7 questions to ask before trusting AI with your listing photos

Every AI editing tool has the same landing page: stunning before-and-afters, a speed claim and a price. The real differences, the ones that decide whether the tool survives contact with a Monday morning of client campaigns, only show up under questioning. These are the seven questions we would put to any provider. Our own answers are on the record, so hold us to them too.

1. Is it generative or deterministic?

The single most important question, and most providers hope you will not ask it. If the AI generates pixels, some part of every delivered image is invented, and you carry the misrepresentation risk on every listing. If it is deterministic, the same input produces the same output and every pixel traces to your camera. Ask the question in exactly those words and watch how straight the answer is.

2. What was it trained on?

An AI editor is only as good as the editing it learned from. Models trained on scraped web imagery learned what average looks like. Models trained on professional production output learned what shippable looks like. LUXengine's models trained on more than 100,000 RAW inputs paired with the finished edits a production house delivered to paying clients over nearly a decade. Ask any provider this question and insist on a number.

3. Do I get full resolution back?

Some pipelines quietly downscale, process, then upscale. It looks fine on a phone and falls apart on a printed window card or a cropped hero shot. Your camera captured the detail. The tool should return all of it, at the resolution you shot.

4. What happens when it gets one wrong?

Every system produces a miss sometimes. The question is what the workflow looks like. Is there a review step before anything reaches your client? Can you flag and fix fast? Does the fix teach the system, so the same miss stops recurring? A provider with no good answer here is a provider who has decided your client relationship is the quality-control step.

5. What does it actually cost per delivered image?

Not the headline rate. The delivered rate, including redos, rush turnaround and minimums. AI pricing should be simple enough to say in one breath. Ours is: plans include a pooled monthly quota, one video counts as 50 images, top-ups are 10 cents an image and 5 dollars a video, and there are no rush fees because everything is minutes anyway.

6. Does it fit how a photography business actually runs?

Editing is one link in a chain that includes booking, shooting, delivery and getting paid. If the AI tool is a silo, you have added a subscription, not removed work. The platforms worth switching to fold editing into the workflow: jobs in, edits reviewed, campaigns delivered through a branded portal, invoice raised, all in one place.

7. Who built it, and have they ever delivered a campaign?

Tools built by people who have never sat in a production seat optimise for demos. Tools built by people who have processed ten thousand jobs optimise for the boring things that keep clients: consistency, deadlines, colour that matches last month. We built LUXengine because we run one of the largest real estate photography networks in Australia and needed it ourselves, 19 years into the industry. That is not a guarantee of perfection. It is a guarantee that the tool was shaped by consequences.

Ask all seven. Any provider worth your campaigns will enjoy answering them. And if one of the answers makes you flinch, trust the flinch. Your name is on every image.