In short
AI in advertising means using machine learning and generative AI to plan, create, deliver and measure ads: from copy, images, video, voice and music to targeting, bidding and testing. An IAB report from July 2025 found that 86% of video ad buyers use or plan to use generative AI for video creative. The gains are speed and more variants; the risks are generic-looking work, rights questions and, in the EU, the duty to disclose realistic AI content since 2 August 2026.
AI in advertising is any use of artificial intelligence along the ad workflow, from research and creative production to media buying and measurement. It covers two quite different things: generative AI that makes assets (text, images, video, audio), and machine-learning systems that decide who sees which ad and what it may cost. This guide explains both, shows documented examples, and ends with a checklist for getting started.
What is AI in advertising?
AI in advertising is the use of machine learning and generative models to create, place and optimise ads. Machine learning has run inside ad platforms for years, in bidding, audience matching and performance prediction. Generative AI is the newer layer: it produces new copy, images, video, voices and music from a brief or a reference.
The two layers meet in practice. A generative tool produces many versions of a spot; a platform algorithm then decides which version reaches which audience. That is why people use terms such as “AI ads”, “AI-generated ads” and “AI-powered advertising” for slightly different things, and why a good plan names which layer it is talking about.
How is AI used across the advertising workflow?
AI can support every stage from the first research to the final report, but the value differs a lot by stage. The table shows where teams use it today and what to watch.
| Stage | What AI does | Typical output | Watch out for |
|---|---|---|---|
| Strategy and research | Summarises market, audience and competitor material; clusters search and social data | Briefing drafts, audience hypotheses | Unchecked summaries; verify every number |
| Copy | Drafts headlines, scripts and variants in several tones and languages | Ad texts, video scripts | Generic phrasing; brand voice needs a human editor |
| Images | Generates or edits key visuals, product scenes and backgrounds | Stills, social formats, banners | Wrong product details, rights and likeness issues |
| Video | Generates shots, animates stills, transforms footage | Spots, social cutdowns | Consistency from shot to shot; short clips need editing into one film |
| Voice and music | Creates voice-overs, translations and music beds | Narration, localised versions | Consent for any real voice; disclosure |
| Personalisation and targeting | Builds versions per audience, format or context; predicts who responds | Variant sets, audience segments | Privacy law (GDPR), bias in the data |
| Media buying and optimisation | Sets bids, budgets and placements automatically, tests creative | Automated campaigns | Less control and transparency; set clear guardrails |
| Measurement | Attributes results, forecasts, flags anomalies | Dashboards, reports | Platform-reported numbers; check against your own data |
Platforms now bundle several of these steps. Google, for example, describes optional asset generation inside Performance Max: you enter a landing page, and its AI suggests headlines, descriptions and images that you can edit or discard. According to the vendor’s help page (checked November 2026), generated images are identified with SynthID watermarking and metadata, and generated assets are not guaranteed to pass ad policy review, so you still check them before publishing. Meta reported that the combined annual revenue run rate of its video generation tools reached $10 billion in the fourth quarter of 2025.
AI in advertising examples, 2024 to 2026
The campaigns below are documented by the brand, its agency or press coverage; we re-checked each against the source listed at the end. For a longer, curated list of finished spots with the tools behind them, see our overview of AI-generated commercials.
