AI can be a brilliant support tool in brand strategy work. It can help you process interviews faster, spot themes and patterns in customer and employee research, summarise long documents, and pull insights together more efficiently.

But here’s the part many brand strategists overlook:

If you’re using AI to analyse your client’s documents or research data, you need their permission.

Not just as a vague “heads up”, but ideally written permission inside your contract, so everyone is clear on what’s happening and why.

In this article, I’ll walk you through what clients are likely to care about, what to include in your contract clause, and I’ll share two ready-to-use clause options you can adapt for your own projects.

Why permission matters (even if your intentions are good)

When you upload or paste anything into an AI tool, you’re often sharing it with a third-party provider.

Even if you’re using paid plans, even if you trust the tool, and even if you’re only using AI to “speed things up”… your client may still have concerns like:

• “Is our information leaving our systems?”

• “Will our confidential docs be stored somewhere else?”

• “Could our data be used to train an AI model?”

• “Is this compliant with privacy laws?”

• “Have our employees/customers consented to this?”

Most of the time, clients aren’t anti-AI. They just want clarity, control, and reassurance.

And the best way to provide that is to make AI use explicit in your contract.

This is getting more urgent, not less. In 2026, a US federal court ruled that conversations with AI tools carry no legal confidentiality protection. That means if something sensitive ends up in an AI tool and there's a dispute later, you can't argue the conversation was confidential. A written clause in your contract is your protection. Without it, you're relying on goodwill and assumptions, and neither of those holds up well in a difficult conversation with a client.

What “using AI” actually means in brand strategy research

In brand strategy projects, AI typically shows up in places like:

1) Analysing client-owned materials

For example:

• brand docs and strategy decks

• existing research reports

• customer feedback

• internal culture documents

• workshop notes

• leadership speeches 

• messaging and positioning drafts

2) Analysing research you conduct during the project

For example:

• customer interviews

• employee interviews

• stakeholder workshops

• survey results

• transcripts (audio/video)

• thematic analysis across responses

This matters because the second category often includes third-party voices: customers and employees.

Which brings in extra sensitivity around consent and confidentiality.

What clients might be concerned about (and how to address it)

Here are the most common worries clients have when they see an AI clause in a contract:

1) Confidentiality: “Are you uploading our sensitive data into external tools?”

A good AI clause should reassure them that:

• you’ll use AI responsibly

• you’ll only share what’s needed

• you’ll use secure/confidential settings

2) Model training: “Will our data train the AI?”

This is the big one.

Many clients are fine with analysis support, but they do not want their documents or research data used to improve an AI product.

So it’s smart to explicitly include a line that says:
“No training / no model improvement.”

Make sure you have that turned off in all your AI tools.

3) Boundaries: “What exactly will you upload?”

Clients may want to avoid sharing:

• financial information

• product roadmaps

• sensitive employee comments

• protected customer data

Even if they don’t ask for it upfront, your contract clause can include a simple mechanism:
they can request exclusions in writing.

4) Accuracy and responsibility: “Are we relying on AI outputs?”

This isn’t always about privacy.
Some clients worry about AI making mistakes and being treated as truth.

The solution is simple:
AI supports your process, but you’re still doing the thinking.
You review, interpret, and apply professional judgement.

Brand strategy is not a skill you can outsource to AI! But AI can help you get to the obvious answers more quickly and speed up parts of the process.

Two contract clause options you can use

Not all clients need a heavy legal clause. Some just want a clear explanation and reassurance.

So here are two versions you can adapt: one “light” and one “enterprise ready”.

Get these as a Google doc you can edit here.

Version A: Client-Friendly AI Permission Clause (for smaller/less cautious organisations)

This is ideal for:

• small business clients

• agile projects

• clients who trust your process but want clarity

AI-Assisted Analysis (Client Permission)
The Client acknowledges and agrees that [Your Company Name] may use artificial intelligence tools (including paid/professional third-party services such as ChatGPT, Claude, Google Gemini, Otter, Notion AI, transcription tools, and survey analysis tools) (“AI Tools”) to support the Services. This may include using AI Tools to summarise, analyse, organise, synthesise, and extract insights from relevant information.

What this applies to
This permission covers:

1. Client materials the Client provides or makes available (including documents, reports, communications, and existing research); and

2. New research conducted as part of the Services, including interview notes, transcripts, survey results, workshop outputs, and other findings gathered from the Client’s customers, employees, or stakeholders.

Confidential and secure use
[Your Company Name] will use AI Tools only under secure and confidentiality-oriented settings and will share only the minimum information reasonably necessary for the relevant task.

No training / no model improvement
[Your Company Name] will not knowingly use Client materials or research outputs in a way that allows an AI provider to train or improve its models, where the AI Tool provides controls or contractual commitments to prevent this.

Data retention
The Client should be aware that AI Tool providers may retain submitted data for a period (which varies by provider and plan type) even where training is disabled. [Your Company Name] will use business or professional plan tiers where available, as these typically offer stronger contractual data protections, and will inform the Client on request of the retention policies applicable to the AI Tools in use.

Client instructions
If the Client wants certain information excluded from AI processing (for example, especially sensitive commercial information), the Client must notify [Your Company Name] in writing and [Your Company Name] will comply with reasonable restrictions.

Version B: Maximum Protection Clause (For larger organisations/ procurement/ cautious/compliance-led client )

A longer version like this may be needed for:

• corporate clients

• procurement-heavy organisations

• clients with formal privacy or compliance reviews

• industries where data risk is higher

This version includes extra details around:

• data minimisation

• personal data handling

• international processing

• retention/deletion

• opt-out options

(You can copy/paste the clause into your contract and adapt it to fit your services.)

AI Tools and Permission for AI-Assisted Analysis
“AI Tools” means third-party artificial intelligence and machine-learning tools used to assist with the Services, including paid/professional plans of tools such as ChatGPT, Claude, Google Gemini, Otter, Notion AI, transcription tools, and survey analysis tools.
“Client Materials” means documents, data, content, and information provided or made available by the Client.
“Research Outputs” means information created or collected as part of the Services, including interview notes, transcripts, survey responses, recordings, workshop outputs, findings, and analysis.

The Client authorises [Your Company Name] to use AI Tools to support delivery of the Services, including to transcribe (where applicable), summarise, analyse, categorise, and extract insights from Client Materials and Research Outputs.

Safeguards
[Your Company Name] will:

• use paid/professional plans where available;

• use confidentiality and secure settings intended to protect the information;

• share only the minimum necessary information; and

• where reasonably practicable, remove or redact personal data prior to processing with AI Tools.

No training / no model improvement
[Your Company Name] will not knowingly submit Client Materials or Research Outputs to AI Tools in a manner that permits training, fine-tuning, or improving the AI provider’s models using Client content, and will enable any available protections to prevent this.

Restrictions and opt-out
The Client may request reasonable restrictions on AI use (including an opt-out for certain categories of information) by notifying [Your Company Name] in writing. Where restrictions materially impact delivery, the Parties will agree any necessary changes to scope, timelines, or fees.

A simple way to reduce friction: add a checkbox option

If you want to make client approvals easier, add a checkbox under the signature section:

• ☐ AI-assisted analysis permitted (per clause above)

• ☐ AI-assisted analysis not permitted (manual analysis only; scope/timing may change)

This makes it feel collaborative rather than hidden.

Practical best practices (what I do)

Even with a clause in your contract, your process matters.

My general approach is:

• Use AI to speed up admin-heavy work (summaries, pattern spotting, clustering)

• Keep sensitive details out of prompts wherever possible

• Only share what’s needed for the task

• Apply human judgement to everything that makes it into final recommendations

Check your AI tool settings (this matters more than you think)

Your contract clause says "no training / no model improvement." But that clause only works if you've actually turned off training in the tools you're using.

