There are six situations where automated retouching is not a slightly worse choice but the wrong category of tool. Most of them are about accuracy, not aesthetics.
The short answer: avoid generative and automatic retouching when the photograph is a record of something, when the subject's identity is the point, when nobody asked, when the file is under an NDA, when the frame will be judged at full size, and when a set has to match. Everywhere else it is a straightforwardly good tool.
Automated retouching tools are useful and this is not an argument against them. It is a list of the situations where they are the wrong category of tool — where the issue is not "slightly worse output" but "an output that is a problem".
1. When the photograph is a record of something
Product photography where the customer receives the object. Documentation. Real estate. Anything used as evidence. Medical or dental before-and-after images.
Generative tools invent plausible detail. In a record, invented detail is a false statement. A generatively "cleaned" product shot can misrepresent what arrives in the box, and that is a commercial and sometimes legal problem rather than an aesthetic one.
The rule: if someone will compare the photograph to the thing, nothing in it may be invented.
2. When the subject's identity is the point
Automatic portrait tools normalise. They smooth toward an average, and the features they remove first are exactly the ones that make a face specific — freckles, moles, asymmetry, scars, lines.
For a portrait commission, a family photograph or a professional headshot, the subject is the point. Removing what identifies them is not a retouch, it is a substitution, and clients notice even when they cannot articulate it.
This applies with extra force to small faces that have been upscaled. Reconstructed features are plausible rather than correct, and "plausible" is not the same person.
3. When you have not been asked
Automatic tools apply their whole treatment by default. Slimming, smoothing, eye enlargement, teeth whitening — all at once, without anyone deciding.
Changing someone's body or face without being asked is a decision made about them, not for them. Ask, and the answer is very often "leave it".
4. When the file is under an NDA or a model release
Cloud-based tools upload your file. Some free ones reserve the right to keep or train on uploads.
For a client's portrait, a product under embargo, or anything with a signed release attached, that is a contractual question rather than a preference. It is not a reason to avoid these tools; it is a reason to decide once which files go where.
5. When the frame will be judged at full size
Automated skin work looks fine at 50% and gives itself away at 100%. If the frame is going to print, to a full-screen crop, or to a model's portfolio, it will be looked at at a magnification where the absence of pore structure is unmissable.
The test is simple and people skip it: open it at 100%, the next day, on a calibrated screen.
6. When the set has to match
Automatic tools work per image. Run forty frames through one and each gets its own interpretation of what that frame needed.
The result is forty individually plausible frames that do not belong together — which is the failure clients notice from across the room, before they notice anything about any single image.
Where automation is fine, for balance
Masking. Denoise. Upscaling within sensible factors. Dust removal. A first-pass cull. Batch consistency on genuinely consistent sets. Personal work. Rough previews. Anything where you will review the output before it goes anywhere.
The common thread: the output is easy to verify, and you are going to verify it.
The honest test
Before running a frame through an automated tool, ask:
Would I be comfortable telling the client exactly what was done to this image?
If the answer is yes, proceed. If it involves a subject's face and the answer is "I would rather not go into it", that is the tool telling you something.
That is the same standard we hold our own work to. Nothing we deliver has had a decision made about a person's face by software, and the reason our beauty work explicitly does not deliver completely smooth skin is that the alternative is a decision nobody made.
One question that covers all six
Before running a frame through an automated tool, ask whether you would be comfortable telling the client exactly what was done to it.
That question resolves every case on this list without needing the list. Denoise and masking survive it easily. A generatively rebuilt product surface does not. Neither does a face that has been quietly slimmed, or a portrait whose freckles were removed by something that treated them as blemishes.
It also scales to tools that do not exist yet, which a list of specific prohibitions does not.
The rule underneath it is simple enough to hold: a tool may execute a decision you made, and it may not make one about a person. Everything in this article is a consequence of that, and it is the same standard we hold our own work to.
Common questions
Is it ever wrong to use generative fill on client work? When the client can compare the image with reality — a product, packaging, a room they were in. Generative tools produce plausible detail, and plausible is the wrong answer when someone can check.
What is the problem with automatic portrait tools? They normalise toward an average, and the features they remove first are the ones that make a face specific. On a portrait commission the subject is the point.
Do I need permission to retouch someone's appearance? It is the right instinct. Slimming, smoothing and reshaping are decisions about a person, and automatic tools apply all of them by default without anyone choosing. Ask, and the answer is often "leave it".
Can I use cloud AI tools on NDA work? That is a contractual question rather than a preference. Some services reserve rights over uploads. Decide once which files may go where.
What is the single best test? Whether you would be comfortable telling the client exactly what was done. If it involves a face and the answer is "I would rather not go into it", the tool has told you something.
Related reading: AI Retouching vs Human Retouching and 10 Retouching Mistakes Clients Notice Immediately.
BeEdits Editorial Team
Post-production studio
A photography post-production company: high-end retouching, creative editing and scalable production support for photographers, studios and brands.
