AI Output Disclaimers For Marketing Agencies

Do you need to tell your clients when you use AI to create their content, ads or strategy work? Usually, yes - or at least, you should be very clear about how AI is being used and what you aren't promising.

The reassuring part is that an AI output disclaimer doesn't need to be scary or full of legal jargon. For a New Zealand marketing agency, it's really about setting sensible expectations around accuracy, originality, approvals, privacy and brand risk.

If you use tools like ChatGPT, Claude, Gemini, Midjourney, Canva AI or AI features inside ad platforms, your client contract should keep up. A good disclaimer can help your clients feel informed, while giving your agency a clearer process if an AI-assisted output turns out to be wrong, off-brand or too close to someone else's work.

Do You Need An AI Output Disclaimer For Client Work?

Yes, if you use AI in client work, you should have an AI output disclaimer or AI terms in your client agreement. This is especially true if AI helps you create copy, images, campaign concepts, reports, research summaries or recommendations.

An AI output disclaimer is a clause that explains how you may use AI tools and what limits apply to the results. It helps answer practical questions like:

• Will the agency use AI to create drafts, ideas or final deliverables?
• Who checks the accuracy of AI-assisted content?
• Can the client rely on AI-generated facts, statistics or recommendations?
• Who is responsible for approving claims before publication?
• What happens if AI output resembles existing content or doesn't match the client's brand?

In New Zealand, there's no single AI law that tells every marketing agency exactly what wording to use. Instead, your obligations come from the usual business laws, your contract and the way you represent your services.

For example, the Fair Trading Act 1986 can apply if your marketing or client communications are misleading or deceptive. If you tell a client that every piece of content is researched from primary sources and hand-written by a senior strategist, but you're relying heavily on AI drafts without review, that could create problems.

The Consumer Guarantees Act 1993 may also matter in some service arrangements, although many agency-client contracts are business-to-business. Where both parties are in trade, you may be able to contract out of certain Consumer Guarantees Act guarantees if the wording is clear, in writing and fair and reasonable. This is something to handle carefully in your terms, not with a casual line in a proposal.

Can An AI Disclaimer Remove Your Liability?

No, an AI disclaimer can't remove all liability, but it can narrow expectations and make responsibilities much clearer. Think of it as a risk allocation tool, not a magic shield.

You can't use a disclaimer to excuse misleading conduct, false claims or work that falls well below what you agreed to provide. If your agency makes a promise, your contract wording should match what you can actually deliver.

For example, if your website says you provide fact-checked medical advertising copy, you shouldn't rely on a disclaimer buried in your terms saying AI outputs may be inaccurate and aren't checked. Those two messages don't sit comfortably together.

What a disclaimer can do is make clear that:

• AI tools may be used to assist with drafts, ideas, research prompts or production workflows
• AI outputs can contain errors, outdated information or unsupported statements
• the client must review and approve final deliverables before use
• the client is responsible for confirming industry-specific, technical, legal, financial or medical claims
• your agency doesn't guarantee that AI-assisted content will be unique, registrable as IP or free from all third party similarities
• your liability is limited in a fair and clearly drafted way

Liability clauses work best when they sit with the rest of your contract, including your statement of work, approval process, payment terms, change requests and any limitations on consequential loss. Templates often miss the detail that matters for your agency's actual services.

What Should Your AI Output Disclaimer Say?

Your AI output disclaimer should explain your AI use, the limits of AI outputs and the client's approval responsibilities in plain English. It shouldn't be so broad that it contradicts the service you're selling.

A practical clause usually covers these areas.

Use Of AI Tools

Say that your agency may use AI-assisted tools to help produce or support the services. You might refer to drafting, ideation, editing, image generation, analytics, workflow automation or content adaptation.

Be careful with promising that you'll disclose every single AI use if that isn't practical. If your team uses AI inside common platforms, like grammar tools, design software or ad platforms, a broad disclosure may be more realistic than a per-output notice.

Human Review

Explain whether your team reviews AI-assisted outputs before sending them to the client. If your service includes human review, say what kind of review it is.

For example, reviewing for tone and brand consistency is different from verifying every scientific claim, checking copyright ownership or confirming compliance with a regulated industry's advertising rules.

Accuracy And Reliance

AI tools can confidently produce wrong information. Your disclaimer should say that AI-assisted outputs may contain errors, omissions or unsupported statements, and that factual claims should be checked before publication.

This is particularly important for sectors like health, finance, property, education, employment and food or cosmetics advertising, where inaccurate claims can create bigger compliance issues.

Client Approvals

Your contract should say that the client is responsible for reviewing and approving final deliverables before they go live. This includes checking that claims about their products, pricing, qualifications, results, customer testimonials and offers are accurate.

Without a clear approval process, you may end up arguing about whether the client signed off on a risky claim or whether your agency should have spotted it.

A Practical Sample Clause

Sample wording: Our team may use AI-assisted tools to help create, edit, analyse or improve deliverables. AI-assisted outputs can contain inaccuracies, omissions, outdated information or content that requires further review. We'll use reasonable care in preparing deliverables within the agreed scope, but we don't guarantee that AI-assisted outputs will be error-free, suitable for every purpose, unique, registrable as intellectual property or compliant with industry-specific rules unless we have expressly agreed to provide that review. You're responsible for reviewing and approving all final deliverables before publication, including product claims, technical information, legal or regulatory claims, pricing and brand requirements.

This wording is only a starting point. Your agency may need tighter or softer wording depending on whether you provide strategy, ad management, social media content, SEO blogs, design, email marketing services, influencer campaigns or regulated-industry work.

How Should You Manage Accuracy And Brand Risk?

You should manage accuracy and brand risk with a review process, not just a disclaimer. A clause helps, but your day-to-day workflow is what usually prevents issues.

