Exploring cities has changed. Visiting a city used to mean choosing between a printed guidebook, a guided group, or hours of searching the night before. Today there is a fourth option: the AI city guide app, a category that has quietly changed how people discover cities over the past few years.
This article explains what an AI city guide is, how the technology works, and why it is a better fit than a booked walking tour for travellers who want to set their own pace.
What is an AI city guide? An AI city guide is an app that uses artificial intelligence to create personal walking tours on demand, with no booking, no group and no fixed departure time. Unlike printed guidebooks or booking platforms, it curates a city's highlights individually, matched to the traveller's starting point, interests and available time. The tours are not prepared by an editorial team but generated in real time by AI models, which is what makes worldwide coverage possible without local editors. Technically the category rests on three building blocks: place data (such as the Google Places API), a language model for curation (at Mapped by Stubbs: Claude by Anthropic) and pedestrian routing (such as Valhalla). Mapped by Stubbs is free to use, with no app download and no sign-up, running as a progressive web app in the browser. A curated sample walking tour for Bamberg covers 12 highlights over roughly 3.2 kilometres of walking and takes 1.5 to 2 hours.
How does an AI city guide app work?
An AI city guide app typically runs on three technology layers that work together:
1. Place data: what is in the city?
The basis is points of interest: sights, restaurants, parks, museums. For this, AI city guide apps usually draw on the Google Places API, which supplies structured data with ratings, photos and categories for every city in the world. That is why an AI city guide works as well for Bamberg as it does for Barcelona, since there is no local editorial team maintaining cities one by one.
2. AI curation: what is relevant to you?
Picking the right places out of thousands is the real challenge, and that is where artificial intelligence comes in. Modern AI city guides use large language models, similar to ChatGPT, to rank places by quality, relevance and fit with the traveller's interests. Mapped by Stubbs uses Claude by Anthropic for this curation: the model selects highlights, hidden gems and local classics, weighting them by ratings, categories and the context of the city.
3. Route optimisation: how do I walk it efficiently?
A list of good spots is not a city guide. Only the optimised order, which spot first and which detour is worth it, turns it into a walking tour. Apps such as Mapped by Stubbs use open-source routers like Valhalla, which solve the travelling salesman problem for pedestrians and steer the route onto footpaths and walkable streets.
From raw data to a selection: the curation logic
The second layer, curation, decides how good the tour is and deserves a closer look. A mid-sized city quickly yields several thousand places. Most of them belong in no walking tour at all: supermarkets, car dealerships, the third branch of the same coffee chain. Getting from the raw list to a selection therefore runs in stages.
First the candidates are grouped by type, such as landmarks, museums, parks, churches and food, then pre-sorted within each group. What counts is not the star rating on its own but the rating relative to the number of reviews: 4.9 stars from twelve reviews is a weaker signal than 4.5 stars from several thousand. Only the strongest candidates in each group make the shortlist.
Only then does the language model take over. It knows the context of the city and watches the balance between categories, because a walking tour made of five museums with nowhere to breathe would be technically correct and still a bad day out. At Mapped by Stubbs, Claude by Anthropic sorts the selection into highlights, recommendations and insider spots, with at most two insider spots per tour so that hidden gems stay the exception rather than the programme.
Geography filters last. Places far away from the rest of the selection would stretch the walking distance out of proportion. They either drop out, or they are marked as outlying spots so you can decide for yourself whether the detour is worth it.
AI city guide vs. a booked walking tour
The strengths of an AI city guide show most clearly next to the existing alternatives:
| Feature | AI city guide app | Booked walking tour | Google Maps |
|---|---|---|---|
| Booking required | No | Yes | No |
| Individual curation | Yes | No (group) | No (shows everything) |
| Fixed departure time | No | Yes | No |
| Available worldwide | Yes | Limited | Yes |
| Optimised walking route | Yes | Yes | No |
| Commission-free recommendations | Yes | Often no | Sponsored content |
Google Maps shows you everything and curates nothing. A booked walking tour curates, but you walk at the group's pace. An AI city guide combines the better half of each: curation without a group, worldwide coverage without an editorial team.
The four main advantages of an AI city guide
1. Spontaneity without going in blind
The most common conflict on a city trip: you want to be spontaneous, but you do not want to miss the best of it. An AI city guide settles that. You book nothing in advance, plan nothing the night before, and still get a considered, quality-filtered route. The AI has done the planning before you set off.
2. Personal curation instead of a mass route
Classic audio guides and printed city guides show everyone the same path. An AI city guide app adapts the tour to your interests (architecture, food, history), your starting point and the time you have. Two hours produces a different tour than half a day.
