AI-Powered contextual smart map
MapAI
A map that understands your travel context, not just your destination.
See it in action
Context
University Project
Team
4 designers
Presented
February 2026
Category
Product Strategy, UX Research, Interaction Design
overview
What it is
A mobile concept combining map-based discovery, personalised recommendations, conversational refinement, an editable preference profile and saved places.
Who it’s for
Travellers who want personalised suggestions. Our research personas: an open-minded traveller who doesn’t know the place, a traveller who wants to see everything on a tight budget, and a traveller who knows the place and wants personalised advice.
What I did
I raised the original MapAI idea, contributed to the concept, and designed the screens and interfaces.
Design Methods
Story interviews, storyboarding, prototyping, Wizard of Oz usability testing
the problem
Why the usual Map apps is
not enough?
Planning a trip often means jumping between maps, reviews, social media, travel blogs, and conversations with friends. Each source offers a different piece of information, leaving travelers to connect the dots by themselves.
Yes, these days AI can simplify this process by suggesting places and building itineraries,
but recommendations are only useful when they reflect the traveler’s actual context.
So, how might we help travelers discover places that fit their needs while keeping them in control of the decisions?
Research
Story interviews
We conducted eight story interviews as a team to understand how people discover places and make travel decisions. I conducted two of these interviews, exploring both group-trip planning and solo travel.
“We used too many apps. I wish there was one place where we could compare destinations, see events, weather, and budget together.”
Interview 1 · 22-year-old student planning a group New Year’s trip
“I couldn’t really tell if the plan was efficient or affordable… I still felt like I had to double-check things elsewhere, like to check costs and travel time outside the app.”
Interview 2 · 24-year-old remote worker planning a solo trip with an AI travel app
What we learned from 8 interviews
01
Fragmented information
Travelers compare maps, reviews, social media, and personal recommendations because no single source answers every question.
02
Context changes everything
A recommendation depends on more than a destination. Budget, companions, dietary needs, transport, and the purpose of a trip can all change what feels suitable.
03
AI helps discovery, but trust needs evidence
AI can reduce the effort of finding places and building itineraries. Travelers still need to verify practical details and understand why a suggestion makes sense.
04
Flexibility matters
Travel plans evolve. People need to adjust their criteria, explore alternatives, and remain free to make the final decision.
The design opportunity
Instead of asking travelers to adapt to a fixed itinerary, MapAI explores a more flexible approach: a map that responds to their context, explains its suggestions, and lets them steer the experience.
design process
From research to prototype
Storyboard
The storyboard help us translate our research into a restaurant-discovery scenario. These are 4 key moments from our 14-frame storyboard.
01
Set the starting point
The traveler chooses a location and transport mode to establish the context for discovery.
02
Explore within real constraints
A travel-time limit narrows the search area, helping the traveler focus on places that fit the journey rather than distance alone.
03
Refine through conversation
Instead of navigating endless filters, the traveler clarifies cuisine, dietary requirements, and the occasion through a conversation with the assistant.
04
Understand and decide
Recommendations include explanations of their suitability. The traveler can save a promising option or reject one that does not fit.
Paper prototype
Before going to any high-fidelity design, we used paper prototypes to explore how the map, contextual constraints, and conversational assistant could work together.
01
Set the start point
The app asks where you start and how you travel: car, walk, bike or public transport.
02
AI suggests a place
A “Suggested by AI” label marks the best match. When she saves Poca, the AI replies with a friendly tip.
03
A gentle warning
She saves a place that doesn’t fit her needs. The AI warns her, but still saves it. She stays in control.
The solution
MapAI's Features
The goal was not to automate every decision. It was to make each decision easier to understand.
Context-aware search
Explore places based on your budget, transport, travel time, and personal preferences… Or skip directly to the map.
Personalized Recommendations
Go beyond map with conversational AI recommendations that explain why a place fits your needs.
Recommendations with context
MapAI's most faved feature. You can manage preferences that adapt to your travel context, without being defined by a fixed profile.
Sushi only with friends? Show me parks but without having to do 100 steps stairs? Gelato but only when its 30 degrees outside?
