Frequently Asked Questions

CorPrompt is an AI-powered service that transforms your rough ideas into professional, production-ready prompts, instantly. We call this process Prompt Refinement. Whether you're using ChatGPT, Claude, Gemini, or any language model, CorPrompt ensures your prompts are structured, clear, and optimized for consistent results every time.

Prompts are instructions you give to AI tools like ChatGPT, Claude, or Gemini. They're questions, commands, or descriptions that tell the AI exactly what you want it to do. The better your prompt, the better your result.

When you ask an AI to do something without a well-structured prompt, you usually get an okay result. Improving it typically requires multiple rounds of back-and-forth adjustments—and even then, you can't easily repeat the same quality next time.

CorPrompt solves this by refining your rough idea into a polished, structured prompt that works flawlessly across all AI models. Once refined, you get excellent, consistent results every single time.

You own everything you create. We only use your prompts to provide the service. We never claim ownership of your content, and your prompts are never used to train AI models.

It's easy:

Usually 1-2 minutes, depending on how complex your prompt is and how fast our AI services are running at that moment.

Elements are helpful snippets of text you can save and reuse—things like your brand guidelines, writing instructions, or anything else you want included in your prompts every time. Think of them as "context" that makes your prompts better. Elements can be created manually or generated by AI.

Examples are sample input-and-output combinations. For instance, if you want to show what good customer service looks like versus bad, you can save examples and include them in your prompts. Providing AI with clear examples can improve results substantially. Examples can be created manually or generated by AI.

Validation is the final quality check in our refinement process. Our AI agents review the refined prompt and decide if it needs more work or if it's ready to go. It usually doesn't take long, and it ensures you get a really great prompt.

Click on "My Prompts" to see everything you've created. From there, you can:

If you're not fully satisfied with your prompt, you can provide several rounds of feedback to further improve your prompt for free. The number of rounds is determined by your subscription, and is noted on the details page of your prompt. Your feedback should be direct and instructional. You can ask our AI agents to remove, modify, or add content to the prompt.

There’s a "Version History" card on the sidebar of the prompt details page. You can view and restore previous versions of your prompt from there.

Go to the prompt's details page and click "Copy." The entire prompt is copied to your clipboard. Now, just paste it into ChatGPT, Claude, or whatever AI tool you're using.

On the prompt details page, click "Download" and choose your format: plain text, Markdown, PDF, or JSON. Download any prompt anytime.

Yes, two ways:

To stop sharing a prompt, click "Manage Share Link" and revoke it, or unpublish if it's public.

When you create an Element or Example, you can check "Make public (share with community)" to let other users see and use them. To stop sharing, edit the item and uncheck that box.

This feature uses AI to write well-crafted Elements or Examples for you. Your subscription includes a generous number of generations per month (including both Elements and Examples combined).

Yes. Go to "My Prompts," check the boxes next to the prompts you want to delete, and click "Delete." Once deleted, they're gone for good, so make sure you download a copy before you do.

Use folders! Go to "My Prompts" and click "New" to create a folder. Then you can move prompts into folders to keep things organized. You can also manage folders from each prompt's details page. For quick categorization, you can add tags from the prompt details page as well.

Our system uses a collection of advanced AI agents designed to align your prompt with LLM (Large Language Model) best practices. Sometimes, a plain English request needs to be restructured significantly or expanded to ensure a computer understands the specific intent, tone, and constraints. If the result feels too different, check that your original request was specific enough; if the AI has to guess your intent, it may drift from your vision.

The "Add Guidance" tool is a powerful, one-shot instruction mechanism. It interprets your commands literally and rapidly. If you say "remove the intro," it will cut it immediately without attempting to preserve context unless you explicitly tell it to. We recommend being very specific with your guidance (e.g., "Rewrite the second paragraph only," rather than "Shorten this").

Our refinement process is powered by "Model Sets"—groups of AI agents that work together to polish your prompt. Depending on your subscription tier, you are accessing different levels of reasoning power. Higher tiers generally utilize more complex Model Sets capable of handling nuanced logic and subtler instructions.

Check to ensure you haven't uploaded conflicting Elements. If you upload a Brand Voice guide that demands "professionalism" and a Tone guide that demands "slang," the refinement agents may get confused or default to a neutral path. Additionally, ensure you select only the Elements relevant to the specific prompt you are building; including too much irrelevant data can dilute the AI's focus.

This is often due to the model selected for the test run. While your prompt is high-quality, not all AI models are created equal. If you choose a smaller, older, or open-source model to run a complex prompt, it may not have the "brainpower" to follow the instructions. Try running the prompt with a more capable model (like a GPT-5 or Claude-class model) available in your tier to see the true potential of the prompt.

This is a phenomenon known as "overfitting." If your provided Examples are too specific or if there are too many of them, the model might think you want it to repeat those distinct patterns rather than follow the logic behind them. Try reducing the number of Examples or ensuring your Examples cover a wider variety of scenarios to show the model the pattern, not just the answer.

Those are placeholders. If your prompt includes variables (like {{Customer_Name}} or {{Article_Topic}})—either because you added them or the refinement process identified a need for them—you must provide the actual data for those placeholders before running the prompt through our system. When using your prompt with other system (e.g., ChatGPT), be sure to manually replace those placeholders with the required context. Without that input, the model will either leave the bracket blank, parrot it back to you, or hallucinate random data.

Our platform allows you to test prompts on specific models. If you test it here on a high-end model (like GPT-5), but then copy-paste the prompt into a less capable or different proprietary system (like a basic chatbot or an older open-source model), the results will vary. Always ensure the "destination" model has similar capabilities to the one you used for testing.

