The Great Umbraco Search Shake-Up
Senior Developer, Busra, took to the Umbraco UK Festival stage to share what she learned comparing three search engines.
The Great Umbraco Search Shake-Up: What we learned comparing three search engines
Search on Umbraco websites is changing more than it has in over a decade. At the Umbraco UK Festival on 16 October 2026, I put three search engines side by side on the same website and measured them. Here’s what I found, and what it means if you run an Umbraco site.
Why search matters now.
For many visitors, the search box is the quickest way to find a product, an article, or an answer. When search misses, they leave. So, the engine behind your search box is a business decision, not just a technical detail.

What’s changing in Umbraco?
For 17 years, every Umbraco website has used the same built-in search engine, called Examine. Its creator, Shannon Deminick, still looks after it, and it has been downloaded more than 23 million times.
Umbraco is now making something you can plug in. A new layer called Umbraco Search sits between your website and the search engine. Your website asks its questions in one standard way, and Umbraco Search passes them to whichever engine you choose. That could be Examine, or a cloud service such as Algolia or Microsoft’s Azure AI Search.
Umbraco Search is available today as an add-on for Umbraco 17 and 18, and it’s planned to become part of Umbraco itself in version 19. Examine stays the default, so nobody has to change.
How we tested it
To keep it fair, I built one Umbraco website with 5,000 articles and products, each with up to 16 pieces of information: categories, prices, ratings, dates, and more. Visitors could narrow results in 12 ways, such as price range or topic, and sort them in five ways.
The same content went into all three engines at the same time. To check the comparison was fair, I confirmed that all three gave exactly the same counts when filtering, down to the last item. Then I ran every type of search 50 times on each engine and compared the typical result.
What we found
Typos are where the engines really differ. We searched for “chiar” instead of “chair”. Algolia found 101 results and Azure found 97. Examine, as it comes in Umbraco Search, found none. For a shop, that’s a visitor who was ready to buy and found nothing.
Speed depends on what you ask for. On a typical search, Examine answered in about 35 millisecond, against about 50 millisecond for Algolia and 55 millisecond for Azure. Examine runs inside your website, while the cloud services sit on the internet, so every search has to travel there and back.
The surprise was the filters. Without the filter options down the side of the page, Examine answered in about one thousandth of a second. With twelve sets of filters, most of its time went on counting them. The cloud services stayed at around 50 either way. So for a simple search box, Examine is hard to beat. For a shop page with lots of filters, the cloud services catch up, and they do it without using your web server’s processing power.
All of these times are tiny, far quicker than a person can notice. What matters more is how each engine behaves as your site and its traffic grow.
Behind the scenes
Keeping three engines in step. Each engine needs to be told in advance about the information it will filter and sort on, and each wants that list in its own format. Rather than keeping three lists by hand, which sooner or later drift apart, I wrote the list once. A small translator for each engine turns it into that engine’s format. Add something once, and all three get it. It’s a simple habit, and it’s why the three engines could give identical results.
Finding and fixing a bug. While building the demo, I found a bug in the new Examine integration: a price filter such as “£1,000 and over” made the whole search fail. I reported it to Umbraco, and it turned out their team was already fixing it. I tested the fix, and it’s on its way into Umbraco Search. That’s the open-source community at its best.
What it cost
All three can be tried for free, but each has a catch:
Examine costs nothing beyond the server you already have. The catch is that it uses that server’s power.
Algolia’s free plan includes 10,000 searches a month and up to 50,000 items. Many of its AI features are on paid plans.
Azure AI Search has a free tier for trying things out, with up to 50 MB of content and 3 search indexes on shared servers. For a live site, you’d move to a paid tier. Prices change often, so check the latest pricing before you decide.
What about AI?
Everyone asks about AI. Algolia can tailor results to each visitor, learning their preferences from what they click and buy, but that’s on its paid plans. Azure offers search by meaning, not just by matching words, on every tier, including free. Umbraco also has its own AI search package built on Umbraco Search.
One honest note: Umbraco Search itself doesn’t yet have a concept of “this particular visitor”. Personalised results today mean working with the chosen engine directly.
Which one should you choose?
Examine is still the right answer for most Umbraco sites. It’s free, it’s the default, and it’s very fast for a simple search box. Just keep the number of filters sensible, and know that it doesn’t forgive typos out of the box.
Algolia suits sites where a polished search experience really matters: typo tolerance, lots of filters at a steady speed, and business teams who want to tune results themselves.
Azure AI Search suits organisations already using Microsoft Azure, and anyone moving towards search by meaning.
Should you try it now?
Yes, carefully. Umbraco Search runs alongside your existing search, so you can try it without ripping anything out. It does upgrade Examine to a new version, so any custom search code should be tested before going live.
Want to see it for yourself?
The demo website, the content generator and the measurements are all available on GitHub here.
If you’re wondering which search engine is right for your website, or how to make your current search work harder, get in touch with the true digital team. We’d be happy to help.


