AI Use Policy

myCFPCoach's core feature — coaching a talk idea into proposal drafts — is powered by a large language model (currently one fixed Anthropic Claude model; we don't let requests pick their own model). This page explains what that means in plain terms.

1. What the AI Actually Does

When you submit an idea, we send it — along with your target conference, format, and any optional document, role, or CFP link you provide — to Anthropic's API. The model uses a web-search tool to research the conference, then drafts three coached proposals: a title, abstract, full description, a fit score, and coaching notes. Nothing is generated without you actively submitting a request, and every request counts against a rate limit designed to keep the tool usable and free.

2. AI Output Is a Draft — Not a Guarantee, Not Advice

Treat every coached proposal as a first draft written by a very well-read intern, not a finished, fact-checked submission. Specifically:

• The fit score and fit band are a directional estimate, not a probability of acceptance — real conference acceptance rates are typically 8–15%, and no model can predict a specific committee's decision.
• Specific numbers, incident details, or claims the model writes may be plausible-sounding but invented if you didn't provide them yourself. Never submit a specific stat, quote, or incident you can't personally verify.
• The model's read of "what this conference accepts" comes from web search (or, if you provided one, a Sessionize feed of real accepted talks) — it can be incomplete, outdated, or wrong.
• This is coaching, not professional, legal, medical, or financial advice of any kind.

You are responsible for reviewing, fact-checking, and editing any coached output before submitting it to a real conference, employer, or anyone else, under your own name.

3. What We Send, and What We Don't

We send the AI provider only what's needed to coach your proposal: your idea text or uploaded document, target conference and format, and optional context (role, experience, tone, language, and Sessionize-sourced accepted-talk titles). We do notsend your password (we don't have one — you sign in via Google), payment details (we don't collect any), or your full account history.

4. How the AI Provider Handles That Data

Per Anthropic's commercial API terms as of this writing, prompts and outputs sent through the API are not used to train Anthropic's foundation models by default. Terms can change — see Anthropic's current commercial terms for the authoritative, up-to-date statement rather than relying solely on this page.

5. Abuse Prevention — Why There Are Limits

AI generation costs real money per request, and unmetered access invites abuse (bots, scraping, resale). That's why: anonymous and signed-in users each get a fixed number of generations per hour and per day, everyone shares one fixed, non-configurable model rather than picking their own, and we log enough (IP or account, timestamp, outcome) to enforce those limits — never the content of a blocked or rate-limited request beyond what's needed for that.

6. No Guarantee of Originality

Language models can occasionally produce text that resembles existing material. We don't warrant that any coached output is free of similarity to other work. You're responsible for ensuring anything you publish or submit doesn't infringe someone else's rights — when in doubt, rewrite it in your own words.

7. Changes to This Policy

As the AI features change (new models, new data sources, new limits) we'll update this page and the "Last updated" date above.

8. Contact

Questions about how the AI works or what data it sees? Email support@mycfpcoach.app.