AI use policy template for writing teams and agencies

An AI use policy template in copyable blocks: what is allowed and banned, human review, fact-checking, sources, client data, disclosure, logging and responsibilities, based on Google’s guidance and on U.S. and EU official sources. It is not legal advice.
Written with AI, reviewed by humans

In this article

See how your brand shows up in AI answers

Book a demo and find out what to fix first, with a GEO specialist.

Book a demo

Reading time: 11 minutes

An AI use policy template tells your team what it can do with AI, who reviews the work, which data stays out and when to tell readers. Copy its blocks into an AI use policy for writing and adapt them. However, the template is not legal advice.

In short

  • An AI use policy for writing is short. It answers six questions: what is allowed, who reviews, how to check facts, which data goes in, when to tell readers and who is responsible.
  • Google does not ban AI. But it asks you to fact-check and review all AI-generated content, and it treats scaled content abuse as spam.
  • According to the official pages, data use in the tools depends on the plan. So the policy lists the approved tools.
  • The template also draws on U.S. and EU sources on copyright, fake reviews and AI labels. It is an editorial starting point, so a lawyer should review the final version.

What is an AI use policy for writing teams?

An AI use policy for writing makes the whole team use the tool in the same way. It sets clear rules on what is allowed and who reviews. So AI use no longer depends on each person’s judgment. It also gives clients a standard they understand.

Without a policy, each writer decides alone which tool to use, which material to paste and whether to tell the client. As a result, errors tend to show up in three places. They are the fact nobody checked, the client data that went into a personal account and the text that shipped without review. A short policy prevents all three.

However, the policy does not replace the process. It sets the rules, and the AI content workflow shows how to follow them, with roles, fact-checking checklists and a log template. So in this guide, the clauses point to that workflow instead of repeating it.

Note: this template is an editorial starting point, not legal advice. After all, client contracts, copyright, privacy and advertising rules depend on your case and on where you publish. So ask a lawyer to review the final version before you adopt it.

To use the AI use policy template, copy the blocks below into a single document. Then replace the fields in brackets, such as [company] and [owner]. Finally, delete whatever does not apply to your team. First of all, the first block holds the header and the purpose.

AI USE POLICY FOR WRITING
[Company] | Version 1.0 | [date] | Owner: [name]
Legal review: [pending / done by ___ on ___]

1. PURPOSE AND SCOPE
- This policy guides the use of generative AI in
  creating content for [company] and for the clients
  we serve.
- It covers staff, freelancers and vendors who
  produce text, image, audio or video for us.
- AI supports the work. Whoever signs the content
  is responsible for it.
AI use policy template: illustration of a document with a checklist and a magnifying glass

What does Google say about AI-generated content?

Google does not ban the use of AI. In its guidance on generative AI, it says the tool can help you research a topic and add structure to original content. But generating many pages with no value for readers may violate the spam policy on scaled content abuse.

On the other hand, the same page asks for human review. It reminds us that generative models do not retrieve facts. Instead, they predict a likely sequence of words, so they can make mistakes. So the conclusion is direct:

“It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing.”

Google Search Central, guidance on generative AI content

This review also applies to the title, the meta description, structured data and image alt text. In addition, Google suggests sharing how the content was created, because that gives readers context.

In its FAQ on AI-generated content, the company says appropriate use of AI is not against its guidelines. The limit is using it mainly to manipulate rankings. If your client asks whether Google penalizes AI-made content, the policy and the studies are in that guide.

So the table shows how each piece of guidance becomes a clause in this AI use policy template.

What Google saysWhere it fits in the policy
AI helps with research and structureSection 2, allowed uses
Many pages with no value violate the spam policySection 4, banned uses
Fact-check and review everything before publishingSections 5 and 6, review and sources
Saying how the content was created helps readersSection 8, disclosure

What is allowed and banned in an AI use policy?

An AI use policy splits uses into three lists. The first holds what the team does without asking. The second holds what needs the editor’s approval, and the third holds what is banned. In general, what is allowed helps with research and structure. What is banned misleads the reader, the client or Google.

The lists in this AI use policy template are our suggestion. Adjust them to the type of client and the risk of each topic. For example, an agency that serves clinics can send every health text to the editor.

2. ALLOWED USES (with human review)
- Research the topic and list the reader's questions.
- Suggest structure, headlines and openings.
- Generate a draft from the brief and the approved
  sources.
- Summarize material the team has the right to use.
- Review grammar, clarity and tone.
- Translate our own text, with review by someone
  who speaks the language.

3. USES THAT NEED EDITOR APPROVAL
- Topics on health, money, law and safety.
- Text with a number, price, date or quote.
- Image, audio or video generated by AI.
- Text signed by a person from the company or from
  the client.

4. BANNED USES
- Publishing AI text without human review.
- Generating many pages just to fill search results.
- Inventing an author, profile, testimonial, study
  or quote.
- Rewriting third-party text to hide where it
  came from.
- Faking experience the team did not have.
- Sending banned data (section 7) to any tool.

