Strengths, limits and when not to use AI imagery

Video lecture

Strengths, limits and when not to use AI imagery

10 chapters · about 8 min · full transcript

Coming soon

Chapter 1 of 10

When NOT to use AI imagery

Knowing when to say no is a senior skill
Where AI shines, where it misleads
A one-minute decision tree

The narrated lecture is in production

Every chapter is scripted and ready. Browse the chapters and read the full transcript now — the video will appear here when it’s published.

Chapters

Transcript of the narration, chapter by chapter.

0:00 When NOT to use AI imagery

Here's a skill nobody puts in their portfolio, but every senior designer has. Knowing when to say no. Imagine a client asks you for before-and-after photos showing their serum clearing acne in two weeks, and they want them generated with AI because it's faster. You could do it in ten minutes. Should you? Absolutely not. In this lecture, you'll learn where AI imagery shines, where it's risky or simply wrong, and a quick decision tree you can run on any brief in about a minute. By the end, you'll be able to push back on a risky brief with confidence, and offer a better alternative instead of just a no.

0:48 Why it matters

Why does this matter? Three reasons. First, trust. Audiences are getting better at spotting AI images, and a brand caught faking evidence loses credibility fast. Second, law and rules. Advertising regulators, like the Advertising Standards Authority in the UK and the Federal Trade Commission in the US, say ads must not mislead, and that applies no matter what tool made the image. Platforms also require labels for realistic synthetic content. Third, effectiveness. Sometimes AI images simply perform worse because they look generic. Think of AI imagery like a special effect in a film. Used well, it creates things that couldn't be filmed. Used as a substitute for reality, it breaks the audience's trust in the whole story.

1:39 Where AI shines

So where does AI imagery shine? First, pre-production. Moodboards, storyboards, set and prop ideas that you then shoot for real. Second, variations at scale. Seasonal backgrounds, new aspect ratios, and scene extensions built around a real product photo. Third, illustration systems. Recurring spot illustrations for a blog or an app, driven by a documented style reference. Fourth, the impossible or impractical. A floating island made of books for a reading campaign. Nobody thinks that's real, and nobody is misled. Notice the pattern. AI is strongest when it's either invisible in the final work, like a background around a real product, or obviously imaginative, where no viewer could mistake it for evidence.

2:27 The risky zone

Now the risky zone. Anything that works as evidence. Before and after results, testimonials, real customer photos, news-like scenes and documentary images of real events. Also real people without consent, real brands and trademarks, and cultural or religious subjects where accuracy and respect matter. Here's a simple test I call the authenticity test. Would a reasonable viewer rely on this image as proof that something is true? If yes, AI is the wrong tool, period. Disclosure does not fix a false claim. If the serum doesn't clear acne in two weeks, a small label saying made with AI doesn't make the ad honest. It just makes it honestly misleading.

3:15 Worked example 1: a Dubai café latte

Let's do a simple worked example. A café in Dubai wants an Instagram post for its new pistachio latte. Option one. Generate a perfect latte with AI. It looks amazing, but it isn't their latte, the foam art is not what customers will get, and the cup doesn't match. Option two. Photograph the real latte on the counter with a phone in good window light, and use AI only to extend the background for a vertical story format. The first option risks disappointed customers and complaints. The second is honest, and it still uses AI to save time. That's the hybrid mindset. Real where it's evidence. AI where it's context.

4:03 Worked example 2: a skincare campaign

Worked example two, a business scenario. Details are illustrative. Sara manages marketing for a mid-sized skincare brand selling in the UK and the UAE. The team wants a campaign with before and after photos, lifestyle images, and a hero visual for the website. Let's split it. The before and after images must be real customers, with written consent, photographed under identical lighting, with results that match what the product can actually do. Lifestyle images? A mix. Real product photography, with AI-generated backgrounds that are reviewed for accuracy. The hero visual? An abstract, clearly artistic scene of botanical shapes, generated with AI from a documented style reference. One campaign, three different decisions. That's what good judgment looks like.

