Former Travel Writers Test AI on France Trip, Find It 80% Accurate
Two former travel writers who left the industry over AI content flooding search results have completed a deliberate experiment: they let AI plan an entire trip across France, then graded what worked and what failed.
Amber and Eric Hoffman, who spent more than a decade writing about food and travel for publications and tourism boards, used Gemini and Perplexity in parallel to plan a multi-city route through France. The tools independently recommended the same itinerary: Bordeaux, La Rochelle, and Nantes, with a day trip to Clisson.
The result was roughly 80% accurate - strong on major decisions, unreliable on specifics.
Where AI Excelled
Both tools selected a country the Hoffmans didn't know well, sequenced three cities logically, and identified a workable train route between them. The recommendations surfaced La Rochelle, a port town neither the Hoffmans nor their well-traveled friends had considered before.
Restaurant recommendations landed roughly 80% of the time. AI directed them to a Bordeaux food market they returned to three days in a row.
"AI was excellent at the big decisions and unreliable on the details," Amber Hoffman said. "It chose the right country, cities, and route - then sent us to a wine tasting that didn't exist."
Where It Failed
Hours and menu details were frequently wrong. When asked to replicate the open-air market dining experience they found in Bordeaux at other stops, AI recommendations missed the mark.
One specific failure: AI suggested a wine tasting that had no actual existence. The tools couldn't grasp nuance or understand how to adapt a successful experience from one city to another.
What This Means for Travel Writers
The Hoffmans see a clear division of labor emerging. AI now handles the structural work - route planning, identifying destinations, basic restaurant discovery. Travelers still need to verify details themselves before booking.
For travel content creators, the implication is direct: AI will absorb logistics and listicles. First-hand storytelling and lived experience become more valuable, not less.
Amber Hoffman, who also consults with companies on AI visibility, approached the experiment as both traveler and researcher - testing not just whether AI could plan a trip, but how accurately it surfaces real-world hospitality businesses.
The full account of what AI got right and wrong across all four destinations is published on Food & Drink Destinations.
For AI for Writers, this experiment offers a practical case study in how generative tools perform on research and verification tasks - core skills for anyone writing about real-world subjects.
Your membership also unlocks:
