All insights

How US and UK brands choose an AI creative studio (2026)

US and UK brands choose an AI creative studio by settling AI labels, likeness consent and file ownership in writing, then testing it against their best ad.

Kevin Dosanjh
Co-founder / CMO, AI Vidia · Updated

US and UK brands choose an AI creative studio through a paid pilot against their best ad, run in their own ad account. Before it starts, get written answers on who applies the AI label in each market, how real faces and voices are cleared, and who owns which parts of each file. New York requires synthetic performer disclosures, California follows on 1 January 2027, and EU-facing deep fakes need a label under the AI Act.

A long oak table in a bright room with production kit laid out in a row: a laptop on a stand, a phone on a small tripod, wrapped white card stacks, a camera beside a bar of soap and a kraft box tied with twine
Image generated with AI: a studio table laid out for product ad production, from laptop and phone to camera and packaging.
On this page9 sections
  1. 01Shortlist studios on your hardest product
  2. 02Which AI labelling rules apply to US, UK and EU ads?
  3. 03Who applies the label: your brand, your agency or the studio?
  4. 04Consent for a real face or voice comes before any label
  5. 05Who owns the files, and what the signed scope must say
  6. 06Bring your agency of record in before the first brief
  7. 07Design the pilot so one number decides it
  8. 08Red flags that should end a shortlist
  9. 09How AI Vidia answers these checks

A US or UK consumer brand picks an AI creative studio through a paid pilot on its own ad account. Before that pilot starts, legal and procurement need written answers on AI labels, consent for real faces and voices, file ownership and the agency of record. Shortlist on relevant work. Then settle the paperwork and test against your best ad.

Shortlist studios on your hardest product

Ask every studio on the list for work that shows the product problem you actually have. Reflective packaging, fabric that has to drape, a pump that has to press and small label text all fail in different ways, and a reel of moody brand films proves none of them. If you have no list yet, start with our co-founder Kenneth Dosanjh's comparison of five AI creative agencies, AI Vidia among them. It sets out each agency's documented work, what it suits best and its published pricing or budget signals.

Ask the same of each sample. Is it client work or a demo? Which parts were generated, filmed or photographed, and what did the studio itself make? Labelled demos are fair evidence.

Then ask for one job that went wrong. A shareable account of the problem, the decision and the delivered fix shows how the team handles a product the model cannot render, and it needs no confidential client data.

Forty grey vertical phone-shaped cards in a grid on an oak desk, with two cards near the centre glowing orange
Image generated with AI: a review grid of vertical ad frames with two marked, the way a buyer flags the few samples that show its own product problem.

Which AI labelling rules apply to US, UK and EU ads?

Each market sets its own rule. Four regimes matter to a US or UK brand that also sells into Europe, and the federal FTC rule sits underneath the US ones.

AI ad disclosure rules a US or UK consumer brand meets, as of 22 October 2026
RuleAppliesWhat triggers itWhat the ad needs
New York GBL section 396-bSince 9 June 2026A synthetic performer: a realistic AI-made person who is not a recognisable real performer, where the business that makes the ad knowsA conspicuous disclosure in the ad. Civil penalties of USD 1,000 for a first violation and USD 5,000 after that
California SB 1050From 1 January 2027A synthetic performer shown prominently, in video or audio adsA clear and conspicuous disclosure, worded close to "this performance features a synthetic performer"
EU AI Act, Article 50(4)Since 2 August 2026A deep fake shown in the EU, including realistic AI people and product images that mislead about how the product looksA visible label at first exposure, applied by the deployer. Fines up to EUR 15 million or 3% of worldwide annual turnover, lower for SMEs
UK CAP and BCAP CodesIn force; CAP guidance of 11 June 2026Any ad, however it was madeNo blanket legal duty to disclose AI. The ad must not mislead, and the advertiser stays responsible for the AI output it uses
US FTC fake reviews ruleSince 21 October 2024 (announced 14 August 2024)Reviews or testimonials by someone who does not exist, such as AI-generated fake reviewsNo AI presenter speaking as a real customer

Two details change most buying decisions. California's SB 1050 reaches audio ads, which New York's section 396-b exempts. And the Commission's Article 50 guidelines (20 July 2026) bring in a company established outside the EU when its deep fake ad is displayed in the Union, so a US brand running EU campaigns plans for Article 50 as well.

