Do I have to tell a client I used ChatGPT?

If the contract or brief bans AI, yes—or do not use it. If the brief is silent, default to honesty on anything that invents claims or replaces your judgment. A light grammar or humanizer pass on text you wrote is different from shipping a raw model draft.

Is using a humanizer the same as AI ghostwriting?

Usually no. Ghostwriting with a model means the model wrote the substance. A humanizer rewrites rhythm after your facts and outline exist. Still disclose if the client asks about any AI, or if the SOW requires it.

What should I put in a freelance contract about AI?

Allowed uses (outline, grammar, humanize), banned uses (invented quotes, fake case studies), disclosure trigger, and which detector mode you will pre-check if they care. Keep it short.

Will Human Writes disclose for me?

No. It edits prose. You own the client conversation and the invoice.

Disclose AI Writing to Clients (Freelance and Agency)

Clients do not buy your prompt library. They buy work they can put their name on. Disclosure is about trust and liability, not about sounding pure.

Academic ethics live here: Checklist for Ethical AI Use in Academia. Who a humanizer is for: AI Humanizer Use Cases. This post is the client deliverable version: freelance and agency.

When silence is a problem

Situation	Disclose?
Brief or MSA bans AI	Do not use it, or get a written exception
You invent testimonials, metrics, or citations with a model	Always a problem; delete, then tell them if it shipped
You outline in ChatGPT, write the argument yourself, humanize once	Tell them if they ask, or if the SOW requires any AI note
Grammar pass on your own draft	Rarely material; still answer honestly if asked
Regulated claims (health, finance, legal)	Counsel + explicit approval; tools are not a defense

If you would be embarrassed to explain the workflow on a Zoom, fix the workflow before you fix the wording.

Humanizer as QA vs ghostwriting

Role	What happened	How to describe it
Ghostwrite	Model produced the substance	“AI-drafted; I edited” (if allowed)
Assist	Model helped outline; you wrote claims	“AI-assisted outline; human draft”
QA / humanize	Your draft; rhythm pass	“Edited with an AI writing tool for clarity”

Human Writes sits in the QA row when you already locked facts. It does not invent a case study. If you used it to polish empty ChatGPT marketing, you are closer to ghostwrite. Be honest about which row you are in.

Sample language (edit with counsel)

Allowed in SOW:

Contractor may use AI tools for outlining, grammar, and clarity editing. Contractor will not invent facts, quotes, or metrics. Final claims remain Contractor’s responsibility. On request, Contractor will disclose which tools touched the deliverable.

Banned:

No generative AI may be used to draft client-facing copy without prior written approval.

Delivery note (when they want visibility):

Drafted by [Name]. Outline assistance: ChatGPT on [date]. Clarity pass: Human Writes on [date]. All product claims taken from your brief dated [date].

Keep a one-line log per job. You will need it when an account manager changes.

Detectors and clients

Some buyers add “must pass Originality.” That is a QA SLA, not a substitute for disclosure. See detectors for agencies. A green score does not fix a fake customer quote.

What not to do

Do not hide AI after they asked a direct question.
Do not put their unreleased strategy into a consumer chatbot.
Do not promise undetectable delivery as a service line.
Do not treat disclosure as a confession of incompetence. Treat it as scope control.

Bottom line

Disclose when the contract requires it, when substance came from a model, or when trust would break if they learned later. Human Writes is a clarity pass after the work is yours. It does not send the email for you.

Lock the claims, run one pass on Human Writes if the account allows it, and keep the log.