4:54 Watch me do it: the decision tree

Watch me do it. I'm going to run the decision tree on three images from a real-world style brief. Image one, a customer testimonial photo. Question one, does it show a real person the audience will rely on? Yes. So it's real photography, with consent. Stop. Image two, a background for a sale banner with abstract shapes. Question one, no. Question two, would a viewer think this actually happened? No. Question three, does it need a specific brand or person? No. Question four, is it a background or concept? Yes. Green light for AI, with review. Image three, a photo of the founder at an event she didn't attend. Question two is a yes, it looks real, and it isn't. That's a hard no. One minute, three clear decisions.

5:50 Hybrid approaches

Let's talk about hybrid approaches, because the answer is rarely all AI or no AI. You can photograph the real product and generate the environment. You can shoot real people and use AI only for cleanup, like removing a distracting sign, keeping retouching honest. You can use AI to plan a shoot, generating ten set ideas before you spend money on props. And you can license stock photography for authentic human moments, then use AI to adapt the format. The rule underneath all of these is the same. Everything that the audience will treat as true, keep true. Everything else, use AI to move faster and try more ideas.

6:37 Common mistakes

Common mistakes. Using AI for testimonials or before-and-after images. Thinking a disclosure label makes a misleading image acceptable. Generating food, fashion or products that don't match what customers will receive. Depicting cultural, religious or national dress inaccurately because the prompt was vague. Using AI for something where real photography is cheap and better, like a phone photo of your actual storefront. And forgetting that AI imagery can look generic. If every competitor uses the same model defaults, your brand disappears into the same glossy look. Sometimes the most distinctive thing you can do is show something real.

7:19 Recap and try this now

Let's recap. AI imagery shines in pre-production, variations around real products, illustration systems and clearly imaginative scenes. It's the wrong tool for anything that works as evidence: results, testimonials, real customers and news-like scenes. Run the authenticity test, remember that disclosure never fixes a false claim, and use hybrid approaches to get speed without deception. Here's your try this now. List five images you're planning for a brand this month. Run each one through the four-question decision tree from the lesson, and write your decision next to it: AI, real photography, licensed stock, or hybrid. If any image lands in the red zone, rewrite that brief today.

Knowing when to say no

Part of professional judgment is knowing when AI imagery is the right choice — and when a photograph, illustration, stock image or simple graphic would serve the audience better. Using AI everywhere can save time but damage trust, especially when audiences expect authenticity.

Where AI imagery shines

Ideation and moodboards: exploring directions quickly before a shoot or illustration commission.
Concept visualization: showing a client what a campaign could look like.
Backgrounds and environments: extending a set, creating abstract textures or scenic backdrops.
Stylized illustration and fantasy: scenes impossible or expensive to photograph.
Variations and localization: adapting a visual to different seasons, settings or formats (with care).
Editing: removing distractions, extending canvases for new aspect ratios, replacing backgrounds.
Low-budget content where no real subject is being represented.

Where AI imagery is risky or inappropriate

Situation	Why it's risky	Better option
Showing real product results (skincare, fitness, food as sold)	Can mislead consumers; may breach advertising rules	Real photos of actual products and results
Testimonials and "customers"	Fake people implying real endorsements is deceptive	Real customers with consent
News, documentary, current events	Can spread misinformation	Real, verified photography
Real people (celebrities, politicians, private individuals)	Likeness, defamation, deepfake laws, platform policies	Licensed photos, consent-based shoots
Property, travel and hospitality listings	Misrepresenting what customers will get	Real photos of the property or location
Cultural or religious depictions	Inaccuracy and disrespect risks	Work with community members and real photography
Brand logos and trademarks	Trademark and similarity risks	Official assets
Highly personal brands built on authenticity	Audience trust	Your own photos and video

The authenticity test

Before using AI imagery, ask:

Would the audience reasonably believe this shows something real? If yes, and it is not real, you risk misleading them.
Does the image make a claim (about a product, result, place or person)?
Would the audience feel deceived if they learned it was AI-generated?
Is there a legal or platform rule requiring disclosure or prohibiting this use?

If any answer is concerning, use real imagery, clearly label the content, or rethink the concept.