Here's the problem. Most AI tools now train on your conversations by default, even if you're paying for them. A lot of freelancers assume that a paid subscription means their data is protected. It doesn't. Paying gets you faster responses and better features. It does not automatically stop your conversations from being used to train the model.

This changed significantly in late 2025 when several major providers quietly shifted to opt-out models. So if you haven't checked your settings recently, do it now.

ChatGPT (OpenAI) Go to Settings > Data Controls > "Improve the model for everyone" and turn it off. This applies whether you're on the free plan or paying for Plus.

Claude (Anthropic) Go to Settings > Privacy Settings > "Help improve Claude" and turn it off. Anthropic changed its default in October 2025. If you didn't actively opt out, your conversations may already be included.

Google Gemini Turn off "Gemini Apps Activity" in your Google account settings. Be aware that turning this off also removes your chat history.

The bigger point: personal plans vs business plans

If you're on a personal plan (even a paid one like ChatGPT Plus or Claude Pro), you need to manually opt out of training. The tools don't do it for you.

If you're on a business or team plan (ChatGPT Team, Claude Team, or Google Workspace with Gemini), training on your data is contractually excluded. You don't need to toggle anything.

Most freelancers and independent strategists are on personal plans. So if that's you, go check your settings before your next client project. This is one of those things that takes sixty seconds and genuinely matters.

Disclaimer

I’m not a lawyer, and this isn’t legal advice. This is my best practical approach at the time of writing, based on how I think brand strategists can use AI responsibly while protecting clients and building trust. If you work with highly regulated organisations or sensitive personal data, it’s worth getting a qualified legal professional to review your contract and process.

Want these clauses as a downloadable, editable Google doc? Click here. 

FAQs

Do I really need permission to use AI in brand strategy work?

If you’re analysing your client’s documents or research data using third-party AI tools, it’s best practice to get written permission. It protects you, sets expectations, and builds trust.

What if I’m only using AI to summarise notes?

Summarising interview notes or transcripts still involves processing information that may be confidential or personal. Permission is still the safest approach.

Will this scare clients or make them say no?

Usually, the opposite happens. Clients tend to appreciate transparency. A short, clear clause often reassures them more than avoiding the topic.

What if my client says “no AI allowed”?

That’s okay. You can still do the work manually. Just be prepared that it may affect the project timeline, depth, or cost.

Which clause version should I use?

• Use Version A for smaller clients who want clarity without heavy legal wording.

• Use Version B for corporate clients or anyone likely to run the contract through procurement.

Can I use these clauses internationally?

They’re a strong starting point, but laws and expectations vary by country. If your client is in a heavily regulated sector or if you’re processing a lot of personal data, get legal review.

Should I name the AI tools I’m using?

Yes, it can help with transparency. Listing tools like ChatGPT, Claude, Gemini, Otter, Notion AI, and transcription/survey tools gives clients confidence that you’re not hiding anything.

Should I include “no training / no model improvement”?

If you can, yes. It’s one of the biggest reassurance points for clients and helps reduce pushback.

Does it matter whether I'm on a free or paid AI plan?

Yes, but not in the way most people think. Paying for ChatGPT Plus or Claude Pro gets you better features, but it doesn't automatically protect client data from being used for training. You still need to manually opt out in your settings. The real privacy distinction is between personal plans and business/team plans. Business plans (like ChatGPT Team or Claude Team) contractually guarantee that your data won't be used for training. If you're doing regular client work, a business plan is worth the small extra cost.

Has anything changed since this article was first published?

Yes. I originally wrote this in early 2025 and have updated it since. The biggest changes: most major AI providers now default to training on your data (even on paid personal plans), a US court has ruled that AI conversations aren't legally confidential, and the EU AI Act compliance deadline (August 2026) is approaching, which means European clients may start asking more detailed questions about AI use. The core advice hasn't changed: get written permission, turn off training, and use professional judgement. But the case for doing it properly has only gotten stronger.