■ HIGH RISK ■ Media & Communication
For the compilation end of the trade — listings, itineraries, '10 best' roundups — yes, and it barely needed to try, since that content was aggregation with adjectives. The travel writing that survives is the kind that requires actually going: firsthand reporting, real recommendations with a name attached, and voice people follow on purpose.
“AI writes travel content from reviews. Your 'authentic' guide was just Yelp in book form.”
Our AI replacement risk score — how we score jobs
Travel guide writing has two halves that the byline conveniently blurs. One half is genuine fieldwork: visiting the hotels, eating the questionable street food, learning which ferry schedule is fictional, and turning it into trustworthy guidance. The other half — the majority, commercially — is desk work: compiling listings, updating opening hours, rewriting destination boilerplate, and assembling itineraries from other people's information. Guidebook publishers, travel sites, and content farms have run for decades on that second half, often paying writers to 'update' places they'd never been.
LLMs do the desk half instantly and for free, which is why travel content was among the first verticals AI flooded. Models trained on the internet's reviews, blogs, and guides generate serviceable destination pages, itineraries, and listicles on demand; travel platforms now auto-summarize reviews into exactly the digest a guide once provided; and trip-planning AI assembles personalized itineraries interactively, which is strictly better than a static guide for that use case. The market signals followed: guidebook publishing was already shrinking, travel content rates collapsed with the rest of commodity writing, and search results filled with synthetic destination content of confidently variable accuracy.
The durable core is verification and trust — the things that require a human body in a place. AI cannot know the restaurant closed last month, that the 'beachfront' hotel faces a parking lot, or that a neighborhood's vibe shifted; its travel hallucinations are a running joke among travelers who followed them to nonexistent attractions. Writers with beats they physically cover, audiences that trust their taste, and distribution of their own (newsletters, video, social) are converting that trust into viable businesses. The 75 risk score falls on the compilation trade; the correspondent trade is smaller, harder, and still standing.
Automatability: our editorial assessment of current and near-term AI capability
The flood already happened — AI destination content saturated the web from 2023 onward, and commodity travel writing rates went down with the rest of the content market. Guidebook publishing continues its long structural decline in parallel. Through the late 2020s expect the compilation tier to finish converting to AI-plus-editor workflows, while firsthand, personality-driven travel media keeps consolidating around individual trusted names rather than mastheads.
Unevenly, and that's being polite. AI travel content is fluent about places as they existed in training data — it confidently recommends closed restaurants, misstates transit details, and occasionally invents attractions entirely. Travelers have followed AI itineraries to places that don't exist. That unreliability is precisely the surviving human writer's pitch: verified, current, firsthand information is now a differentiator instead of an assumption.
As a freelancer selling destination articles to whoever's buying — barely, and shrinking. As a travel media operator, yes: writers with a defined beat, an owned audience, and layered revenue (newsletters, video, affiliate, trip-planning services, guiding) are building sustainable businesses on trust AI can't replicate. The job stopped being 'write about places' and became 'be a specific person people trust about places.'
Because so much of it was already secondhand. Destination roundups and itineraries were compiled from reviews, blogs, and other guides — exactly the sources LLMs trained on — so models reproduce that tier natively. Combine that with travel SEO content's enormous volume and thin margins, and it was the first place publishers cut human writers. The firsthand-reporting tier was hit far less, because models can't visit anywhere.
Three assets AI lacks: presence (you were physically there, recently, with photos and receipts), relationships (locals, fixers, tourism sources feeding you what isn't online), and identity (a voice and track record audiences deliberately seek out). Practically: pick a coverable beat, publish on owned channels, document verification visibly, and treat static listicle work as gone. The market rewards correspondents now, not compilers.