Two of these items come straight from Google’s documentation. The spam policies treat the use of generative AI to create many pages with no value for users as scaled content abuse. They also treat as scraping the practice of copying content from other sites, making small changes and republishing it.

If the team plans to automate publishing, the guide on automated blogs with AI shows the controls.

How does an AI policy handle review, facts and sources?

An AI policy requires a person to review all AI-made content. In addition, someone other than whoever generated the draft checks each fact at the source. Every piece of third-party information also carries its source in the text. In other words, an AI tool never counts as a source.

Google treats factual errors as a quality problem. On its helpful content page, it asks whether the content has any easily-verified factual errors. So we treat the fact-check clause as the most important one in the template.

The fact-checking steps, with the signs of invented data or quotes, are in the guide on the AI content workflow. The AI use policy for writing only sets the rule and the owner.

5. HUMAN REVIEW AND FACT-CHECKING
- All AI-made content goes through human review
  before publishing, including the title, meta
  description, structured data and image alt text.
- Whoever checks the facts is not whoever generated
  the draft.
- Every number, date, name and quote has an open
  source that someone read and logged.
- Data without a source leaves the text.
- Sensitive topics also go through a specialist.

6. CITING SOURCES
- Every piece of third-party information carries
  its source in the text, preferably with a link to
  the primary page.
- A quote in quotation marks uses only the original
  text, checked, and a translation is marked as such.
- If the source sells what it praises, the text
  says so.
- We do not paraphrase someone else's text without
  adding something of our own.
- The AI tool is never the source.

Which client data can go into AI tools?

Under the template, no client data goes into an AI tool unless the tool is on the approved list. The official pages show that what happens to the data changes with the tool and the plan. So the policy records the plan, what the page says about training and the date of the check.

ToolWhat the official page says
ChatGPT, Free, Plus and Pro plansAccording to OpenAI’s help page, the company may use the content to train its models. You can turn this off under “Improve the model for everyone”
ChatGPT Business, Enterprise and Edu, and APIAccording to the same page, by default OpenAI does not use inputs and outputs to improve its models
Claude, Free, Pro and Max plansAnthropic uses chats if the user allows it or if they are flagged for safety. Incognito chats are not used
Claude, commercial terms“Anthropic may not train models on Customer Content from Services”
Gemini Apps, personal accountHuman reviewers look at some chats. Google asks you not to enter confidential information

The table reflects what each vendor says on its own pages, and we did not test the tools. These rules change, so the check needs a date and an owner. Personal data also calls for care. The European Commission defines personal data as any information that relates to an identified or identifiable living individual. Its examples include a name, an IP address and a cookie ID. So ask your legal team or data protection officer what you can send under the laws that apply to you and your clients.

7. CLIENT DATA AND PRIVACY
- We only use tools and plans on the approved list
  (appendix A).
- We use the work account, never a personal account.
- These do not go into AI tools: personal data of
  clients and readers, passwords, contracts,
  unpublished numbers and confidential material.
- Before using client material, we check what the
  contract says about AI. If in doubt, we ask.
- We turn off the use of chats for training when
  the tool allows it.
- In tests and examples, we use fictional data.

APPENDIX A. APPROVED TOOLS (one per block)
Tool: [name]
Plan: [plan]
Training on our data: [yes/no, according to the page]
Allowed data: [public / internal / client]
Page checked on: [date] | Next: [date]
Approved by: [name]

When should you tell readers about AI use, and how do you log it?

Tell the reader when they would expect to know how the content was made. Google says AI or automation disclosures are useful for content where someone might think “How was this created?” An internal log lets you fix errors fast and shows who reviewed each text.

The byline needs care, though. According to the same FAQ, giving AI an author byline is probably not the best way to show it took part. In addition, the helpful content page asks you to avoid deceptive authorship information, such as fabricated creator profiles.

In e-commerce, however, there is a labeling rule. Google Merchant Center requires the IPTC TrainedAlgorithmicMedia metadata on AI-generated images and an “AI-generated” label on product data. If you publish an open editorial policy, the context is in the guide to E-E-A-T. The text log template is in the workflow guide, and the clause only requires that a log exists.

8. DISCLOSURE
- We tell the reader when they would expect to know
  how the content was made, with a short note at
  the end.
- Sample note: "This text used AI support for
  research and the first draft. The team checked the
  data and is responsible for the content."
- AI does not sign. The author is a real person.
- We agree with each client whether, where and how
  the note appears.
- AI-generated images and product data follow the
  label each platform requires.
- Where a law requires an AI label, we follow it.
  The owner checks this with a lawyer.

9. USE LOG
- Each text has one line in the log: address, date,
  tool, steps done with AI, who checked the facts
  and who approved.
- For work a client wants to protect by copyright,
  the log also notes what a person wrote, chose or
  changed.
- The log is kept in [location] and stored for
  [period].
- The owner reviews the log every [period].

