Yes, as long as each photo still shows what a guest will find. Fixing light, colour, tilt and temporary clutter is fair. Adding a view, enlarging a room, repairing worn finishes or inventing an amenity is misleading and breaks platform rules. Ask whether a guest arriving would feel the room doesn't match.
AI is not the line. A traditional retouch that removes a stained carpet misleads exactly as much as an AI one, and an AI colour correction misleads nobody. Platforms and advertising rules look at the result: does the photo change something a guest weighs when booking, such as room size, view, condition or facilities? AI makes misleading edits cheaper and harder to spot, which is why platforms have begun naming it in their policies.
A two-question test for any edit
Before you accept or publish an edited hotel photo, ask two things.
Will the guest find this on arrival? This is the permanence question. Removing a phone charger, a bin liner or a guest's forgotten towel is fair, because nobody expects those to be there. Removing a power line, the car park, a radiator or a worn carpet is not, because the guest will find all of them.
Would the guest decide differently if they knew? This is the materiality question. A guest choosing between two rooms weighs the view, the size, the bed, the bathroom and the facilities. If the edit touches any of those, it is misleading even when the object technically exists somewhere in the building.
A third rule of thumb helps with grey cases: could a competent photographer with a normal lens and decent light have taken this picture? If yes, an AI tool that gets you there faster is fine. If only a generator can produce it, the image is an illustration, not a photograph of your room.
Fair, grey and misleading: a verdict for 14 common edits
The table gives my call on edits that hotels actually ask AI tools for. "Grey" means I would avoid it or fix it at source, not that it is banned.
| Edit | Verdict | Why | Safer route |
|---|---|---|---|
| Brighten, balance colour, lift shadows | Fair | Corrects what the camera missed | Avoid halos and glowing windows |
| Straighten verticals, fix lens distortion | Fair | Shows the room as a person sees it | None needed |
| Remove a cable, bin, stray bag or personal items | Fair | Temporary items staff would clear anyway | Tidy before shooting where you can |
| Blur or remove strangers from lobby or pool shots | Fair | Protects privacy of identifiable people | Shoot when empty or get consent |
| Sharpen, reduce noise, blur the background | Fair | Technical clean-up, no content change | Keep it light so surfaces stay real |
| Replace a grey sky with a blue one | Grey | Weather changes, but it sells a climate you cannot promise | Reshoot on a clear day |
| Add a sunset glow or golden-hour look | Grey | Only honest if the view really faces that way | Photograph at that hour |
| Add a sea, mountain or city view in the window | Misleading | The view is a main reason to pay more | Photograph the real view; call it partial if it is |
| Widen or extend the room, exaggerate wide-angle | Misleading | Room size is a main characteristic | Use a fair lens and state the floor area |
| Add furniture, art or plants (virtual staging) | Misleading | The guest gets what is in the room | Dress the real room |
| Change bed size, linen or pillow count | Misleading | Beds and linen are material to choice | Make up the real bed |
| Remove a stain, scuff, damp or worn carpet | Misleading | Condition is part of what you sell | Repair or replace, then reshoot |
| Remove a power line, neighbouring building or current scaffolding | Misleading | Guests will see it; a current building site should be disclosed | Pick another angle, or say works are under way |
| Generate a pool, gym, terrace or breakfast spread | Misleading | It invents a facility or its scale | Photograph what you serve and offer |
Two further cases sit outside the table. Generated people, such as AI guests relaxing by the pool or avatars in place of staff, are a poor idea: Google's guidance on business photos suggests showing your real team, and an invented crowd implies an atmosphere you cannot guarantee. And "enhance" buttons that apply a style across the whole photo deserve the same suspicion as a heavy filter: you cannot tell from the button what they changed.
Uses of AI that never touch what the photo shows
The safest AI help with hotel photography leaves the pixels alone. Use it to cull a shoot of 400 frames down to the best three per room, to crop each image to the ratios a channel asks for (cropping is fair so long as the crop does not hide a flaw), to rename files sensibly, to order a gallery so the strongest honest picture comes first, and to draft alt text and captions from facts you supply. A prompt such as "Write alt text for this photo of room 12 in one short sentence: a double bed, a desk, a window facing the courtyard" gives you something usable, and the facts come from you rather than from the tool's guess about what the picture shows. Check each description against the photo, because tools describe what they expect to see, and "sea view" is a common invention. The alt-text tutorial has the checking routine.
