Travel is the one stock category where the most valuable word in your keyword set is something no image recognition model can read off the pixels: the name of the place. A buyer looking for a hotel campaign types a city, a season and a feeling in the same query, and a file described only as mountain sunset beautiful landscape never enters that search. Getting travel metadata right is mostly about supplying geography the machine cannot know, then layering the commercial concepts around it.
Why travel files behave differently from other categories
In most categories the picture describes itself. A bowl of soup is a bowl of soup wherever it was shot, and an AI tagger can get most of the way there on visual evidence alone. Travel breaks that assumption twice over.
First, the buyer's query is usually a proper noun. Tourism boards, airlines, travel publishers and hotel groups search for a destination by name, then narrow by season or activity. If your file does not carry the place name, the country and the region, it is invisible to the people with the largest budgets in the category.
Second, travel photography runs into more rights questions than almost any other subject. Recognisable people, trademarked architecture, branded signage and protected sites all live in the frame at once, and how you describe the image affects whether it can be licensed commercially or only editorially.
So travel keywording has two jobs that pull in different directions: be as specific as possible about where this is, and be honest about what the image can legally be used for.
Settle the geography before you write anything else
Place names are the only part of a travel keyword set where you can be factually wrong, and agencies treat that as a quality problem rather than a style preference. A cathedral labelled with the wrong city, a mountain attributed to the wrong range, a beach given the wrong island: all of these can get a file pulled after a buyer complains, and repeated errors damage the account rather than the single image.
Write the hierarchy from narrow to broad, and put it in the caption as well as the keywords. Landmark or district, then city or town, then region or state, then country, then continent. Buyers search at every level of that ladder, and a file that only carries the country name loses the most valuable searches to files that carry the street.
Include the local spelling alongside the English one where both are in genuine use, and resist the urge to pile on every nearby town. Keyword spam in travel is easy to spot and easy to report: if the picture was taken in one valley, do not list the four valleys next to it.
The four layers a travel keyword set has to carry
Layer one: what is literally there
The physical subject, described plainly. Harbour, fishing boat, cobbled street, market stall, terraced rice field, cable car. This is the layer an AI tagger handles well and the layer buyers use least on its own, but leaving it out breaks broad searches that feed into everything else.
Layer two: where and when
The geographic ladder plus the temporal markers a travel buyer plans around: season, time of day, weather, and whether the scene reads as high season or low season. Tourism campaigns are commissioned months ahead against a specific season, so autumn, off season, golden hour and monsoon are commercial terms in this category, not decoration.
Layer three: who the traveller is
Travel buyers shop for an audience, not a landscape. Solo traveller, couple, family with young children, group of friends, older couple, backpacker, business traveller, digital nomad: each of those is a campaign brief, and each is a different search. If there are people in your frame, say who they read as and what they are doing, because that is what turns a scenic file into a licensed one.
Layer four: the concept
The abstract idea the picture is bought to illustrate. Wanderlust, escape, adventure tourism, sustainable travel, slow travel, overtourism, remote work, homesickness, arrival. This layer is where the same photograph earns a second and third life, and it is the layer AI taggers are weakest at because the concept is not visible in the frame.
A worked example
Take a photograph of two people in their thirties sharing coffee on a narrow terrace above a harbour, early morning, small fishing boats below, shot in Kotor, Montenegro, in late September.
A weak set stops at the surface: coffee, terrace, harbour, boats, sea, morning, holiday, beautiful, view. Everything there is true and almost none of it is searched by a buyer with a budget.
A working set covers all four layers: cafe terrace, fishing boats, Bay of Kotor, Kotor, Montenegro, Adriatic, Balkans, Europe, shoulder season, late September, sunrise, couple in their thirties, breakfast outdoors, slow travel, off season travel, European city break, Mediterranean lifestyle, quiet tourism. The caption carries the same facts as a sentence, so the file reads correctly even where keywords are not displayed.
The difference is not length. It is that the second set answers questions a buyer actually types, and each of the four layers gives the file a separate route into search.
Keywords that get travel files rejected or restricted
- Brand and trademark names. Airline names, hotel chains, theme parks and visible logos. Naming them in a commercially licensed file is a reliable way to get it rejected or reclassified as editorial.
- Landmarks with usage restrictions. Some famous buildings, monuments, interiors and lighting installations carry restrictions that vary by country, and a number of protected sites require a permit for commercial photography. If you are unsure, the safe route is editorial rather than a keyword set that implies a commercial licence.
- Recognisable faces without a release. Street scenes are where this bites hardest, because a face in the middle distance is still a face. Either crop, blur, get a release, or submit the file as editorial and describe it accordingly.
- Aspirational adjectives with no search behind them. Stunning, breathtaking, paradise, must see. These are how a caption writer talks, not how a picture researcher searches, and they crowd out terms that would have worked.
- Nearby places you did not photograph. Adding the famous city forty kilometres away is straightforward keyword spam and gets treated as such.
The wording of your title matters as much as the tags here, since moderators read the two together. There is more on writing titles that survive moderation and on attaching model and property releases, which is the step most travel contributors skip until an image is pulled.
The tools, and what each one cannot do for travel
Every tool below is genuinely useful for some part of this. None of them can tell you where you were standing, which is why travel is the category where the human input matters most.
