How our search works
You type in a few words, and a moment later matching pages appear. Behind this are three steps that happen long before your search: collecting pages, building an index and ranking the results. Here we explain how this works at Polias and where the limits are.
1. Collecting pages
A search engine can only find what it has seen before. That's why our crawler, the PoliasBot, continuously visits public web pages and reads them just as any visitor sees them. It starts with websites that many people know, such as pages of cities, authorities, clubs and associations, newspapers and reference works, and follows the links from there to further pages.
It follows rules while doing so: it respects the robots.txt file, in which website owners specify
what may be collected. It only fetches a few pages per website at a time and pauses in between so that
no server is overloaded. It never submits forms and never logs in anywhere. That's why Polias doesn't
know anything that sits behind a login.
2. Building the index
The collected pages become the search index. You can think of it like the index at the back of a thick book: for every word, it lists the pages it appears on and where, for example in the title, in a heading or in the text. That way, nobody has to read through the whole web for a search, just look it up in the index.
Words are reduced to their base form. So anyone searching for „Fahrräder“ (bicycles) also finds pages about a „Fahrrad“ (bicycle). For each page we also keep a short snippet, which you see below the title in the results, and how many other pages link to it.
Our index is our own: we collect the pages ourselves and run it on servers in the EU. That comes at a price: an index of your own has to grow first. Some pages, especially small local businesses, are still missing.
3. The order of the results
For almost every search, a great many pages match. What matters is which ones come first. To decide, we look at, among other things:
- How well the page matches your words. If they appear in the title, the address or a heading, that counts for more than a single mention far down in the text.
- How many other pages link to it. A page that many others link to is usually an important source for its topic.
- Which results others found useful. We count how often a result is clicked for a search term so that we can improve the order. This is done only as a daily total and without any link to the person who clicked.
What we don't do: we never reorder anything for money. Paid entries only exist as clearly labelled ads, separate from the normal results. And we don't tailor the order to you personally: anyone who searches for the same thing sees the same results.
4. The AI answer
For some questions, a short answer appears above the results. It's written by an AI, which receives your search query and the snippets of the best results for this, never any details about you. Where a statement comes from a result, a small number after it leads to the source. Under "AI answer" there's a more detailed version, after which you can ask follow-up questions without starting your search again.
An AI can make mistakes, which is why there's a note about this above every answer. You should check anything important in the sources. If you spot an error, you can send it straight to us with "Report an error".
5. What we know about you in the process
As little as possible. We don't store IP addresses and we don't build a profile. Your recent searches are kept only in your own browser. All we count is how many searches there were and which country they came from. All the details are on the Privacy made simple page and in our privacy policy.
Where the limits are
- New pages take time. The index often only knows about what's published today a few days later. For breaking news, a news website is therefore often the better choice.
- Local coverage still has gaps. Some doctors, pharmacies or tradespeople near you are still missing. We're working on it.
- Images and videos are still being built up.
- Mainly German and European. In English, the focus is on technical topics (developer documentation, programming, software, IT, English Wikipedia).
If you're missing a page or think a result is wrong, write to us: hallo@polias.eu. Tips like these help us most.
Read on: Search tips · A short search engine glossary · FAQ