What Google, Expedia and the AI Act actually say
Google Business Profile. Google's page on business photos says a photo should be in focus, well lit and have "no significant alterations or excessive use of filters or AI. In other words, the image should represent reality." For hotels it adds that guests rely heavily on room photos and that yours should truthfully show what it is like to be in the room as a guest. The page is Tips for business-specific photos on your Business Profile, and it is the one to read before you upload anything there.
Google's hotel results. Google's platform policy for partners that feed its hotel listings says images must reflect the honest, real-world condition of the property and, in its words, "strictly prohibits" AI-generated imagery or visual assets that have been digitally modified, falsified or heavily edited to distort room dimensions, scenic views or property conditions. A flagged image is blocked from showing until it is replaced or a successful appeal clears it. The rule binds the aggregators that send hotel data to Google, so if an online travel agency or similar partner passes your photos on, your edits are judged by it.
Expedia Group. The partner provided content policy covers any text, image or video partners upload, including AI-generated content. It prohibits content that is fake, misleading or exaggerates or omits material details, and specifically content that uses generative AI or AI editing tools to mislead, misrepresent or fabricate, including to misrepresent a property. It also asks partners to disclose significant known disruptions, with ongoing construction as its example, and to update content promptly when conditions change.
Booking.com and other channels. I could not find an AI-specific photo rule on the public Booking.com pages I could read. Its partner pages carry photo requirements, so read the current version in your own account before uploading, and assume the same honesty standard applies.
Meta ads. If you promote rooms with paid ads, Meta's ads help says it labels ads it detects were made with third-party AI tools. A label does not rescue a misleading picture; it only tells viewers the image was AI-made.
The EU AI Act. If you market to guests in the EU, Article 50 has applied since 2 August 2026. It requires anyone who uses an AI system to generate or manipulate image content that is a "deep fake" to disclose that it was artificially generated or manipulated. The Act defines a deep fake as AI-generated or manipulated content that resembles existing persons, objects, places or events and would falsely appear authentic. A doctored picture of your real pool or lobby that passes as genuine matches that wording, and it is the same question as the guest test above. How the Act treats ordinary corrections is a matter for your adviser.
How common is all this? According to travel trade-press reports, a May 2026 analysis of 25,550 photos from 700 hotels on major booking platforms, run by a marketing agency with a forensic AI-verification firm, flagged about 19% as showing at least one sign of AI generation or editing. A photo was flagged if any single signal turned up, from suspicious metadata to irregular pixels, and the study did not separate fully generated images from enhanced ones, so read the figure as a sign of how widespread AI use is, not of how many photos mislead. The coverage also described the distortions worth auditing for: rooms made to look larger, furniture made to look newer, gardens made lusher and pools made bigger. The wider legal picture for marketing images is in this tutorial on using AI images legally.
Why a photo counts as a promise
Advertising and consumer-protection rules generally treat a hotel photo as a claim about what the guest will get. The question those rules ask is whether the overall impression is likely to mislead an ordinary customer about something that matters to the purchase. That is why the weak defences do not work: "everyone retouches", "it is only a little sky" and "the description says otherwise" do not change what the picture itself communicates. I would treat a caption as an addition to accuracy, never as a substitute for it. For anything contractual, such as a dispute about a refund, ask a solicitor.
The practical consequences are mundane and costly: guests asking for partial refunds, chargebacks, reviews that quote your photo back at you, and a channel asking you to remove content. None of those needs a regulator; they only need one guest with a phone standing in the wrong room.
Where the disclosure line sits, with wording to copy
Disclosure is for things a guest could mistake for a photograph of your property. It does not belong on every corrected image, and it cannot rescue an edit that changes what the guest receives. The table sets out four tiers.
| Tier | Example | Label? | Wording you could use |
|---|---|---|---|
| A. Corrections | Exposure, colour, straightening, sharpening | No label on each photo; log it | Optional gallery note: "Photos are colour-corrected and tidied. Rooms are shown as they are." |
| B. Small clean-up | Removing a cable or bin with a generative tool | Log it; mention in the photo policy page | "We remove temporary items such as cables and bins before publishing." |
| C. Illustration | Concept render of a planned refurbishment, event graphic, seasonal banner | Yes, on the image | "Illustration created with AI. Not a photograph of the hotel." |
| D. Changes what the guest gets | Added view, furniture, pool, bigger room | Do not publish | Reshoot; a label does not fix it |
Keep illustrations out of room galleries and channel listings altogether, and write alt text that describes what the real room contains. The mechanics of labelling AI images on social channels are covered in this labelling tutorial.