Xpiks
An open source desktop editor that is very good at the mechanical half: bulk editing, presets, spell checking and writing metadata into the file itself. For travel that is a real advantage, because a place hierarchy you build once can be applied to an entire shoot in seconds. What it will not do is generate the concept layer or decide the release question for you. It is a precise instrument for a photographer who already knows what to type.
PhotoTag.ai
Fast, clean AI tagging with a simple interface and good results on the literal layer. On a harbour scene it will find the boats, the water and the architecture. It has no way of knowing the harbour is in Kotor unless you tell it, and it does not deliver to agencies, so it sits at the start of a workflow rather than covering it.
StockSubmitter
The long standing answer for multi agency distribution, with broad platform coverage and fine grained control over how each agency receives a file. Contributors with large travel archives often run it precisely because per agency category mapping is tedious by hand. It is a Windows desktop application and the metadata itself is largely your job.
Wirestock
A managed pipeline that handles keywording and distribution for you in exchange for a share of royalties. For a photographer who travels constantly and does not want to touch metadata at all, that trade can be rational. The cost is that you give up a percentage indefinitely and hand over control of how your work is described.
Generic vision APIs
Cheap per image and trained on the open web, which means some of them do recognise very famous landmarks. They are also the easiest way to end up with confidently wrong geography, because a plausible guess and a correct identification look identical in the output. They know nothing about agency keyword rules or editorial classification.
Rastock AI
Generates titles, descriptions and keywords against each agency's own rules, embeds them as IPTC so the data travels inside the file, and delivers over FTP or SFTP to more than 10 agencies with per agency status. For travel the useful part is the control layer: banned and mandatory keyword lists, so an editorial set never picks up commercial concept terms, and rollback if a batch comes out wrong. It still cannot see where you were. You supply the place hierarchy once and it propagates across the shoot.
When to use which
- A few hundred travel files a year and one or two agencies: Xpiks or PhotoTag.ai with your own place hierarchy is enough, and paying for a pipeline would be paying for capacity you do not use.
- Thousands of files across many agencies, with editorial and commercial mixed in the same trip: a tool that enforces per agency rules and keeps IPTC intact earns its place, because the failure mode you are avoiding is a whole shoot described under the wrong licence.
- Shooting constantly and unwilling to do metadata at all: a managed service such as Wirestock is the honest answer, provided you accept the revenue share and the loss of control over wording.
- Whatever you choose, write the geography yourself. It is the single input no tool can supply and the one buyers search on first.
Fixing a travel portfolio you already uploaded
Start with the files that already sold, since those prove a buyer exists, and check whether their neighbours from the same shoot carry the same place hierarchy. Then go after the trips where you know the geography is thin, working destination by destination rather than file by file. Fix the caption at the same time as the keywords, because agencies weight them differently and an accurate caption is what a picture researcher reads before deciding.
If the question you are really asking is which software to run rather than how to write the terms, the companion piece on tools for travel stock content covers capture to delivery for this category.
Rastock AI was built for the part of this that is repetitive rather than creative: writing the layers, enforcing each agency's rules and getting files delivered without the metadata falling off in transit. You can see how the pipeline works or check the plans and the 14 day free trial, which does not ask for a card. The place names are still yours to get right.
Frequently asked questions
How many keywords should a travel stock photo have?
Aim to fill all four layers rather than to hit a number. In practice a well described travel file lands somewhere around thirty to fifty terms, because the geographic ladder alone takes five or six. Agencies differ on limits, so check the ceiling for each one you deliver to. What matters more than count is that every term is defensible: if you cannot point to it in the frame or in the location, drop it.
Can AI keywording tools identify a location from a photo?
Sometimes, for very famous landmarks, and not reliably enough to trust. A model that has seen thousands of images of a particular cathedral may name it correctly, but a plausible guess and a correct identification look the same in the output, and a wrong place name is the one travel metadata error that gets files removed. Supply the geography yourself and let the tool handle the layers it can actually see.
Do I need a property release for a famous building?
It depends on the building and the country, which is why travel is harder than it looks. Some monuments, interiors, lighting installations and privately managed sites carry restrictions on commercial use, and some protected places require a permit to photograph commercially at all. When you cannot confirm the position, submitting the file as editorial and describing it as editorial is the safe route.
Should travel keywords include the season?
Yes, and it is one of the most commercially useful layers in the category. Tourism campaigns are planned against a specific season months in advance, so terms like shoulder season, off season, autumn foliage and high summer match real briefs. Add time of day and weather alongside it, since a destination in soft morning light and the same destination in heavy rain sell to different buyers.
What is the most common travel keywording mistake?
Describing the scenery and forgetting the people. Files that carry the landscape but not who is in it, what they are doing and who they read as miss the searches that tourism, airline and hotel buyers actually run, because those campaigns are built around an audience rather than a view. The second most common is stopping at the country name when the buyer was searching for the town.
Is it worth re-keywording old travel photos?
Usually yes, because travel archives age well: a destination that was quiet five years ago may be in demand now, and the file is already shot and approved. Work destination by destination rather than image by image, start with trips near files that have already sold, and fix captions in the same pass. Most agencies re-index updated metadata within about a day.