emptygpt
Pricing

Word-credit pricing.
No surprises.

You pay per word, not per click. Both tiers cover the actual LLM cost we incur and leave a small margin to keep the lights on. The monthly word budget refills on a rolling 30-day window — no calendar surprises, no auto-charges for going over.

Standard
$0one-time

Sign in and try emptygpt on real drafts. A free onboarding allowance — once you've used it up, upgrade Premium to continue.

  • 1,000 Standard words (single-pass)
  • 300 Premium words to sample the best-of-3 picker
  • Citations, terms, and numbers preserved exactly
  • Auto-translate-safe + B&W minimal UI
  • 8,000 character cap per rewrite
  • .docx + .txt upload, .docx export
Sign in to use
Premium
$14.99/ month

For grad students and researchers writing dense academic prose. Best-of-3 picker selects the most human-sounding variant.

  • 20,000 words / month
  • Best-of-3 picker (three independent rewrites, top model)
  • More structural variety on long inputs
  • Section-aware rewriting (Methods, Results, Discussion)
  • Same hard preservation rules as Standard
What you're paying for
The actual rewrite

Every call runs through our humanizing pipeline — a stack of language-model passes plus deterministic post-processing that we keep refining as we learn. Premium fans out into several independent rewrites and returns the one that reads most like a human wrote it.

Not detector bypass

We do not promise outcomes on Turnitin, Originality.ai, GPTZero, or any other AI-detection platform. Detection algorithms differ, change without notice, and operate outside our control. We optimize for human readability.

What we don't promise
  • emptygpt is for legitimate humanizing of your own AI-assisted drafts — not for academic dishonesty, plagiarism, or evading institutional AI-disclosure policies. See Terms §4.
  • We do not guarantee any outcome on Turnitin, Originality.ai, Copyleaks, or any other AI-detection service. These platforms change without notice and are not in our optimization loop.
  • Single-shot detector scores are noisy — the same input can score very differently on consecutive calls of the same detector. Treat them as rough signals, not predictions.
  • You are responsible for complying with your institution's AI-disclosure policies. emptygpt is an editing aid, not a replacement for original work or for honest disclosure when your institution requires it.