A retouched photo starts from a real shot and improves its rendering, while an AI-generated visual creates an image from a written description, with no original photo required. For a brand, the key difference is fidelity to the real product. Since 2 August 2026, the European AI Act also requires labelling of deepfakes.
What is a retouched photo?
A retouched photo is an image taken from a real shot (a dish, a product, a shop, a team) whose rendering has been improved afterwards. Retouching usually works on brightness, framing, colour balance, sharpness, or removing distracting elements in the background. A retouched photo keeps a direct link to reality: the customer who buys the product finds what they saw on Instagram, LinkedIn or TikTok, give or take some enhancement.
What is an AI-generated visual?
An AI-generated visual is an image created by a generative artificial intelligence model from a prompt, that is, a written instruction describing the expected scene, style, colours and format. An AI-generated visual needs no source photo: it can depict a mood, a setting, a concept or an editorial format in a matter of seconds. According to Eurostat's "Use of artificial intelligence in enterprises" (2025 data), 9.55% of European Union enterprises with 10 or more employees used image, video or sound generation technologies in 2025.
What is the difference between an AI visual and a retouched photo?
The difference between an AI visual and a retouched photo lies in the image's origin, and therefore in its degree of fidelity to reality. The table below compares the two approaches on the criteria that matter for a small or medium business.
| Criterion | Retouched photo | AI-generated visual |
|---|---|---|
| Starting point | Real shot | Written description (prompt) |
| Fidelity to the real product | High if retouching stays light | Must always be checked |
| Needs a photoshoot | Yes | No |
| Suitable uses | Products, dishes, venues, team | Moods, illustrations, editorial formats |
| Main risk | Excessive retouching that misleads the customer | Depicting a product that does not exist |
The practical conclusion is simple: the more a visual shows what the customer will actually buy, consume or visit, the more a retouched photo is the right option. The more a visual illustrates an idea, a season or a message, the more AI generation becomes relevant.
When should a small business prefer a retouched photo?
A small business should prefer a retouched photo for any content showing a product, a dish, a venue or a person that the customer will actually see. A retouched photo guarantees that what is shown matches what is sold: a restaurant publishing its dish of the day, a craftsperson showing a finished job, a fashion shop presenting an item in stock. Light retouching (brightness, framing, colours) improves the rendering without betraying reality, whereas retouching that changes a product's size, texture or composition raises the same problem as a fully invented image.
When is an AI-generated visual suitable?
An AI-generated visual suits illustrative content that makes no precise promise about a product. The most common use cases for a small business are the following:
- Mood and seasonal visuals (back to school, holidays, sales).
- Backgrounds and settings to stage a message.
- Editorial formats: quotes, tips, event announcements, simple infographics.
- Variations of the same message across several formats for several social networks.
An AI-generated visual then allows a brand to quickly produce varied content without booking a photographer for every post. It becomes a problem only when it is used to depict a real product in a misleading way.
Must an AI-generated visual be disclosed?
Whether an AI-generated visual must be disclosed depends on its nature, and this article does not constitute legal advice. According to the European Commission (31 July 2026), the AI Act's transparency rules have applied since 2 August 2026: deepfakes must be labelled, and content generated or modified by AI must carry machine-readable markings. The Commission also states that more than 180 organisations have signed the code of practice on transparency of AI-generated content. For a specific case, particularly a visual depicting a real person or product, legal advice remains recommended.
What is the common point of caution for both approaches?
The common point of caution is fidelity to the real offer: neither a retouched photo nor an AI-generated visual should ever suggest a feature, size or quality the product does not have. This is why a human validation workflow before publication remains necessary in both cases. In Isalis, every image, short video or text generated from the brand DNA goes through a configurable validation circuit before being published on social media.
What I often see while building Isalis is that brands getting the best results do not pick a single camp: they reserve real photography for their products and use AI to multiply the mood visuals and editorial formats around them.
Why does this question concern more and more small businesses?
This question concerns more and more small businesses because AI use is growing fast among French small firms. According to the France Num 2025 barometer presented by the French Directorate General for Enterprises, 26% of French micro, small and medium businesses use artificial intelligence solutions, twice the share recorded in 2024. The more generated visuals become commonplace, the more the ability to clearly distinguish illustration from product representation becomes a trust marker for customers.
In short, 3 questions
What is the main difference? A retouched photo starts from a real shot, an AI-generated visual starts from a written description.
Which one should be chosen to show a product? A lightly retouched photo remains the reference for anything the customer will actually buy or consume.
Must an AI visual be disclosed? Since 2 August 2026, deepfakes must be labelled according to the European Commission; for a specific case, seek legal advice.
So, do you reserve real photography for your products and AI for your mood visuals, or do you mix the two?
To organise the review of your visuals before publication, read our FAQ on setting up an AI validation workflow.
Sources
- Eurostat, Use of artificial intelligence in enterprises, 2025 data extracted in December 2025.
- European Commission, Commission starts enforcing AI Act rules and new transparency requirements on 2 August, 31 July 2026.
- French Directorate General for Enterprises, France Num 2025 barometer: French micro, small and medium businesses confirm their commitment, September 2025.