Using AI-Generated Images and Video in Marketing Honestly

Why honesty about AI images matters more than the images themselves

A restaurant generates a glowing AI photo of a dish that does not quite look like what leaves the kitchen. A hotel generates a drone-style video of a pool deck that does not yet have the loungers shown in the clip. A shop uses an AI voice reading a script in an announcer's exact cadence without ever asking the announcer. Each of these saves time and looks professional, and each one quietly breaks the one thing marketing depends on: a customer believing what they are shown is close enough to what they will get. AI images and video are a legitimate, useful tool for a small marketing team with no studio and no budget for a shoot. The rules below are not about whether to use them — they are about using them in a way that keeps customer trust intact once the picture is clicked or the video is skipped past.

Treat these as a sequence, the same way a kitchen treats food safety: skip one link and the rest stop protecting you. Labelling a video as AI-generated does little good if the caption still claims it shows "our actual branch." Getting a model's consent for one photo shoot does not cover reusing their face in a different ad next year. Walk through all six steps before anything goes live, then keep the short checklist near whoever approves posts.

Step 1: label it as AI-generated, every time

If an image or video was generated or substantially altered by AI, say so in the post, caption or on the asset itself — plainly, in the same language as the rest of the post, not buried in a bio or a terms page nobody reads. This applies whether the whole image is synthetic or only the background, lighting or a person in it was AI-altered. A short, visible note is enough: something like "AI-generated image" or "created with AI" placed where a viewer actually sees it before they decide to trust what's shown. Do not rely on a platform's own AI-content tagging to do this for you — those tags vary, get missed, and are not a substitute for your own caption. The test is simple: could a reasonable customer tell, from what you published, that this is a generated image and not a photograph of the real thing? If the answer depends on reading fine print, the label is not working.

Step 2: never show AI output as the real product, venue, person or result

This is the rule that protects the business, not just the customer. An AI-generated image may illustrate a concept, a mood, a "coming soon" idea, or a style direction — but it must never stand in for an actual photo of the actual product on the shelf, the actual room a guest will sleep in, or an actual before-and-after result a service produced. If a customer books a room because a video made it look larger or better-furnished than it is, that gap becomes a complaint, a refund, or a public review — and it was entirely avoidable. The same applies to results: an AI-generated "after" image of a repaired engine, a landscaped garden, or a renovated office sets an expectation the real work must then match exactly, and usually can't. When in doubt, ask: would we be comfortable if the customer stood next to the real thing holding this image? If the honest answer is "it would look different," either use a real photo, make the AI origin unmistakable, or don't publish it.

Step 3: get consent before any real face or voice becomes reference material

AI tools can generate a face that looks like a specific real person, or a voice that sounds like a specific real announcer, presenter or staff member, especially when a real photo or recording was fed in as a reference. Before doing this with anyone — a staff member, a model, a customer who agreed to a testimonial shoot, or anyone recognisable — get their clear agreement first, in writing where possible, covering what the generated version will be used for and where it will appear. This matters even for a staff member who is "just helping out" for an internal video; their likeness being turned into an AI-altered version is still a use of their image, not a neutral editing step. It especially matters for anyone outside the company, where there is no employment relationship to fall back on. Consent given for one piece of content does not automatically extend to a different ad, a different year, or reuse by a different team — if the use changes meaningfully, ask again.

Step 4: check the rights of every piece of reference material

Many AI image and video tools let you upload a reference photo, logo, product shot or sample clip to guide the result. Before uploading anything, confirm your organisation actually has the right to use that specific file as a basis for new content — not just the right to look at it. A photo taken by a freelance photographer, a product shot sent by a supplier, or a clip downloaded from somewhere online often comes with limits on how it can be reused or altered, even inside the same company. If nobody can say where a reference image came from or what it was licensed for, treat it as unusable until that is confirmed, rather than assuming it is fine because it already existed on a shared drive.

Step 5: keep the originals and the prompts

For anything that gets published, keep a simple record: the original reference files used (if any), the prompt or instructions given to the tool, the date, who approved it, and where it was posted. This is not bureaucracy for its own sake — it is how your team answers a question like "did we actually have rights to that photo?" or "who approved showing that result?" six months later, without guessing. It also makes it far easier to update or retract content if something turns out to be wrong, and to show, if ever challenged, that the generated content was handled carefully rather than carelessly. A shared folder with a consistent naming pattern is enough for a small team; it does not need special software.

Step 6: have a person review before anything goes out

No AI-generated image or video should be posted without a staff member looking at it first, specifically checking for the issues above: is it labelled, does it avoid claiming to be the real product or result, was consent handled if a real person is involved, and are the reference rights and records in order? This person should be someone other than whoever generated the content, even if that just means a second staff member glancing at it before the post button is pressed. A generated image can look polished and still contain a problem invisible to the person who made it — a sign in the background that implies a false claim, a face that resembles a real customer by coincidence, or a detail that misleads about size, colour or condition. One short review step catches most of these before a customer ever sees them.

A short checklist before you post

1. Is this image or video AI-generated or AI-altered, even partly?
2. If yes, is there a visible "AI-generated" label on the post, not just in a bio?
3. Does the caption avoid implying this is a real photo of our actual product, venue, person or result?
4. If a real person's face or voice was used as reference, do we have their clear agreement, specific to this use?
5. Do we know where every reference file came from, and that we have the right to use it this way?
6. Have we kept the original reference files, the prompt, and who approved this?
7. Has someone other than the creator reviewed it against this checklist?

Common mistakes we see

1. Treating "AI-assisted" as not needing a label. Touching up or extending a real photo with AI still counts; label it.
2. Putting the label somewhere a customer won't see it, such as only in a watermark corner or a linked page.
3. Reusing a staff member's AI-generated likeness in a new campaign without checking whether the original agreement covered it.
4. Generating a "finished result" image before the real work is done, then publishing it as if it already happened.
5. Losing track of which reference photos were licensed for reuse, especially ones a previous staff member downloaded or received from a supplier.

How we can help

If your team wants to use AI images and video for marketing without this becoming a trust problem, our AI content studio service builds these steps into the workflow from the start: clear labelling by default, a review step before anything posts, and a simple record of references, prompts and approvals your team can actually keep up with. It pairs well with a short AI staff training session so everyone creating content, not just one designer, understands where the line sits.