What do U.S. and EU rules add to an AI use policy?

U.S. and EU sources add three checks to an AI use policy template: copyright, fake reviews and AI labels. None of them replaces a lawyer. However, they explain why the template keeps a record of human work, bans invented testimonials and follows any label a law requires.

Copyright. The U.S. Copyright Office says copyright does not extend to purely AI-generated material. It also excludes material where human control over the expressive elements is insufficient. Its January 2025 report adds that, with current generally available technology, prompts alone do not provide sufficient control. However, human authors keep rights in what they wrote, and in their creative selection, arrangement or changes to AI output. For registration, its 2023 guidance asks applicants to disclose AI-generated material that is more than de minimis. So clause 9 of the AI use policy template notes the human contribution.

Fake reviews. In August 2024, the Federal Trade Commission announced a final rule against fake reviews and testimonials. According to the FTC, the rule covers reviews that misrepresent that they come from someone who does not exist, such as AI-generated fake reviews. The rule lets the agency seek civil penalties against knowing violators. So the ban on invented testimonials in clause 4 is not only an editorial choice.

AI labels. The EU adds a duty in some cases. Under Article 50(4) of the EU AI Act, deployers of an AI system that generates text published to inform the public on matters of public interest must disclose it. The duty does not apply if the text went through human review or editorial control and a person or company holds editorial responsibility. Whether a marketing blog falls under it is a question for a lawyer.

These rules change, so the owner rechecks them every quarter.

Who owns the AI use policy, and how do you roll it out?

One named person owns the AI use policy for writing. That person keeps it up to date, maintains the tool list and takes questions. So each role in the newsroom has a clear duty, and whoever publishes answers for what goes live. Finally, management sets the consequences.

To roll out the AI use policy template for writing, follow five steps:

  1. Adapt the template: replace the fields, delete what does not fit and add each client’s specific rules.
  2. Validate with a lawyer: ask for a review of the contract, copyright and data protection points.
  3. Present it to the team: a 30-minute talk, with examples of allowed and banned uses.
  4. Add it to onboarding and contracts: freelancers and vendors accept the policy before they deliver text.
  5. Review it every quarter: check the tool pages and Google’s guidance.
10. RESPONSIBILITIES
- Writer: uses only approved tools, flags the data
  to check and follows section 7.
- Fact-checker: checks each fact at the source and
  logs the result.
- Editor: approves voice, disclosure and publishing.
- Policy owner: keeps the tool list, answers
  questions and updates this document.
- Management: sets the consequences and provides the
  work accounts.

11. INCIDENTS AND POLICY REVIEW
- Whoever sent banned data tells the owner the same
  day.
- The owner assesses the case, tells the client if
  the contract requires it and logs what changed.
- We review this policy every 3 months and when a
  Google or tool rule changes.

One last caution: the policy should not rely on text detectors, because their score does not prove who wrote the text. The guide on how Google detects AI-made text explains why.

Frequently asked questions

Do we need an AI use policy if the team rarely uses AI?

Yes, but a short version is enough. The risk of error exists even with occasional use, because Google asks for review of all generated content. In addition, the individual plans of some tools may use chats for training. So a page with sections 2, 5 and 7 already covers the essentials.

Does the AI use policy apply to freelancers and vendors?

It does, and it should be in the contract or the brief. After all, text bought from third parties carries the same risks as in-house text, and whoever publishes answers for the result. Also ask freelancers to accept the policy, name the tools they used and deliver the sources for each fact.

Do we have to tell the client we use AI?

In general, agree on this in writing. The policy can require that the client knows when the team uses AI on their material. However, the contract rules on what is mandatory, and here a lawyer helps. For readers, Google recommends disclosure when they would expect to know how the content was created.

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Main sources

  • Google Search Central, guidance on using generative AI content on your website (2026).
  • U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (2025).
  • Federal Trade Commission, final rule banning fake reviews and testimonials (2024).
  • European Commission, AI Act Service Desk, Article 50 on transparency obligations (2026).
  • OpenAI, how your data is used to improve model performance (2026).
  • Anthropic, data use for training on the individual plans (2026).
  • Anthropic, commercial terms, section on customer content.
  • Google, Gemini Apps Privacy Hub (2026).
About this articleWritten with AI and reviewed by humans, with verified sources. Found something out of date? Write to the NotHuman team.

Get new NotHuman articles by email

By subscribing, you agree to receive blog articles by email and to the Privacy Policy. Unsubscribe any time.

Keep reading

Demo

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Book a demo

Your privacy

We use necessary cookies to make the site work and Google Analytics and Microsoft Clarity statistics cookies to improve our content. You can decline the statistics ones; in the EU, UK and Switzerland they are only turned on after you accept. Cookie Policy.

Cookie preferences

Cookie Policy