| Brand and campaign | Date | What AI did | Outcome or note |
|---|---|---|---|
| Deutsche Telekom, Einstein film | June 2026 | A film of about 90 seconds, fully AI-generated, with Einstein’s voice recreated | Made by Jung von Matt with the AI production company PROMPTR; promotes Telekom’s partnership with ElevenLabs |
| McDonald’s Netherlands, holiday spot | Dec 2025 | A 45-second, fully AI-generated Christmas ad | Removed within days after criticism of the AI use and the tone |
| Lacalut (Dr. Theiss), anniversary spot | 2025 | Text-to-image scenes animated by image-to-video, with AI music and voice-over | Described by Seven.One as the first fully AI-generated spot on German TV; made with Havas |
| Coca-Cola, “Holidays Are Coming” remake | Nov 2024 | Three AI studios produced versions using several generative models | Criticised online as lacking human artistry |
| Toys”R”Us, brand origin film | June 2024 | Almost entirely generated with OpenAI’s Sora, with some VFX fixes | Premiered at Cannes Lions; made with Native Foreign |
| Mango Teen, “Sunset Dream” | July 2024 | A model trained on photos of the real garments placed them on AI-generated models | Retouched by the in-house art team; collection sold in 95 markets |
| Under Armour with Anthony Joshua | March 2024 | Mix of AI video and photos, AI voice-over, CGI and film footage | Credit dispute over reused footage; the director added the names |
| Klarna, marketing content | 2024 | AI image tools across marketing | Company figures: image cycle from six weeks to one, about $10 million saved per year |
The pattern across these cases: AI shortens production and multiplies versions, but the strongest results still come with a clear idea, human direction and careful selection. Lacalut’s own case page, for example, stresses detailed prompting, selection and revisions after client feedback.
What does the data say about AI in advertising?
Adoption is high among agencies and advertisers, while hard evidence on effectiveness is thinner than the headlines suggest. These are the figures we could trace to a primary source.
- 86% of video ad buyers use or plan to use generative AI for video creative, and about half already do. Buyers expect 40% of all ads to use generative AI creative by 2026, and 42% create audience-specific versions this way (IAB, report of 15 July 2025; survey date and sample size are not stated on the page).
- 98% of 201 surveyed German agencies use generative AI, 95% of them in creative processes (BVDW, 30 April 2025, fieldwork January to February 2025).
- Klarna reports that AI cut its image production cycle from about six weeks to one week (company statement, 28 May 2024).
- Meta reports a $10 billion combined run rate for its video generation tools in Q4 2025 (Meta, January 2026).
Treat vendor and company figures as claims by interested parties. None of the sources above shows that AI ads convert better than conventionally made ads; that depends on idea, audience and execution, and you should test it yourself.
Looking for help with AI commercials? Tell us about your project through the contact form.
What are the benefits and risks of AI in advertising?
The main benefits are speed, lower cost for certain formats and the ability to test many variants; the main risks are sameness, reputational backlash, rights issues and legal duties. Here is how they compare.
| Benefit | Matching risk | What helps |
|---|---|---|
| Faster production and iteration | Output looks generic or “soulless” | A strong concept and human art direction |
| Many versions for formats and audiences | Inconsistent brand look, errors in details | A review step for every version |
| Productions that would otherwise be out of budget | Backlash if AI replaces visible craft | Use AI where it makes the idea possible, not just cheaper |
| Automated targeting and bidding | Less transparency, data protection and bias issues | Own measurement, clear data rules |
Where AI ads went wrong
Coca-Cola’s 2024 remake drew criticism for replacing artists with AI. McDonald’s Netherlands pulled its 2025 holiday spot within days after viewers objected to the AI use and the mood of the ad. Under Armour faced a credit dispute, because footage from an earlier film had been reused without naming its makers. Vodafone Germany’s AI presenter on TikTok drew attention when a viewer noticed a mole that appeared and disappeared; Fortune reported that the clip did not disclose the presenter as artificial. The common thread is not the technology but the lack of craft, credit or transparency around it.
Do AI ads have to be labelled in the EU?
In many cases yes: since 2 August 2026, advertisers who publish deepfakes, meaning realistic AI-generated or manipulated images, audio or video, must disclose that. This follows from Article 50(4) of the EU AI Act, which puts the duty on the deployer, in our case the advertiser or agency that publishes the content. The disclosure must be clear and given at the latest when people first see the content.
For evidently artistic, creative, satirical or fictional works the duty is lighter: the notice must not spoil the enjoyment of the work. Providers of generative tools carry a separate duty under Article 50(2) to mark outputs in a machine-readable way. According to the European Commission’s FAQ, providers of systems already on the market before 2 August 2026 have until 2 December 2026 for that marking, and the Commission states that deployers cannot rely on those machine-readable marks to meet their own visible disclosure duty.