The UK works differently. CAP's guidance of 29 May 2025 says "There is no blanket legal requirement in the UK to disclose the use of AI in ads", and asks whether the audience is likely to be misled without a disclosure. It adds that "disclosure alone is very unlikely to mitigate the harm caused by a fundamentally misleading message." CAP wrote on 11 June 2026 that "the advertiser remains responsible for any output used in their ads".

The statutes are set out clause by clause in what New York's and California's synthetic performer laws require, and the EU roles in who has to label an AI ad under the EU AI Act.

Who applies the label: your brand, your agency or the studio?

Name one party per market. Under Article 50(4) of the AI Act, the visible label on a deep fake is the deployer's duty, and a deployer is whoever uses the AI system under its authority, such as the studio or agency that decides how AI is used. The guidelines add that a brand that "merely commissions an advertising agency to produce an advertisement", without deciding whether and how AI is used, is not a deployer.

On our reading of that passage, a brand that chooses the synthetic presenter is taking those decisions and becomes a deployer next to its supplier. New York puts its duty on the business that produces or creates the ad, where it has actual knowledge of the synthetic performer. If your team briefed or approved the AI presenter, plan as a party that knows.

Platform settings help. They do not settle it. Google's help page on AI label settings warns that the setting "doesn't guarantee your compliance with specific regulations". Put the disclosure in the creative itself, then mark the asset on each platform as well.

Need new ads?
Describe your product and what you want to test.
Send your brief

A disclosure line covers an invented person. A real person needs consent. New York's Civil Rights Law section 50 bars using "the name, portrait, picture, likeness, or voice of any living person" for advertising without first getting their "written consent".

California's Labor Code section 927 makes vague digital replica clauses in a performer's contract unenforceable for new performances fixed on or after 1 January 2025, unless the performer had a lawyer or a union. In the UK, CAP tells advertisers to "consider the legal risks of using a public figure's likeness without permission".

AI presenters must not pose as customers either. The FTC's rule on fake reviews and testimonials covers reviews that "misrepresent that they are by someone who does not exist, such as AI-generated fake reviews". Script an AI presenter as a brand spokesperson describing documented features, and ask the studio where each consent record is kept and which delivered file it covers.

Who owns the files, and what the signed scope must say

Ownership needs a file-level answer. In Part 2 of its report on copyright and AI (29 January 2025), the US Copyright Office concluded that "copyright does not extend to purely AI-generated material", and that with current tools "prompts do not alone provide sufficient control". On the Office's reading, what a person wrote, shot, selected, arranged or edited is what US copyright can protect.

So an "all rights" clause tells you little by itself. Ask the studio to record, file by file, which parts were generated and which a person made, and to list every deliverable: finished files, project files and a record of the AI use. The guide to AI ad usage rights has the complete rights schedule.

Tick each line below only once the answer is in the signed scope.

ChecklistWhat the signed scope should answer before the pilot

0 of 9 done

AI labels
  • New York now, California from 1 January 2027, and the EU for ads shown there

  • On screen from the moment an AI person appears, readable on a phone

Faces and voices
  • Kept with the delivered file it covers

  • No invented customer experiences in the first person

Files and ownership
  • Ask for project files now if another supplier will edit the ads later

  • In the US, the human work is what copyright can protect

  • A new market or language version is a new use to check

Pilot and agency
  • Run in your own ad account

  • One approver on the brand side

Bring your agency of record in before the first brief

Many larger brands buy media and strategy through an agency of record, and an AI studio joins as a production partner. Settle the roles in one page. Your agency keeps the strategy, the media buying and the testing; the studio writes concepts to the brief, makes and checks the ads and delivers the files to whoever runs them. The ads stay in your own ad account.

The label chain needs the same clarity. The Commission's Article 50 FAQ says a deployer "remains a deployer even if third parties (e.g. contractors, freelancers) are involved", so an agency that decides how AI is used and hires a studio to run the tools keeps its duty. Write down which of you applies the label in each market.

