> ## Documentation Index
> Fetch the complete documentation index at: https://playbook.pharmatools.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Freelancer AI-Use Statement

> A practice-level AI-use statement for freelance and independent medical writers: what you use AI for, what you don't, and how you verify it. For your website, onboarding pack or contracts.

The [disclosure snippets](https://playbook.pharmatools.ai/templates/disclosure-language.md) cover single deliverables: one manuscript, one poster, one CSR section. A freelancer also needs a standing answer to the question clients now ask before the work starts: *how do you use AI?* This page gives you a template for that answer: a practice-level statement setting out what you use AI for, what you never use it for, how you handle client data, and who is accountable for the output.

It does not replace per-deliverable disclosure. A statement on your website says how you work in general. The disclosure in a manuscript says what happened on that manuscript. You need both, and they must agree.

> [!NOTE]
>
> **Why this is now expected, not optional.** The [AMWA Position Statement on the Use of AI in Medical Communication](https://cdn.ymaws.com/www.amwa.org/resource/resmgr/statements/amwa_ai_position_statement.pdf) (approved October 2026) states that disclosure obligations may extend to clients, employers, collaborators, publishers, journals and regulators, and treats failing to disclose AI use where it is expected as inconsistent with professional standards. It also tells writers to seek explicit client guidance before putting sensitive material through external AI systems. A written statement is the simplest way to show a client you meet both expectations before they have to ask.

> [!WARNING]
>
> **The client's policy wins.** Your statement describes your default practice. Where a client's SOP, contract or master services agreement is stricter (no external AI tools, approved tools only, AI use logged per project), that governs the work. Say so in the statement, and mean it.

## How to use this page

* **Write it from what you actually do.** A statement that says "I never use AI for first drafts" is a commitment. If it stops being true, change the statement before you change your practice.
* **Edit every bracketed variable.** Delete the lines that don't apply rather than leaving them vague.
* **Name the limits, not just the uses.** Clients want to know where the line is more than they want a list of tasks. The "What I don't use AI for" section does most of the work.
* **Keep client names and projects out of it.** The statement is about your practice, not your portfolio. Naming the brands you used AI on can breach confidentiality just as easily as pasting their data into a tool.
* **Date it and review it.** Tools, journal policies and your own habits all change. Put a "last reviewed" date on the statement and revisit it at least every six months, or whenever a major journal or professional body updates its position.
* **Link it to your audit trail.** The statement says you document AI use. The [AI Audit-Trail Log](https://playbook.pharmatools.ai/templates/ai-audit-trail-log.md) is how you do it.

***

## Public, private, or both?

There is no single right answer, and the choice depends on who you write for.

| Where it lives | Good for | Watch out for |
| - | - | - |
| **Website (public)** | Showing prospective clients and agencies how you work before the first call; signalling that you have thought about it | Anything you publish is a public commitment. Keep it to principles and limits, not tool-by-tool detail that dates quickly |
| **Onboarding pack or proposal (private)** | Fuller detail: named tools, data-handling settings, how you log use. Tailored to the client's sector | Must be kept in sync with the public version. Two statements that disagree are worse than none |
| **Contract or MSA schedule** | Making the commitments enforceable on both sides, including the client's own restrictions | Have it reviewed. A clause you can't comply with on every project is a liability, not a reassurance |

A workable default: a short public version on your website (below), the full statement in your onboarding pack, and a one-paragraph clause in your terms that defers to the client's policy.

***

## Full statement (onboarding pack or website page)

> **How I use AI in my medical writing work**
>
> *Last reviewed: \[date]*
>
> I use AI tools in parts of my work. This statement sets out where I use them, where I don't, and how I make sure the work you receive is accurate and that I am accountable for it. Where your organisation's AI policy or SOP is stricter than this, your policy applies.
>
> **What I use AI for**
>
> * \[Literature searching and screening, e.g., identifying candidate papers, which I then read in full]
> * \[Summarising source documents I have provided, as a starting point for my own reading]
> * \[Copy-editing and proofreading of text I have written: clarity, consistency, house style]
> * \[Checking consistency, e.g., numbers across a document, abbreviations, reference-to-claim matching]
> * \[Structuring and outlining before I draft]
> * \[Other uses: translation first pass, slide layout suggestions, plain-language rewording]
>
> **What I don't use AI for**
>
> * \[Writing first drafts of manuscripts, patient-facing materials or \[other deliverable types]. These are written by me]
> * Generating, formatting or selecting references. Reference lists are built in \[reference manager]
> * Creating or altering clinical images, data or figures
> * Interpreting results or drawing clinical conclusions
> * \[Opinion pieces, editorials and Letters to the Editor]
> * \[Peer review, or any task involving a manuscript I have received in confidence]
>
> **Tools and your data**
>
> * I use \[Tool name(s), e.g., Claude (Anthropic)], on \[plan type, e.g., a business plan with model training on my data switched off].
> * I do not enter unpublished data, patient-level information or confidential commercial material into any AI tool unless you have approved that tool for that material.
> * \[Where a client provides an approved or in-house tool, I use that instead.]
>
> **Verification and accountability**
>
> * Every factual claim, number and reference in work I deliver has been checked by me against the primary source, whether or not AI was involved in producing it.
> * AI does not make editorial or scientific decisions on my projects. I do, and I take responsibility for the accuracy and integrity of everything I deliver.
> * I keep a record of AI use on each project: what tool, when, and for which part of the work. It is available to you on request.
>
> **Disclosure in your deliverables**
>
> * Where AI use needs to be declared (in a manuscript, poster, regulatory document or promotional piece), I will draft the disclosure in line with the target journal's or code's requirements and flag it for the author group or your review team.
> * I follow the [ICMJE Recommendations](https://www.icmje.org/recommendations/), [GPP 2022](https://doi.org/10.7326/M22-1460) and the [AMWA](https://cdn.ymaws.com/www.amwa.org/resource/resmgr/statements/amwa_ai_position_statement.pdf) and [ISMPP](https://doi.org/10.1080/03007995.2023.2273139) positions on AI use.
>
> **Questions**
> If you'd like more detail on any of this, or need me to work under different restrictions for your project, \[contact details].