Prompt refinement optimizes structure and clarity, but it cannot cure an AI's inherent tendency to hallucinate if the source material is missing. If you ask for a summary of a document but don't paste the document into the prompt (or provide it as an Element), the AI will try to be helpful by inventing plausible-sounding content. Always verify that the model has the source information it needs to answer the question.

As the name implies, "Experimental" sets use cutting-edge or beta-testing logic that hasn't been fully standardized. While this can yield creative or breakthrough results for complex problems, it can also behave unpredictably or misinterpret simple instructions. For critical tasks, we recommend sticking to the Standard or Premium model sets.

Note: Only the Pro subscription currently offers multiple model sets. Free and Casual use the CorPrompt Standard model set, which already utilizes some of the latest and most advanced AI model available.

Your subscription tier comes with a generous (but limited) number of guidance rounds you can apply per prompt. However, we advise against applying guidance too many times. Each guidance step is a literal, one-shot modification, and over-refining can eventually degrade the prompt's core logic or make it overly verbose. If you find yourself adding guidance more than once or twice, it is usually more effective to go back and edit your original plain English request, then restart the refinement process.

Some variation is normal in Generative AI (it is probabilistic, not deterministic). However, if the variance is extreme, check the Elements and Examples you included in your original prompt. For example, if your Tone or Brand Element is vague, the model has too much "wiggle room." You may need to use the "Add Guidance" feature to lock down the format strictly (e.g., add guidance: "Ensure the output is always exactly three bullet points").

Credits are tokens you use to run refinements. Each refinement costs one (1) credit. Applying guidance to create new versions are free, so is sharing, publishing, AI generation*, and organizing your prompts. They come with your subscription.

*AI generations are capped, but every subscription comes with a generous monthly limit.

Each month you get a fresh set of credits that expire at the end of that month (they don't roll over). If you run out before the month ends, you can buy a top-up bundle—those last 30 days from when you buy them, regardless of when your regular subscription renews.

We use credits in the order you buy them (first-in, first-out).

Upgrade: Click "Billing" and subscribe to a higher plan. You're charged immediately and get the new credits right away. Your renewal date updates to today.

Downgrade: Subscribe to a lower plan anytime. The change takes effect at the start of your next billing period.

No problem. Go to "Billing" and scroll to "Top-up Credits." You can buy extra credits anytime. They last 30 days.

Yes. Go to "Billing" and cancel your subscription. You keep any remaining credits until the end of your current billing period (or until any top-ups expire). You can reactivate any time before that period ends. And no, there are no lock-in contracts—cancel whenever you want.

We don't offer refunds on any purchases—subscriptions or top-ups. This is because our operational costs vary based on how much work our AI agents need to do. That's why we give you free trial credits: you can test the service before you pay. If you're unsure, use your free credits first.

If your account hasn't been accessed for 12 straight months, we'll automatically delete it to respect your privacy. We'll send you a warning email at least 30 days before this happens. Just log in before the deletion date to prevent it.

We only collect what we need to run the service: your name, email, the prompts you create, and basic usage information. We take security seriously:

You can delete any of your content anytime, or delete your entire account. When you delete, your data is removed from our active system and cleared from backups within 90 days.

Only you keep your prompts. We store them to run the service, but only while your account exists. When you delete a prompt or your account, it goes away (within 90 days for backups). Your prompts are never trained on by any AI provider we use.

This is important: we've enabled Zero Data Retention (ZDR) with our LLM provider, OpenRouter. This means your prompts are never stored or used to train the AI models—they're just processed and discarded. Your data stays yours.

Yes. Payment processing is handled by Stripe, a major payment processor. We don't store your full credit card number—Stripe does that. Your charge will show up as "MADNESS*CorPrompt" on your statement.

MADNESS (FZE) is the legal entity that owns and operates CorPrompt.com, which is why our company name shows up on your credit card statement when subscribing or purchasing top-up credits.

Absolutely. You can download your prompts in plain text, Markdown, PDF, or JSON anytime. You can also delete your entire account whenever you want through the Billing section.

Yes. If you're in the EU, UK, or Switzerland, you have GDPR rights including the right to access, correct, delete, and port your data. If you're in California, you have CCPA/CPRA rights to know what we collect, delete it, correct it, and opt out of selling (which we don't do anyway). Contact [email protected] to exercise these rights.

Your email address can't be changed once you verify it—this prevents fraud and abuse. If you want to use a different email, you'll need to delete your account and create a new one with the new address.

Please don't use the service to create content that is:

AI refinement is creative and results can vary. We can't guarantee every prompt will be perfect on the first try. That's why refinement includes a validation step and why you can restart or apply guidance to create new versions. Always review the refined prompt before using it.

We aim to keep things running smoothly, but like any online service, we may have scheduled or unscheduled maintenance, technical issues, or unexpected problems. We try to minimize downtime, but we can't guarantee 100% uptime.

Absolutely! Once you download or copy a prompt, it's yours to use anywhere. You can paste it into ChatGPT, Claude, or any other LLM. What happens with it after that is up to you—those tools' terms of service apply.

Zero Data Retention is a critical privacy feature. We've enabled ZDR with our LLM provider (OpenRouter). This means:

Yes. Your payment information is processed by Stripe, an industry-leading secure payment processor. We never see or store your full credit card details.

To change your cookie preferences, click the Cookie Settings link at the bottom of the CorPrompt homepage, or visit the Settings section of your account (if logged in). 

For registered users, your cookie preferences are saved to your account. For non-logged-in users, preferences are stored in a cookie. If you clear your browser cache or switch browsers, you'll need to set your cookie preferences again.

For privacy questions:
[email protected]

For general questions:
[email protected]

We aim to respond as soon as possible.