Say you run a Wellington digital agency and create a Facebook ad campaign for a skincare client. An AI tool suggests the phrase clinically proven to reverse skin ageing. It sounds persuasive, but the client has no clinical trial data to support it.

If that claim goes live, the issue isn't just that AI wrote it. The practical problem is that an unsupported claim may mislead customers, which can create Fair Trading Act risk and brand damage for your client. Your agency may also face a difficult conversation about who approved the wording.

A sensible AI workflow might include:

• a prompt rule that tells your team not to ask AI to invent statistics, testimonials or scientific claims
• a fact-checking step for claims about performance, pricing, comparisons and results
• a client approval step before publication
• a special review process for regulated or higher-risk industries
• a record of final approved copy, creative and campaign settings

Brand risk also deserves attention. AI can produce content that's technically fine but wrong for the client's voice, values or market. Your terms should say that the client must provide brand guidelines, product details and feedback in time for you to do the work properly.

If the client gives you limited information, your contract should let you rely on what they provide. Otherwise, your agency can be left carrying the risk for missing details only the client knew.

Who Owns AI-Generated Content In New Zealand?

It depends, because ownership of AI-generated content can be less clear than ownership of traditional creative work. Your contract should deal with IP rights directly rather than assuming the answer is obvious.

New Zealand copyright law sits under the Copyright Act 1994. Copyright generally protects original works, such as written content, images, graphics and videos. AI creates extra complexity because some outputs may involve limited human authorship, and the tool's own terms may affect what rights users receive.

For marketing agencies, the main IP risks are usually practical:

• the AI tool's terms may give you a licence to use outputs, but not a strong warranty that they're unique
• AI outputs may resemble existing content, slogans, logos, images or campaign concepts
• the client may want exclusive ownership of deliverables, even though full exclusivity may be hard to promise for AI-assisted ideas
• AI-created logos or brand assets may be harder to protect if they're generic or too similar to existing marks
• trade mark registration through IPONZ may still require a distinctive sign and proper clearance searches

Your marketing services agreement should separate different types of materials. For example, you may keep ownership of your templates, methods, know-how, prompts, pre-existing materials and internal tools. The client may receive ownership or a licence to final approved deliverables after payment.

If you create logos, taglines or brand names, be extra careful. An AI tool can suggest a catchy name that someone else already uses. Before a client invests in packaging, ads or a website, it's usually worth checking availability through trade mark searches and business name searches, including IPONZ and Companies Office records where relevant.

What Privacy Rules Apply When You Use AI Tools?

The Privacy Act 2020 applies if you put personal information into AI tools or use AI outputs that involve identifiable people. Personal information can include names, email addresses, customer data, photos, voice recordings, behavioural data and sometimes social media handles.

Many AI tools process data through overseas servers or use inputs to improve their systems, depending on the plan and settings. That can matter under the Privacy Act's Information Privacy Principles, especially around purpose, security, disclosure and overseas disclosure.

For client work, you should be clear about what your team can and can't put into AI tools. A practical internal rule might be:

• don't enter client customer lists into public AI tools
• don't upload confidential campaign data unless the tool and account settings have been approved
• remove names, contact details and identifiers where possible
• use business accounts with suitable privacy and security settings
• check whether the client has given you permission to use their data in that way

Your client contract should also deal with confidential information. AI use can blur the line between normal production work and disclosure to a third party platform. If you're uploading a client's unreleased campaign, product roadmap or customer data into an AI tool, your confidentiality clause needs to match that reality.

If you handle personal information at scale, or you work with sensitive sectors, it may be worth having an AI use policy as well as client-facing terms. The policy can guide your team on approved tools, prohibited data, review steps and escalation if something goes wrong.

Where Should Your AI Terms Sit In Your Client Contract?

Your AI terms should sit in your main client agreement, with shorter reminders in proposals, scopes of work and approval emails where needed. Don't rely on a single line in a footer.

For most agencies, the key documents are:

• Master services agreement: your main legal terms covering scope, fees, approvals, IP, confidentiality, liability and termination
• Statement of work: the project-specific details, deliverables, timelines, review rounds and client responsibilities
• Proposal: the commercial summary, which shouldn't overpromise what AI-assisted work can achieve
• Privacy policy or data terms: how personal information is collected, used, stored and disclosed
• Internal AI policy: team rules for using AI safely and consistently

Your contract should also deal with marketing performance claims. If you use AI to support campaign strategy or ad optimisation, avoid guaranteeing results unless you're prepared to stand behind that promise. Many factors sit outside your control, including the client's budget, website conversion rate, product-market fit, stock levels and platform changes.

If you use freelancers or contractors to help with AI-assisted content, make sure their agreements cover confidentiality, IP ownership, tool use and client data. The label you use for a worker also needs to reflect the real arrangement. Employment status can affect your obligations under the Employment Relations Act 2000, and New Zealand's 2026 contractor gateway changes make employment contract and contractor arrangements even more worth getting right from the start.

Key Takeaways

• An AI output disclaimer in New Zealand should explain how you use AI, what you check and what the client must approve.
• A disclaimer can't remove all liability, especially for misleading claims under the Fair Trading Act 1986.
• Your contract should deal with accuracy, IP ownership, privacy, confidentiality, client approvals and liability limits.
• AI-generated content can raise copyright, trade mark and originality issues, so avoid promising full uniqueness unless you have done the right checks.
• If you put personal information or confidential client data into AI tools, the Privacy Act 2020 and your confidentiality obligations may apply.
• Your AI terms work best when they sit in your main client agreement, supported by clear scopes of work and an internal AI policy.

If you'd like help with your AI clauses, client services agreement, privacy terms or internal AI use policy, you can reach us at 0800 002 184 or team@sprintlaw.co.nz for a free, no-obligations chat.