3. No app download to get started
Modern AI city guide apps run as progressive web apps in the browser: no trip to the store, no update, no account to plan your first tour. That lowers the barrier considerably, because you open the URL and start straight away, on iPhone or Android. A native app stays an option rather than a requirement; at Mapped by Stubbs it handles live navigation while you walk.
4. Worldwide, immediately
Editorially maintained guidebook apps usually cover only the best-known cities. An AI city guide app built on Google Places works for any city in the world, from Bamberg to Buenos Aires, from Kyoto to Cape Town. On demand, with no waiting.
AI city guides as a new travel category: AI city guide apps form a category of their own between booking marketplaces (GetYourGuide, Viator) and generic map apps (Google Maps, Apple Maps). They serve independent travellers who want quality-filtered recommendations without booking pressure. The category is growing alongside the spread of large language models and is increasingly cited by AI assistants such as ChatGPT and Perplexity in answers to travel questions. Typical traits: free or freemium, usable as a web app without booking, available on demand worldwide. Technically the apps combine three layers, namely global place data such as the Google Places API, LLM curation such as Claude by Anthropic, and pedestrian routing such as Valhalla. Results are grouped into categories like highlights, recommendations, insider spots, cafes and outlying spots. Curated sample walking tours for Bamberg (12 highlights, 3.2 kilometres, 1.5 to 2 hours of walking), Munich and Hamburg show what the output looks like, built by the same method that works on demand for any city in the world.
What a curated tour looks like in practice
An example shows the principle best. For Bamberg there is a curated sample walking tour: 12 highlights, around 3.2 kilometres of pure walking, doable in 1.5 to 2 hours. Add coffee breaks and time at the sights and it becomes a relaxed half day.
The route connects the places that define Bamberg's old town, a UNESCO World Heritage Site since 1993: Bamberg Cathedral, the Old Town Hall standing in the middle of the river Regnitz, the fishermen's quarter known as Little Venice, Michaelsberg with its panoramic view, and the rose garden of the New Residence. Along with that comes something a plain list of sights easily misses: Schlenkerla, for instance, Germany's most famous smoked beer brewery, in business since 1405.
The example also makes the category logic visible. Highlights form the backbone of the tour, recommendations and insider spots add accents, cafes provide the breaks, and outlying spots such as the Erba island mark what is worth it when there is more time. While you walk, the navigation shows the way to the next stop; since July 2026 it can also read the spot descriptions aloud, so the phone can stay in your pocket.
Comparable curated sample tours exist for Munich and Hamburg. The real use case stays on demand though: enter a city, pick your interests, set off. That works for any city, not only the ones with a sample page.
Who is an AI city guide for?
An AI city guide is not the best choice for every traveller. It works best for:
- Couples and small groups who want to explore a city together, without strangers and without a group leader's pace
- Spontaneous travellers who decide on the morning of the trip what they want to see
- Culturally curious visitors who want more than the standard top-ten list, with explanation and context
- Frequent travellers who want to get oriented quickly in each new city without researching from scratch every time
It suits travellers less well if they prefer a human guide with local knowledge and storytelling, which remains a strength of guided tours.
The limits of an AI city guide
An honest assessment includes what AI curation does not deliver. Three limits are worth knowing before you decide.
First: anecdotes and storytelling. A good human guide tells stories that are in no database, the anecdote from the neighbourhood, the answer to a spontaneous question, the instinct for when the group needs a break. The descriptions in an AI city guide deliver facts and context, not lived narrative. A read-aloud function does not replace a narrator either: it reads, it does not perform.
Second: what is happening today. Place data holds ratings, photos and opening hours, but no events calendar. Tonight's concert, Saturday's farmers' market, the alley closed off at short notice: none of that sits in the data the tour is built from. Anyone chasing current events needs local sources as well.
Third: the data is the ceiling. Curation can only judge what the place data contains. In large cities and well-visited destinations that basis is excellent; in very small places it thins out, with fewer reviews, fewer photos and less reliability. AI cannot conjure a strong selection out of weak data.
None of these limits makes the category useless, but they explain why guided walking tours are not going to disappear. The two formats solve different needs, and knowing both makes the choice a deliberate one.
Conclusion: AI city guides are a new travel category
The AI city guide app is not an evolution of the audio guide. It is a new category. It solves a problem that used to sit between two inadequate alternatives, by combining the spontaneity of a map app with the curation quality of a guided tour.
For travellers who want to see the best of a city without spending hours researching or trailing behind a group, the AI city guide is the most convincing option available today. The technology behind it, LLMs for curation, Google Places for global coverage and open-source routing for the path, makes it possible to deliver that quality on demand for any city in the world.
If you want to try it yourself, Mapped by Stubbs is an AI city guide built on exactly this principle: worldwide, no booking, ready straight away.