***

## Short version (website snippet)

*For a services page, footer or "How I work" section. Roughly 80 words; link to the full statement if you publish one.*

> **AI use.** I use AI tools for \[research support, summarising sources and copy-editing]. I don't use them to \[write first drafts, generate references, or create clinical images or data]. Every claim, number and reference I deliver is checked by me against its primary source, and I take full responsibility for the work. I never put confidential client material into an AI tool you haven't approved, and where your AI policy is stricter than mine, yours applies. \[Read my full AI-use statement →]

***

## Contract or terms clause

*For your standard terms or a schedule to an MSA. Have it reviewed before you rely on it.*

> **Use of AI tools.** The Contractor may use AI tools in the course of the Services as described in the Contractor's AI-Use Statement dated \[date], save that (a) where the Client's written policies, SOPs or instructions restrict or prohibit the use of AI tools, the Contractor will comply with those restrictions; (b) the Contractor will not input Client Confidential Information into any AI tool not approved in writing by the Client for that purpose; and (c) the Contractor remains responsible for the accuracy and integrity of all deliverables regardless of any AI tool used. The Contractor will maintain a record of AI use on the project and provide it to the Client on request.

***

## How the guidelines map onto the statement

None of the core publication guidelines was written with a freelancer's website in mind, so the statement has to translate them.

| Guideline | What it says | Where it lands in your statement |
| - | - | - |
| **[ICMJE Recommendations](https://www.icmje.org/recommendations/)** | AI cannot be an author; authors disclose AI use and remain responsible for the content | *Disclosure in your deliverables*: you draft the declaration, the authors own it |
| **[GPP 2022](https://doi.org/10.7326/M22-1460)** | Published in August 2022, before generative AI tools were widely available, so it does not address them directly. Its principles of transparency and acknowledging writing support are the ones that carry over | *Verification and accountability*: the same transparency you apply to acknowledging your own role applies to the tools you used |
| **[AMWA position statement on AI](https://cdn.ymaws.com/www.amwa.org/resource/resmgr/statements/amwa_ai_position_statement.pdf)** (Oct 2026) | Disclosure may be owed to clients as well as journals; get explicit client guidance before using external AI on sensitive material; organisations should have written AI policies | The whole statement, and especially *Tools and your data* |
| **[ISMPP position statement](https://doi.org/10.1080/03007995.2023.2273139)** (2023) and [enhanced guidance](https://www.ismpp.org/artificial-intelligence-ai-) (2025) | Human oversight, appropriate disclosure, confidentiality | *What I don't use AI for* and *Tools and your data* |
| **Journal policies** (see [Declaring AI Use](https://playbook.pharmatools.ai/principles/declaring-ai-use.md)) | Some uses are now prohibited outright, not just disclosable | *What I don't use AI for*: if a use is barred by a journal you write for, don't list it as something you do |

***

## Notes on adapting this template

* **Limits age better than tool lists.** "I use Claude Opus 4.7" will be out of date within months. "I don't use AI to generate references" stays true. Put tool names in the private version, principles in the public one.
* **Scope by deliverable type if your practice differs.** Many writers use AI more freely on internal briefing documents than on manuscripts or patient materials. Say so explicitly rather than writing the statement to the most permissive case.
* **Don't promise what you can't evidence.** "Every claim is verified" is only credible if your audit trail shows it. If a client asks for the log, you should be able to produce it.
* **Agencies are clients too.** If you subcontract through an agency, their policy, and their client's policy above it, may both apply. Ask which governs before the project starts.
* **Practise what you publish.** This playbook applies the same standard to itself: see [How AI was used to write this](https://playbook.pharmatools.ai/about.md#how-ai-was-used-to-write-this).

***

## Related

* [Disclosure Language Template](https://playbook.pharmatools.ai/templates/disclosure-language.md): per-deliverable wording for manuscripts, posters, regulatory and promotional material
* [Declaring AI Use](https://playbook.pharmatools.ai/principles/declaring-ai-use.md): journal, regulator and sponsor expectations, including what is now prohibited
* [AI Audit-Trail Log](https://playbook.pharmatools.ai/templates/ai-audit-trail-log.md): the record that backs up the statement's commitments
* [Review and Accountability](https://playbook.pharmatools.ai/principles/review-and-accountability.md): why the named human stays accountable

***

*Last reviewed: 18 November 2026 · 6 min read*


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.