From discovery to itinerary
Turn saved places into a flexible plan, reorder stops, choose transport, and review travel times. Your trip, your decisions.
Compare places
Save your favorite places and compare them side by side. Make more informed decisions before choosing where to go.
Testing and iteratioNS
Testing with Wizard of Oz
We conducted a comparative interaction study with three participants aged 23–25 who were frequent travelers or food enthusiasts.
Participants used a high-fidelity Figma prototype to explore restaurants near the Louvre, apply travel constraints, find an Italian or vegetarian option, and save or reject recommendations. AI responses were simulated using the Wizard of Oz method rather than generated by a working AI backend.
The study helped us observe how people navigated the experience, interpreted recommendation explanations, and managed their choices.
What we found and refined
Finding 01
Make the AI easy to find
What we saw
People missed the Suggest bar. They looked at the pins first.
What we changed
We swapped the bar for a floating button with a clear bot icon.
Finding 02
Don’t repeat the map
What we saw
Price and rating were already on the pins, so the AI felt extra.
What we changed
Pins now use colour: green for a good match, grey for no vegetarian options. Next idea: the AI shares what pins can’t, like the menu or the vibe.
Finding 03
Explain, kindly
What we saw
People liked hearing why a place was picked, like “Has vegetarian options”.
What we changed
Every suggestion shows its reasons, and the AI stays friendly when people ignore a warning.
Finding 04
Compare saved places
What we saw
Comparing two saved places was hard without a side-by-side view.
What we changed
We added a Compare view once 2 places are saved, showing travel time and diet fit.
Finding 05
Stop asking twice
What we saw
Typing “Italian” again felt “annoying” when it was already in the profile.
What we changed
Idea, not built yet: use the profile to filter the map before Suggest.
The design principlES
The user is the pilot. AI is the navigator.
MapAI is designed to assist decisions, not make them on the traveler’s behalf.
Listed principles were based on Nielsen’s usability heuristics.
User agency
Related to: User control and freedom
Travelers choose their constraints, refine requests, and decide which recommendations to keep.
Example: In the storyboard, Anna saves a restaurant the AI flagged as unsuitable. Her choice stands.
Explainable recommendations
Related to: Recognition rather than recall
Suggestions should communicate why a place may be suitable, helping travelers judge the result rather than simply trust it.
Example: “Italian, has vegetarian food, closest to the museum” or “no vegetarian options”.
Editable preferences
Related to: User control and freedom · Flexibility and efficiency of use
Personalization should remain visible and adjustable, so travelers can correct assumptions as their needs change.
Example: “Vegetarian only when travelling with friends.” Explored as a concept; not demonstrated in the tested prototype.
Recoverable decisions
Related to: User control and freedom · Help users recognize, diagnose, and recover from errors
Rejecting a suggestion or changing a constraint should leave room to explore another option.
Example: Editing inputs during use was added to the prototype after the mini-study.
Final design
See it in action
Set your trip context in a few taps, then explore matching places on the map.
Clickable Figma Make prototype. AI replies are scripted.
What I learned
A useful recommendation is more than a
good match
01
Context is part of the interface
Preferences only become useful when the product considers the circumstances in which they apply.
02
Transparency supports trust
Explaining why a place is recommended gives people something they can evaluate, question, and correct.
03
Interaction details shape the experience
A useful AI feature is not useful if people cannot find it, and saving options does not automatically make comparing them easy.
04
AI needs boundaries as well as intelligence
Designing a conversational experience means deciding what the system should suggest, what it should explain, and where the user should remain in control.
Disclaimer: MapAI is a high-fidelity prototype. The AI conversational responses, live data integrations and production are all simulated.
Fun fact!
Google announced Ask Maps shortly after our final presentation. Our independent academic project explored a narrower question: how can contextual preferences and clear explanations keep travellers in control?
Show & Tell · BitCommit 2026
MapAI was selected for BitCommit’s Show & Tell
BitCommit is a mini-conference by AlumniEI-FEUP on how AI is changing our work. MapAI got one of the 10 Show & Tell stands, where I showed it to FEUP alumni and people working in AI in Porto on October 10 2026.