To choose an AI marketing agency, ask who does the work, whose accounts it lives in, who approves it before it goes live and which one number they measure each month. Then check the contract: a data processing agreement under GDPR Article 28, the EU AI Act transparency rules, and who keeps the prompts and files. Ask for proof you can open, and start with one small project.
Key takeaways
- People: ask for the names of the people who do your work and of the person who checks what AI makes.
- Accounts: the work runs in accounts you own, and you keep the prompts, briefs and files at the end.
- Guarantees: Google says no one can guarantee a first place in its search results.
- Law: a data processing contract under GDPR Article 28, and AI Act transparency rules that apply since 2 August 2026.
- Proof: ask for live work with dates and sources, including the numbers the agency has not measured yet.
- Bottom line: ask every agency the same 10 questions, and start with one small project measured by one number.
The work of an AI marketing agency
An AI marketing agency makes marketing and sales work with AI tools, while people with industry experience decide what to make and check the result. Typical work is ad content, product images and video, SEO and AI search, websites, AI workflows for the team and AI training.
AI makes production fast. A product video, a set of ad images or the first draft of a page now takes hours. People still choose the angle: who buys, why they buy and what makes them act. That choice needs someone who knows your market.
This guide gives you 10 questions to ask any agency, what a clear answer sounds like, and the EU rules that apply to AI work. Where it helps, we describe how we work at AI for Marketing Ops, so you have one example to compare with.
Questions to ask an AI marketing agency
Ask every agency on your list the same 10 questions, in writing if you can. They cover people, review, approval, accounts, ownership, measurement, data, AI rules, proof and the first project. A clear answer names a person, a place or a number.
Some questions come from Google’s own advice on hiring an SEO. Google suggests asking for examples of previous work, experience in your industry, the results to expect and the timeframe, and how you will communicate.
| Question | A clear answer includes |
|---|---|
| Who does the work on my account? | The names of the people, their industry experience and the tasks AI does |
| Who checks the AI output before I see it? | A named reviewer and the checks for facts, claims and brand |
| Who approves before anything goes live? | You do, for every page, post, ad and message |
| Whose accounts does the work live in? | Yours: website, ad accounts, analytics, Search Console and social profiles |
| Who owns the prompts, briefs and files? | You do, during the work and after it ends |
| Which one number do you measure each month? | One number in the contract, its starting value and how it is measured |
| How do you handle personal data? | A contract under GDPR Article 28 and a list of the AI tools that see the data |
| How do you meet the EU AI Act rules? | When people are told they talk to AI, and which content needs a label |
| What proof can I open and check? | Live work with dates and sources, and the numbers not measured yet |
| How do we start, and what happens at the end? | One small first project, then documentation and a handover session |
People, AI and approval in an AI marketing agency
Ask who does the work, who checks what AI makes and who approves it before it goes live. A clear answer names the people and their experience, describes the checks, and gives you the final approval. Google says you are responsible for the actions of any company you hire, so the last approval should be yours.
Google’s guidance on generative AI content asks you to check and review all AI-generated content for accuracy before you publish it. The same review applies to titles, meta descriptions, structured data and image alt text. Google also says that many pages made without value for users can break its spam policy on scaled content abuse.
At AI for Marketing Ops, there are no account managers between you and the work. You talk to Lautaro or Vytas, the people who do it. AI does the production, one of us reviews everything it makes, and nothing goes live without your approval. This is how we make AI content for ads and sales: the buyer and the angle first, then the variants to test.
Ownership of your accounts, prompts and files
Your website, ad accounts, analytics, Search Console and social profiles should stay in your name, with you as the owner. The agency gets its own user with the access it needs. Prompts, briefs, templates and source files should sit in a folder you own, so the work can continue if the agency stops.
Google Ads requires third parties to use a separate account for each advertiser they manage. For an SEO audit, Google suggests that you give only read access to Search Console, and no write access at that stage. In both cases, you own the account and the agency gets limited access.
Ask what the handover includes before you sign. A clear answer lists documentation, the prompts and briefs, the source files, and a session where your team learns how the work runs. Every AI for Marketing Ops project ends with documentation and a handover session. If your team wants to run the work itself, our AI workshops train it on its own tasks, and the team keeps the prompts, briefs and notes.
GDPR data processing contract with a marketing agency
If the agency handles personal data for you, such as customer lists, CRM records or email addresses, the GDPR requires a contract under Article 28. The contract says what the agency may do with the data and what it must do at the end. Ask for it before the agency gets access to any personal data.
Article 28 says the contract must set out the subject, duration, nature and purpose of the processing, the type of personal data and the groups of people concerned. The agency may only act on your documented instructions. It cannot use another processor without your prior written authorisation. Ask which AI tools see your data, because each tool provider can be such a processor.
At the end of the work, Article 28 lets you choose: the agency deletes all the personal data or returns it to you. The agency must also give you the information you need to check that it meets these rules, including audits. This guide is not legal advice, so check the contract with your lawyer or your data protection officer.
EU AI Act transparency rules for marketing content
Article 50 of the EU AI Act applies since 2 August 2026. It covers AI that talks to people, AI-made content and deep fakes. Your agency should tell you which of your work falls under it and how each case is disclosed.
EU AI Act, Article 50
Updated November 2026aimo
AI Act transparency rules cover chatbots, AI output, deep fakes and public text
What each paragraph asks for. Check which rows match the marketing work your agency does for you.
