Using AI-Edited Photos in Financial Marketing

The right question is not which tool, it is what changed

Firms tend to start by asking whether AI editing is allowed, which is not a useful question. A tool is not the risk; the change is. Correcting the exposure on a team photo with a slider and correcting it with an AI model produce the same image and the same negligible risk. Removing a person from a photograph of a property is a substantive change whether it took twenty minutes of manual work or one prompt.

So the policy should be written in terms of what an edit does to the meaning of the image. Does a reasonable viewer now understand something different about the world from what the camera recorded? If the answer is no, it is presentation. If the answer is yes, or might be, it needs the same scrutiny as a sentence making the equivalent claim.

The reason AI editing feels newly urgent is not that it enables new categories of change, but that it makes substantive changes as easy as cosmetic ones. The friction that used to keep people from casually reinventing a photograph is gone, and the policy has to replace it.

Where it is genuinely low-risk

There is a large category of everyday work where AI editing raises no meaningful issue. Correcting exposure and white balance on a photograph taken under mixed office lighting. Cleaning or replacing a cluttered background behind a portrait. Matching the look of team portraits shot in different homes so a distributed team appears as one company. Straightening, cropping and resizing for different placements. Removing a lens smudge or a piece of lint.

What these have in common is that nothing depicted changes in a way anyone would rely on. The person is who they are, the office is the office, and no claim is being made about a product, an outcome or an asset. This is presentation, and it is the same work firms have been paying photographers to do for decades.

For this tier, a tool like Foxi AI is a sensible fit. You upload a photo you already have and either apply a named look or describe the change in words, and it edits that photograph rather than inventing a scene. It is well suited to exactly this: office and team imagery, event photos, portrait backgrounds and lighting consistency across a set. Keep the originals, note what was done, and this tier can safely operate without a case-by-case approval.

• Exposure, brightness and contrast correction
• White balance and colour correction on mixed lighting
• Background cleanup or replacement behind a portrait
• Matching lighting and tone across a set of team photos
• Cropping, straightening and resizing for a placement
• Removing transient blemishes such as lint, dust or a lens smudge

Where it is high-risk and should never be casual

The high-risk tier is defined by subject rather than technique. Anything depicting a client, a property, a claim, a result or an asset. A photograph in an investment offering. A before-and-after. An image of an insured loss. A picture accompanying a testimonial. A screenshot of a product. Anything where a viewer's understanding of a real thing is affected by what the image shows.

In these contexts, edits that would be unremarkable elsewhere become substantive. Replacing a grey sky over an investment property changes the aspect and the impression of the site. Brightening a client's home into something more prosperous changes what the image says about them. Removing an object from a photograph of a loss changes the evidence. Smoothing a person's appearance in a before-and-after changes the depicted outcome.

Claims and inspection photography deserves a separate, absolute line. Those are evidentiary images and content-altering edits do not belong anywhere near them, regardless of intent. Keep that workflow entirely separate from marketing tooling, and take the specifics of what is permitted, including whether even brightness correction is acceptable and how it must be documented, from your own legal and compliance function.

• Any image of a client, or of a client's home, business or possessions
• Property and asset photography used in an offering or a listing
• Before-and-after imagery of any kind
• Claims, loss and inspection photographs
• Images paired with a testimonial, case study or outcome statement
• Product screenshots and anything depicting the service working
• Any image where the edit changes what a viewer concludes about a real thing

The disclosure question, which is genuinely unsettled

Should you tell people an image was AI-edited? At the moment there is no single answer that holds everywhere. Some jurisdictions and sectors are developing disclosure expectations around synthetic or altered media, some advertising platforms impose their own labelling requirements independent of law, and financial promotion rules add another layer that varies by regulator and licence. The position is moving.

That uncertainty is an argument for internal record-keeping rather than for guessing at a public standard. If you always know what was changed, you can add disclosure later, answer a question quickly, or comply with a new platform requirement without archaeology. If you do not, a future requirement becomes a project.

Take the specific question to your compliance officer and check your regulator's current guidance directly, along with the current terms of any advertising platform you use. Ask specifically whether the answer differs between an editorial post, a paid advertisement, and material that forms part of a financial promotion, because in several regimes it does.

A two-tier internal policy you can adapt

The workable structure is simple. Tier one is presentation editing on non-client, non-asset imagery, permitted without individual approval, subject to keeping the original and noting what was done. Tier two is everything else, requiring named approval before publication, with the edit described in writing and the original retained.

Write the tiers by subject and by change, and give people examples rather than principles, because principles get interpreted generously under deadline. Name who approves tier two, name where originals are stored, and state the one rule that covers the gaps: if you are unsure which tier something is in, it is tier two.

Then make it easy to follow. Put the storage location in the same place people already work, keep the note requirement to one line, and make approval a same-day process. A policy that adds a week to a social post gets bypassed, and a bypassed policy is worse than none because it creates a false record of compliance.

• Tier one: presentation edits on team, office and event imagery. No individual approval. Keep the original and a one-line note.
• Tier two: anything depicting a client, property, asset, claim, result or product. Named approval before publication, edit described in writing, original retained.
• Never: content-altering edits to claims, inspection or evidentiary photography.
• Default rule: if unsure, treat it as tier two.

Training the people who actually post

The person most likely to breach an image policy is not a marketing specialist. It is an adviser posting from their phone, a branch manager sharing a community event, or a new joiner with initiative. Phone editing tools now offer object removal and generative fill by default, and someone tidying up a photo before posting will not think of it as an edit at all.

So the training has to be about recognising a moment rather than remembering a rule. The practical message is short: if you are removing or adding anything in a picture of a client, a property, a claim or a result, stop and ask. Everything else, keep the original and say what you did. Give real examples from your own firm, because abstract policy statements do not survive contact with a phone at an event.

Review it periodically. Tools change, defaults change, and an edit that required deliberate effort last year may now be applied automatically by a phone's camera app before anyone opens an editor at all.

Frequently asked