HomeOperational AICreative AITrainingBlogContactStart a project
← All articles

Creative29 September 2026 · 7 min read

AI images and video for marketing: a guide for UK brands

The new AI video and image tools, how YouTube, Google Ads, Meta, TikTok and the EU label AI content, and a practical plan for UK marketing teams.

The tools for making marketing images and video with AI moved quickly through 2026, and so did the ways platforms label what they produce. The new models make short video and product imagery quick and affordable to test, and the labelling rules decide how that content appears once it goes live.

Here is what arrived between February and September 2026, and how to put it to work.

Video: one model to generate and edit

A major video launch of the year came at Google I/O on 19 May 2026, when Google announced Gemini Omni. It is a family of models that can take any combination of text, images, audio and video and turn it into video, with editing built in. You describe a change in plain English and the model reworks the clip, so you can refine an idea one instruction at a time.

The first version, Gemini Omni Flash, began rolling out the same day to Google AI Plus, Pro and Ultra subscribers in the Gemini app and Google Flow. Google also began rolling it out at no cost in YouTube Shorts and the YouTube Create app that week, and said developers and enterprise customers would get API access in the following weeks.

For a brand, a short social clip can now start as a product photo, a line of script and a few rounds of plain English feedback. That makes it realistic to test five concepts for Reels or Shorts and back the one that performs.

Images: legible text and consistent products

Image models have become much stronger at two things marketers need for finished work: text and consistency.

Google's Nano Banana 2 image model, released in February 2026, was pitched on more precise text rendering for uses such as "marketing mockups", and on keeping the likeness of up to five characters and the fidelity of up to 14 objects consistent within a single workflow. OpenAI's ChatGPT Images 2.0, launched in April 2026, focused on better text rendering, multilingual text and built in visual reasoning.

In practice this opens up:

  • Lifestyle scenes built around a real product shot, in several seasonal or regional variants.
  • Social posts and ad resizes with a readable headline in the image.
  • A recurring character or presenter who looks the same across a series of posts.

A campaign may need dozens of images that match on product, model and colour grade, and the consistency features help most here.

How platforms label AI content now

Most of the major platforms now read provenance data that travels inside the file. The open standard is C2PA, often shown to users as Content Credentials, and several AI tools write it automatically.

  • From May 2026 YouTube began using new internal signals to spot AI content. Where a creator hasn't disclosed it but YouTube's systems detect significant photorealistic AI, YouTube applies a label itself, and the label now sits in a more prominent position. Videos carrying C2PA data showing they were fully AI generated get a permanent label.
  • On 9 July 2026 Google began rolling out an AI content label setting across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Ads Editor. Google linked this to rules in the EU, India and New York that require disclosures or labels on ads with certain AI generated or edited assets, and noted that the setting on its own does not guarantee compliance with local rules. Images and video made with Google's own AI tools carry an invisible SynthID watermark and C2PA data.
  • Meta's Facebook and Instagram show an "AI info" label when they detect industry standard signals, such as C2PA or IPTC metadata, that mark content as AI generated, and people can also disclose AI use themselves.
  • TikTok reads C2PA Content Credentials so it can recognise and label AI generated content made on other platforms automatically.

If your tool writes provenance data, expect these platforms to label the file. Leave the metadata in place and use each platform's own disclosure setting.

The rules in the UK and the EU

The UK has no AI specific advertising law, and the CAP Code has no separate AI rules. In June 2026 the Advertising Standards Authority set out its position: the CAP Code is media neutral and applies in full to ads made with AI, and the advertiser remains responsible for any AI output that ends up in its ads. There is no blanket duty to disclose AI use. CAP's guidance is that disclosure is needed where leaving it out would mislead the audience.

The ASA is also clear that a label can't rescue a misleading ad. CAP's guidance gives the example of an AI generated image showing the effect of a cosmetic product that does not reflect real world results. Adding a note that AI was used does nothing to fix that. What matters is whether a customer would be misled about what they are buying.

UK brands advertising into Europe have a second set of rules. Article 50 of the EU AI Act, covering transparency for AI generated content, has applied since 2 August 2026. The EU's Digital Omnibus on AI kept that date, with a grace period to 2 December 2026 on machine readable marking for generative AI systems that were already on the market before 2 August. Under Article 50, businesses that use AI to create deepfakes must disclose that the content has been artificially generated or manipulated. In June 2026 the European Commission also published a voluntary Code of Practice on marking and labelling AI generated content, with guidance on how and where to display labels.

What to try first, and how to keep it on brand

Start with work that is frequent and low risk. Good first projects:

  1. Rough out three or four Shorts or Reels ideas from existing product photos with a tool such as Gemini Omni Flash. Treat them as tests and let the numbers pick the winner.
  2. Use AI for the background and setting of product imagery. Keep the real product photo as the anchor, and keep the product itself exactly as it is sold.
  3. Generate ad sizes, seasonal versions and headline variations from one approved master.

To keep it consistent, build a small brand kit for AI work before you scale:

  • A folder of approved reference images: products from several angles, any recurring characters, and examples of your preferred lighting and colour.
  • A shared prompt library. When a prompt produces something good, save it with the settings and the reference images used.
  • One named person who approves AI output before it goes out, checking product accuracy, text spelling and anything that looks slightly off.

For labelling, write a one page policy your team can follow without a meeting:

  • Label anything showing realistic people or scenes that were generated, and anything a customer could mistake for a real event or real result.
  • Use each platform's built in AI disclosure setting and keep provenance data in the file.
  • For campaigns shown in the EU, treat deepfake disclosure as a legal requirement.
  • Keep a simple log of which tool made which asset, and when.
  • Check claims first. A label never makes a misleading claim acceptable.

What to do next

Pick one channel and one month. Produce a batch of AI assisted content alongside your usual output, label it properly, and compare cost, turnaround and performance. That gives you real numbers to decide how far to go.

Your reference images, saved prompts and labelling policy will still be useful as the tools change.

If you'd like a hand setting up that workflow or producing the first campaign, our Creative AI team can help with this kind of work.

Sources

Read next

More from the blog.

Want to put this into practice in your business?