Rule by type of work
- Chatbots and AI assistants People are told it is AI, unless obvious 50(1)
- AI images, audio, video, text The tool marks its output as AI-made 50(2)
- Deep fakes You disclose that AI made or changed it 50(4)
- AI text that informs the public Disclosed, unless a person reviewed it 50(4)
Applies since 2 August 2026. This is a summary, not legal advice. Check your own case with your lawyer.
Paragraph 50(1) says people must be told when they talk to an AI system, unless it is obvious. Under paragraph 50(2), the providers of AI tools mark their output in a machine-readable format as AI-made. Paragraph 50(4) asks you to disclose deep fakes as AI-made or changed.
The AI Act defines a deep fake as AI image, audio or video that resembles real people, objects, places or events and could seem authentic or true. Paragraph 50(4) also covers AI text published to inform the public on matters of public interest. That text needs a disclosure, unless a person reviewed it and someone holds editorial responsibility for it.
Proof from an AI marketing agency
Ask for proof you can open and check: live work, dates, the state before the work and numbers with their source. A clear answer also says what the agency has not measured yet. For each number in a case study, ask for the page, the account or the report behind it.
Here is our own example, with what we can and cannot show. For RNJ Transport, a road freight company in Lithuania, the first call was on 19 October 2026. The new website went live on its domain on 21 October, with 944 of 944 pages answering correctly. It runs in 11 languages, and Lighthouse scored it 97 for mobile performance and 100 for SEO.
When we collected the case facts on 3 November 2026, no traffic, lead or AI citation numbers were recorded yet, so we claim none. CreatorFlow, the other case on our site, is the SaaS company our co-founder Vytas runs, not a client.
- A live link. You can open the site, the ad or the profile yourself.
- Dates. Each number has the date it was measured.
- The state before. You can see what the work changed.
- The source. Each number names the tool or the report it comes from.
- The gaps. The agency says which results it has not measured yet.
First project with an AI marketing agency
Start with one small project, one number to measure and a date to check it. A website revamp, a set of ad content or one AI workflow is enough to see how an agency works. Google says a good time to hire an SEO is when you plan a site redesign or a new site, and the earlier the better.
Small projects also protect the budget. In June 2025, Gartner predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027. It named rising costs, unclear business value and weak risk controls. Gartner also estimated that only about 130 of the thousands of vendors that sell agentic AI offer real agentic products.
At AI for Marketing Ops, the first step is a free 30-minute discovery call. Before the call, we check how Google and AI search see your company, and you keep the short plan from the call. Most projects go live in 10 working days. Monthly plans name one number in the contract, have a 3-month minimum and then run month to month. You can book the discovery call when your list of questions is ready.
Before you sign with any agency, run these five checks on the proposal and the contract.
- Goals, not guarantees. Google says no one can guarantee a first place in Google, so check that the contract names a goal and a number.
- Named people. Check that the proposal names who does the work and who reviews it.
- Your business in the plan. Check that the plan uses your products, buyers and numbers.
- Your accounts. Check that you own every account the work uses.
- Data and AI rules. Check for the Article 28 contract and the list of AI tools that will see your data.
Common questions
Can an AI marketing agency guarantee rankings in Google or ChatGPT?
No. Google says no one can guarantee a first place in its search results, and no agency controls what AI assistants such as ChatGPT or Gemini say. A clear agency names a goal, one number to measure each month and the work it will do. It then reports the number every month, also when it falls.
Do I need a data processing agreement with a marketing agency?
You need one when the agency processes personal data for you, such as customer lists, emails or CRM records. Article 28 of the GDPR requires a contract that sets out the data, the purpose and the duration. If the agency never touches personal data, ask it to confirm that in writing.
Does AI-generated marketing content need a label in the EU?
Some of it does. Since 2 August 2026, the EU AI Act requires disclosure for deep fakes, for chatbots unless it is obvious, and for AI text that informs the public without human editorial review. Google also asks that AI-generated product titles and descriptions in Merchant Center be labelled as AI-generated.
Who owns the content an AI marketing agency makes for me?
The contract decides. Check that it names you as the owner of the finished work, the prompts, the briefs and the source files. Ask which AI tools made the work and check that their terms allow commercial use. Keep everything in folders and accounts that you own.
How long does an AI marketing agency take to show results?
Ask the agency to write the result it expects and the timeframe, as Google suggests when you hire an SEO. Production work such as a website or a set of ads has a clear delivery date. Search results take longer and need a number measured each month. With AI for Marketing Ops, most projects go live in 10 working days.
Should I hire an AI marketing agency or train my own team?
Many teams do both. An agency delivers the first project fast, and training lets your team run the work after the handover. If no one on your team has hours each week for the work, hire help first. If someone does, training keeps the skills inside the company.
Sources and checks (8)
- Google Search Central: Do you need an SEO?
- Google Search Central: guidance on using generative AI content on your website
- Google Ads Policies Help: about Google third-party policies
- GDPR, Regulation (EU) 2016/679, Article 28: processor (EUR-Lex)
- European Commission, AI Act Service Desk: Article 50, transparency obligations
- European Commission, AI Act Service Desk: Article 3, definitions (deep fake)
- European Commission, AI Act Service Desk: implementation timeline (checked November 2026)
- Gartner: over 40% of agentic AI projects will be cancelled by end of 2027 (June 2025)