A photo edit log that settles a dispute
Keep every original untouched in a folder nobody edits, and record each change. A log of five minutes per room is the difference between "we only brightened it" and "trust us". This one is filled in for six illustrative photos.
| File | Edits applied | Tool | Tier | Approved |
|---|---|---|---|---|
| room-12-main | Exposure +0.6, white balance, verticals | Photo editor | A | Duty manager, 4 Nov |
| room-12-bath | Exposure, towel rail cable removed | Generative remove | B | Duty manager, 4 Nov |
| lobby-02 | Two strangers blurred | Photo editor | A | Owner, 5 Nov |
| terrace-05 | Crane removed, then reverted to the original | Generative remove | D then A | Owner, 5 Nov |
| breakfast-03 | Colour correction only | Photo editor | A | Duty manager, 5 Nov |
| spring-offer-hero | Fully generated banner | Image generator | C | Owner, 6 Nov (labelled) |
The log needs one approver who is not the person who edited. That single habit catches most overreach, because the person who has been looking at the room for twenty minutes is least likely to notice the sky has changed.
Photographers, contracts and photos that age
Many hotel photographers now run generative clean-up tools as routine, so the brief is the place to say what you expect. Ask three things before you book: whether they use generative fill or object removal, whether they will list every generative edit they make, and whether they will hand over the untouched originals with the finished files. One paragraph in the brief covers it, and a solicitor can tidy the wording if you want it in the contract. Illustrative wording, not legal advice: "Retouching is limited to exposure, colour, straightening and removal of temporary items. No generative additions, replacements or alterations to rooms, views, furniture, surfaces or facilities. The photographer will supply untouched originals and a list of any removals."
Photos also age. A picture that was honest on the day it was taken can turn misleading after a refurbishment, a change of beds, or a new building next door, and Expedia's policy asks partners to update content promptly when material features, conditions or amenities change. Put the shoot date in the log, set a reminder twice a year to compare each room with its photos, and reshoot any room after refurbishment. It is a cheaper habit than answering the guest who holds up an old photo at the front desk.
Auditing 90 photos in a 24-room hotel, with the counts
Say a 24-room hotel has 90 images online: 72 room photos (three per room) and 18 of the lobby, breakfast room, garden and exterior. The owner sorts all 90 by the two-question test and the table above, spending about 100 seconds on each, which comes to 2.5 hours. The counts, invented for illustration:
- 71 fair. Exposure, colour, straightening and small tidy-ups of temporary items.
- 13 grey. Six sky replacements, four generative clean-ups that also removed permanent items (a wall heater in one, a smoke alarm in another), and three heavy HDR looks with glowing, unnatural windows.
- 6 misleading. Two virtually staged rooms (an added armchair and desk in one, a rug and artwork in the other), one window in room 21 swapped for a sea view when it faces the car park, a crane removed from the terrace shot while building works continue next door, a bathroom where limescale and a cracked tile had been "cleaned", and a pool photo with ten generated loungers where the hotel has four.
71 plus 13 plus 6 is 90. The fix list is short. The six misleading photos are reshot, which takes about half a day. The 13 grey ones are re-edited from the original files in about an hour, because the hotel kept the originals; without them they would have been reshot too. All 90 get a row in the log, and the corrected photos are swapped on the website and each channel in one afternoon. Without the audit, the hotel would have carried six photos that arriving guests could contradict at a glance.
How to keep an AI editor on a short lead
If you use a generative tool for tidy-ups, tell it exactly what it may change and what it must not. A prompt like this works in most editors that accept text instructions:
Edit this hotel room photo for a booking website.
Allowed: brighten the exposure, correct the white balance, straighten the verticals, remove the phone charger cable on the desk and the waste bin by the door.
Not allowed: changing the furniture, bed linen, walls, floor, ceiling, windows, view, room size or proportions. Do not add or remove anything else.
Keep the same framing and camera angle. Return one image.Then check the result against the original rather than trusting it. In one illustrative run the tool removed the cable and the bin, but it also smoothed a crease in the bedspread and merged two bedside lamps into one. Neither change is dramatic, and both would have gone out unnoticed if nobody had looked. A five-point check catches most of it:
- Count the objects: pillows, chairs, lamps, glasses, bottles, sockets.