Advertising and consumer protection

Consumer-protection and advertising rules in many markets prohibit misleading ads regardless of how the image was made. For example, UK advertising codes require that ads not mislead, and the ASA has ruled against ads where images exaggerated product effects; the US FTC treats deceptive representations — including fake reviews and endorsements — as unlawful. Similar principles exist across the Gulf and South Asia. AI does not create an exemption: if an image exaggerates what a product does, it is a problem.

Practical hybrid approaches

Often the best answer is AI plus real assets:

Real product photo, AI-generated background (clearly not claiming the product was photographed there, if that matters).
Real creator photo, AI-assisted cleanup or canvas extension.
AI-generated illustration for a concept, with real screenshots or data for claims.
AI moodboards that guide a real photoshoot.

Worked example: a skincare brand campaign

The brand wants a campaign showing "glowing skin after 4 weeks".

Rejected: AI-generated "after" faces — this would misrepresent results and could breach advertising rules.
Chosen: real customer photos with consent and representative results; AI used only to create abstract, clearly stylized background textures and to extend image canvases for vertical formats; claims supported by evidence.

Decision checklist

[ ] Does the image depict a real product, place, person or result?
[ ] Could it mislead if viewers assume it is real?
[ ] Are there platform or legal disclosure requirements?
[ ] Would real photography or licensed stock serve better?
[ ] Is a hybrid approach possible (real subject + AI support)?
[ ] Who reviews for accuracy and cultural sensitivity?

Common mistakes

Using AI "customers" in testimonial-style ads.
AI-generated property or food images that don't match reality.
Assuming disclosure fixes a misleading claim (it may not).
Using AI imagery for sensitive cultural or religious content without community input.

Hands-on: the use-AI decision tree

Run every planned image through this before you open a generator. It takes about a minute per image.

1. Does the image show a real product, result, place or person
   that the audience will rely on as evidence?
      YES → real photography (AI may assist with cleanup only). STOP.
      NO  → go to 2
2. Would a reasonable viewer assume this actually happened?
      YES → AI allowed only with clear disclosure + platform label,
            and never for testimonials, before/after or news-like content.
      NO  → go to 3
3. Does it need a specific real person, brand, artwork or landmark?
      YES → needs consent/permission, or use licensed material. Consider
            whether AI adds anything.
      NO  → go to 4
4. Is it a concept, mood, background, illustration or variation?
      YES → good AI use case. Plan references, review and provenance.

Before/after: rewriting a risky brief

	Before	After
Brief	"Generate before-and-after photos showing our serum clearing acne in two weeks"	"Photograph real customers (with written consent) at day 0 and day 14 under identical lighting; use AI only for an abstract hero background"
Risk	Misleading product claim; likely breach of advertising rules such as the UK CAP Code and FTC guidance on deceptive claims	Evidence is real and documented; AI used where it cannot mislead
Disclosure	Would not fix the problem — the claim itself is false	Standard ad disclosure; AI background may need no label, check platform rules

Where AI imagery is strongest in 2026

Pre-production: moodboards, storyboards, set and prop concepts you then shoot for real.
Variations at scale: backgrounds, seasonal themes and aspect-ratio extensions around a real product photo.
Illustration systems: recurring spot illustrations built from a documented style reference.
Impossible or impractical scenes: clearly fictional or conceptual visuals where no one could be misled.

Summary

Use AI where it adds creative value without misleading — ideation, stylized scenes, backgrounds and editing — and prefer real imagery where authenticity, product claims, real people or cultural accuracy matter. Apply the authenticity test and the decision checklist on every project.

AI imagery is strongest in pre-production, variations around real products, illustration systems and clearly imaginative scenes.
Anything that works as evidence — results, testimonials, real customers, news-like scenes — needs real, documented imagery.
Disclosure never fixes a false or misleading claim; advertising rules apply whatever tool made the image.
Hybrid approaches (real product, AI context) deliver speed without deception.
Run the four-question decision tree on every planned image before generating.

List five images you might create for a brand. For each, run the authenticity test and decision checklist, and decide whether to use AI, real photography, licensed stock, or a hybrid.

Enrol for free to save your progress

Reading is always free. Enrol to keep your place, take the final assessment and earn a verifiable certificate.

Strengths, limits and when not to use AI imagery