscript.com with the generation ID and I'll look at it.

You shouldn't, yet. Trust isn't something a stranger on the internet can claim. It's something earned over time by acting honestly in situations where acting dishonestly would be easier. What Döppelscript can do right now is tell you exactly how the data flow works (sections above), not sell your writing to anyone, not train any models on it, and give you a one-click delete that actually deletes. That's the contract. The trust has to come from seeing the contract held over time.

The next question below has more on why the business structure makes it possible to keep that contract without needing to betray it later.

Don't. Voice profiles must be built from your own writing. Uploading somebody else's posts to mimic them is a terms-of-service violation, and accounts doing it will be closed without refund.

The one grey area is legitimate ghostwriting for pay, where a writer is contracted to write in a client's voice with the client's knowledge and consent. That is a real profession, it predates AI by centuries, and Döppelscript doesn't have a problem with it in principle. But the voice profile in that case still needs to be built by the person who will be doing the writing, with content they have the right to use, and the output still needs to be approved by the person whose name will go on the post. Anything sneakier than that is impersonation, and impersonation is out.

If you believe an account is doing this, report it at /report-abuse. Every report is reviewed by a human within one business day. You do not need a Döppelscript account to submit a report.

Döppelscript was built by Ian Greenough, a software engineer who got tired of LinkedIn sounding like it was written by a committee of chatbots. It is a one-person product right now. Every line of code, every design decision, every piece of copy on this website, and every position this document takes is signed by one human. When the FAQ says “we,” that's a convenience; in practice it's “I.”

You can find me on GitHub at github.com/nullvoidundefined. Döppelscript's code is not open source, but the account is public so you can verify that a real person with real commit history is behind this.

Bootstrapped. No venture capital, no angel investors, no growth-at-all-costs pressure, no “we need to ship user tracking so our metrics look good for the next round” conversations. Revenue comes from you paying for credits. That is the entire business model. The advantage of bootstrapping is that Döppelscript will never need to sell user data, run ads, or chase vanity metrics to look good for someone else. The constraint is that Döppelscript grows at the pace paying customers let it grow, which is slower than the alternative but sustainable without needing to betray the people who made it work.

Email support@doppelscript.com. The human you reach is Ian. Responses usually come within the same business day. For feature requests specifically, feature-requests@doppelscript.com goes to the same inbox but gets triaged into a different folder. For press, partnerships, or anything that isn't a support question, hello@doppelscript.com. All three addresses are monitored by one person, which is fast when things are quiet and slower when things are busy, and that tradeoff is honest.