This is general information, not legal advice. How the rules apply to specific ad formats is still being clarified, and national unfair-competition law can apply on top. Our guide Do AI ads have to be labelled in Germany? goes through the German side, and our free AI labelling kit contains ready-to-use labels.
How do you start with AI in advertising?
Start small, with one format and one clear goal, and add steps only when you can measure them. This checklist works for brands and for teams briefing an agency.
- Pick one goal and one format. For example a 30-second spot for social media, or ad variants for one product.
- Decide where AI helps and where people stay in charge. Concept, brand decisions and final approval remain human.
- Check rights and data. Clarify licences of tools and training data, consent for voices and likenesses, and what personal data goes into targeting (GDPR).
- Plan the labelling. Decide early which assets are deepfakes in the sense of Article 50 and where the notice appears.
- Produce a pilot. Generate options, select, edit and mix them into one consistent piece instead of publishing raw output.
- Measure against a baseline. Compare with a conventionally made version and use your own data, not only platform-reported results.
- Document what you did. Keep a record of tools, prompts, approvals and labels for later audits.
If you want to see how an AI commercial gets from brief to broadcast, read AI commercial production from idea to broadcast. For explainer content, see AI explainer videos that convert.
How much does AI advertising cost?
Costs depend on length, the amount of live action and how much human work goes into concept, selection, editing and sound. At That Works Media, an AI spot of 20 to 60 seconds starts at 2,500 EUR net, and typical projects range from 3,000 to 12,000 EUR net. Shoot days, travel and locations fall away, and the effort moves into concept, image development, generation and post-production. Our article on what an AI commercial costs breaks down the drivers.
Frequently asked questions
What is AI in advertising?
It is the use of artificial intelligence to create ads, such as generated copy, images, video, voice and music, and to decide who sees which ad at what price. The first part is generative AI, the second is machine learning inside ad platforms.
How is AI used in ads?
Mainly for creative production, many variants per audience, voice and music, automated bidding and placement, and measurement. The workflow table above lists each stage with typical outputs and risks.
What are examples of AI-generated ads?
Documented examples include Deutsche Telekom’s Einstein film (2026), McDonald’s Netherlands’ holiday spot (2025), Lacalut’s anniversary spot (2025), and Coca-Cola’s remake, Toys”R”Us’s Sora film and Mango’s Sunset Dream campaign (all 2024).
Do AI ads need to be labelled?
Realistic AI-generated or manipulated images, audio and video that qualify as deepfakes must be disclosed under Article 50(4) of the EU AI Act since 2 August 2026. Evidently artistic or fictional works get a lighter form of notice. This is not legal advice.
Are AI ads better than traditional ads?
Not automatically. AI can make production faster and enable more variants, but we found no primary source showing that AI ads perform better as a rule. Test against a conventional baseline with your own data.
Plan an AI advertising campaign with us
If you are weighing AI for a spot, a social campaign or a set of ad variants, our team can help with concept, production and labelling. See our AI advertising service, or send us your brief through the contact form.
Sources
- IAB: Nearly 90% of advertisers will use gen AI to build video ads (15 July 2025)
- BVDW: study on generative AI use in agencies (30 April 2025)
- Meta: 2026, AI drives performance (January 2026)
- Google Ads Help: build a Performance Max asset group using generative AI
- Klarna press release (28 May 2024)
- Horizont: Jung von Matt and the Telekom Einstein film (5 June 2026)
- NBC News: McDonald’s Netherlands AI Christmas ad backlash (December 2025)
- Seven.One: Lacalut case
- NBC News: Coca-Cola AI-made ad controversy (November 2024)
- PR Newswire: Toys”R”Us Studios and Native Foreign (24 June 2024)
- Mango Fashion Group: Sunset Dream press release (10 July 2024)
- TechCrunch: Under Armour AI ad controversy (14 March 2024)
- Fortune: Vodafone Germany AI influencer (8 September 2025)
- EU AI Act, Article 50 (artificialintelligenceact.eu)
- European Commission: FAQ on transparency obligations under Article 50 AI Act
This article is general information and not legal advice. For your specific situation, please consult a lawyer.
Planning a project?
We produce AI commercials in Berlin. Tell us about your project through the contact form.