- Compare windows and the view outside, including frames and anything beyond the glass.
- Look at reflections in mirrors, screens and glass, which generators often get wrong.
- Read any text, including signs, labels and fire-exit markings.
- Flip between original and edit at 200% zoom, looking at straight edges, skirting boards and door handles for warping.
The same habit, applied to any generated image, is in the tutorial on checking AI images for errors. Write your room descriptions with the same care, because the caption and the picture are read together; this tutorial on room descriptions covers the text side.
A realistic slip: the sea view that was a car park
A misleading edit usually shows up like this (an illustration). A hotel replaces the window view in room 21 with a sea view for its website, because three other rooms do have one. Over a month, three guests who booked room 21 for its view arrive, see the car park, and ask for money back. If each is refunded one night at $140, that is $420 in partial refunds, plus a review quoting the photo and a note from a booking channel asking the hotel to check its images. The fix cost nothing but a reshoot: replace the picture with the real, partial view, add a caption saying "Courtyard and car park view; sea views are in rooms 4, 9 and 14", and reply to the review with the corrected image link. The hotel now sells room 21 at a lower rate than the sea-view rooms instead. How to word replies like that is in this tutorial on replying to hotel reviews.
If you only change one thing after reading this, start the log. It takes minutes, it forces the two questions to be asked, and it protects you whichever platform rule changes next.
Hotel photo editing questions owners ask
Do I have to label every retouched hotel photo?
Ordinary corrections such as exposure, colour and straightening don't normally need a label on each picture, because the platform rules I read target AI used to mislead or misrepresent a property. Keep the originals and a short edit log, and label anything that is illustrative rather than a photograph of your hotel. If you market to guests in the EU, ask an adviser how the AI Act's rules on manipulated images apply to you.
Can I remove scaffolding or a crane from a photo?
Not while it is there. Expedia's partner policy asks you to disclose temporary disruption such as ongoing construction and to update content when conditions change, so editing out a current building site hides what guests will see. Reshoot when the work ends, or keep the honest photo and say in the description that works are under way and when they should finish.
Is AI virtual staging acceptable for a hotel room?
Not for rooms you let as they stand. Adding furniture, art or equipment that is not in the room changes what the guest receives, and Google's hotel policy bans AI-generated or heavily edited images that distort room dimensions, views or property condition. Staging is reasonable for a clearly labelled concept image of a planned refurbishment, kept out of your room galleries.
What should I do if a guest says the photos do not match?
Treat it as evidence. Compare their room with your original file and your edit log, replace the photo quickly if it overstates anything, and reply politely without arguing. Offer what you would offer for any gap between promise and room. The log is what lets you show, calmly, that you only changed light and tidied temporary items.
Further reads
- AI Virtual Staging for Property Listings: Costs and Disclosure — Costs and disclosure when furniture is added digitally.
- How Independent Hotels Win More Direct Bookings With AI — Honest photos are one part of winning direct bookings.
- Hotel Guest Feedback: Turn Post-Stay Surveys Into Fixes With AI — Find out whether guests feel the rooms match what you show.
- Can a Restaurant Use AI Food Photography on Menus and Delivery Apps? — The same fair-edit question for dishes and menus.
- Jewellery Photography With AI: Clean Backgrounds, True to Life — Keeping colour true when you tidy backgrounds with AI.
- Should You Tell Customers When You Use AI in Your Marketing? — How to decide what to tell customers about AI in your marketing.
- AI Visibility for Hotels: Get Recommended in ChatGPT Answers — Free checks and a fix list for a small hotel: what ChatGPT and Google AI say about you today, where they get it, and what you can and cannot change.
- AI-Generated Marketing Photos: A Cafe Test of What Looks Honest — One brunch dish photographed three ways, scored by customers and checked against platform and advertising rules, to show where AI photos stay honest.
- What Software Do Hotels Use? PMS, Channel Managers and AI Add-Ons — A map of hotel software by job (bookings, rooms and payments), how the systems connect, and the few places where AI earns its keep.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Business Profile help (photo tips and photo policy pages); Google for Developers hotel aggregator platform policies; Expedia Group partner provided content policy; Meta ads help on AI labels; text of Article 50 of the EU AI Act; travel trade-press coverage (June 2026) of a May 2026 analysis of 25,550 hotel photos on booking platforms.