Put dates on the review loop across time zones. Agree in writing the day and local time when consolidated feedback is due, when the corrected version comes back, and who decides a product question while your approver is offline. Two people sending conflicting notes across a time gap can cost a full revision round.

How this works with AI Vidia is on the pages for brand teams and for agency partners.

Design the pilot so one number decides it

A pilot answers one question: can this studio beat your best ad on your own account? Fix four settings before launch. The control is your current best-performing ad. The metric is one number, such as CTR or CPA against the control, and both the variant count and the test window are set in writing.

Run the arithmetic first. Say your control ad runs at a USD 40 CPA and you can put USD 400 a day behind the test. Over 14 days that is USD 5,600; split evenly across the control and ten variants, each ad gets about USD 509, or roughly 13 purchases at the control's CPA. If your analyst finds 13 purchases too thin for a CPA verdict, agree CTR as the metric or test fewer variants.

Platform mechanics matter too. How to test a new AI video against your best ad covers the built-in split tests on Meta, TikTok and YouTube and when CTR stands in for CPA.

AI Vidia's First Batch gives you material for this design: one concept made as 4 videos and 4 images for EUR 999 one-time, which you can run against your control in your own ad account, plus your own media spend. Hold any studio you test to the same four settings.

Red flags that should end a shortlist

Drop a studio that does any of these.

  • Guarantees a winning ad or a platform's approval. Approval is each platform's decision.
  • Answers the ownership question with "all rights" and cannot say which parts of a file were generated.
  • Has no answer on who applies the AI label in New York, California or the EU.
  • Scripts AI presenters to describe their own experience with your product.
  • Quotes by file count and cannot say how many distinct concepts a batch holds.
  • Promises finished ads in languages nobody on its team reviews.
  • Shows aggregate results with no named client behind them, or will not test against your current best ad.
Three groups of small vertical photo cards on a white surface: a loose dozen on the left, a dense pile in the middle and a packed block of well over a hundred on the right
Image generated with AI: three piles of ad cards that grow from a dozen to well over a hundred, the way a file count can climb without a single new idea.

A studio that turns down a shot it cannot render is protecting your product. Ask for that judgement during the pilot, and count it in the studio's favour.

How AI Vidia answers these checks

AI Vidia is the brand of Vidia ApS (CVR 44873443), a Copenhagen studio founded in 2024 that makes AI video and image ads for consumer brands. None of our published cases is an enterprise brand yet.

Every delivered file comes with a usage note that lists what in it is AI-generated, every ad with an AI presenter carries an on-screen AI label, and we agree any extra platform or market labels with you before delivery. AI presenters are AI-generated, not customers. A real person's face or voice is used only with their written consent.

You send the product photos you already have and sign a written product fact sheet before anything is generated. On a monthly plan, the first review batch arrives on day 3, and brief to delivery takes 7 days. Every plan includes one revision round per video. All our rights in the finished ads transfer to you on delivery, and you can edit them and run them with any vendor or in-house team; AI presenters are licensed as they appear in your ads, and our prompts, workflows and project files stay with us. The signed scope of work sets out the details.

We deliver in 9:16 as standard, and other formats can be agreed per client on request. English and Danish are reviewed in-house, so English versions for US and UK audiences are checked by us. Your team or ours launches the ads and runs the tests, agreed per client, and footage you already have can be cut in.

Our published plans are priced in EUR, from the EUR 999 First Batch to Studio from EUR 12,000 a month. Larger programmes, several markets or invoicing in USD or GBP are quoted on a call. Legal and procurement teams can start with our AI governance page, and market details are on the United States and United Kingdom pages.

Your next step takes about 20 minutes. Write one page for your shortlist with your control ad, the metric, the number of variants, the test window and the name of the person who will answer the label, consent and ownership questions. Send the same page to every studio and compare the written replies.

Frequently asked questions

01Does an AI creative studio need to be based in the US or UK to work for a US or UK brand?
A studio does not need to be in the US or UK to work for you: its work, its written answers and a pilot on your own account tell you more than its address. The rules follow the market too, and the EU guidelines give the example of a company established outside the EU whose deep fake ad is displayed in the Union. What location does change is the review loop: agree feedback deadlines by date and local time, and name one approver on your side.
02Who has to disclose an AI presenter in an ad shown in New York?
New York's rule applies to any business that, for a commercial purpose, produces or creates the ad, where it knows the presenter is a synthetic performer. New York General Business Law section 396-b has applied since 9 June 2026 and requires a conspicuous disclosure in the ad itself. Civil penalties are USD 1,000 for a first violation and USD 5,000 for each one after. California adds a similar rule from 1 January 2027 that also covers audio ads.
03Do UK ads have to say they were made with AI?
UK ads carry no blanket legal duty to disclose AI use, according to CAP guidance published on 29 May 2025. The test is whether the audience is likely to be misled without a disclosure, and CAP adds that a disclosure alone is very unlikely to fix a fundamentally misleading message. On 11 June 2026 CAP said the advertiser remains responsible for any AI output used in its ads.
04Who owns the AI-generated ads a studio makes for us?
Your signed scope decides what the studio hands over, and in the US Copyright Office's view only the human contribution in each file is protected by copyright. The US Copyright Office concluded in January 2025 that copyright does not extend to purely AI-generated material and that prompts alone do not give enough control. Ask the studio to record which parts of each file were generated and which a person wrote, shot, selected or edited, and to list every deliverable.
05How long should a paid pilot with an AI studio run?
Fix the window before launch and size it to the purchases your budget can buy. At a USD 40 CPA and USD 400 a day, 14 days buys roughly 13 purchases per ad across a control and ten variants. If that is too few for a CPA verdict, judge on CTR or test fewer variants, and write the choice down before the first ad goes live.
06Can an AI studio work under our agency of record?
An AI studio can work under your agency of record as its production partner, with the agency keeping strategy, media buying and testing. The ads still run in your own ad account. Agree in writing who briefs, who approves, who uploads and who applies the AI label in each market, because under the EU rules the party that decides how AI is used stays a deployer even when contractors do the work.

How we made this article

The legal facts are quoted from primary sources: the New York and California statutes, the EU AI Act, the Commission's Article 50 guidelines and FAQ, the US Copyright Office report and the Google Ads help page, read on 21 October 2026, and the CAP and FTC pages and the Federal Register notice of the FTC rule, read on 22 October 2026. Which party counts as a deployer in each case is our reading of the EU texts. The pilot example is arithmetic on made-up figures.

We research and draft with AI assistance and check figures against the sources below.

Sources

  1. 01New York State Senate: General Business Law section 396-B, synthetic performers in advertisements (revision of 12 June 2026)
  2. 02California Legislative Information: SB 1050 bill text, Chapter 246, approved 16 September 2026
  3. 03EUR-Lex: Regulation (EU) 2024/1689, the Artificial Intelligence Act, Official Journal 12 July 2024 (Articles 3, 50 and 99)
  4. 04European Commission: Guidelines on the transparency obligations of providers and deployers of AI systems, C(2026) 5054 final, 20 July 2026
  5. 05European Commission: Transparency obligations under Article 50 of the AI Act (FAQ), last updated 24 July 2026
  6. 06CAP: Disclosure of AI in advertising, striking the balance between creativity and responsibility, 29 May 2025
  7. 07CAP (on asa.org.uk): AI and deepfakes, four things advertisers need to know before they hit run, 11 June 2026
  8. 08Federal Trade Commission: Federal Trade Commission announces final rule banning fake reviews and testimonials, 14 August 2024
  9. 09Federal Register: Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 89 FR 68034, 22 August 2024 (effective 21 October 2024)
  10. 10New York State Senate: Civil Rights Law section 50, right of privacy
  11. 11California Legislative Information: Labor Code section 927 (AB 2602), digital replica contract provisions
  12. 12US Copyright Office: Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025)
  13. 13Google Ads Help: Use AI content label settings and disclosures (answer 17140115)

Next step

Get a proposal for your next ads.

Tell us about your product, audience and ad production needs. We review your brief and reply with a proposed next step.

Send your brief

Read next