Marginalia Search: A Small Search Engine for the Independent Web
Marginalia Search is a small, independent, open source search engine that finds the parts of the web the big engines bury: personal blogs, hobbyist pages, forums and text-heavy, non-commercial sites. It runs its own crawler and index, shows no ads, and lets you filter out sites with affiliate links, tracking or heavy JavaScript. Use it with short keyword searches, not questions, at marginalia-search.com.
This guide explains how Marginalia works, how to get good results from it, and how it compares with other small search engines such as Wiby, Mwmbl, searchmysite.net and Kagi Small Web.
What is Marginalia Search?
Marginalia Search is a non-commercial search engine built to help people discover websites that Google and Bing rarely surface. Swedish developer Viktor Löfgren started it in 2021 and still develops and operates it largely on his own. The project's about page describes it as a complement to the big engines rather than a replacement, a "minority report that keeps them honest".
The code is open source, mostly under the AGPL 3.0 license, with the sources on the MarginaliaSearch GitHub repository. The project has no investors. It is funded by donations, a grant from the NGI0 Entrust Fund (set up by NLnet with support from the European Commission's Next Generation Internet programme) and commercial API licenses.
The search engine moved from search.marginalia.nu to marginalia-search.com, and the old address redirects to the new one. The developer explained in the redesign announcement that the project had outgrown its blog subdomain and that the .nu domain made some people think the site was Russian. The older, lighter interface is still available at old-search.marginalia.nu, which suits text-based browsers.
How Marginalia Search works
Marginalia Search uses its own crawler and its own index, so its results are not a reskin of Google or Bing results. The homepage sums up the approach as "Simple technology, no AI": a keyword-based engine that favors text-heavy pages over commercial, ad-driven ones.
According to the crawler documentation, a full refresh of the index runs every 8 to 10 weeks. Between refreshes, the crawler checks RSS feeds daily and applies some live updates, so active blogs show up faster than static pages.
The results page looks nothing like Google. There are no ads, and organic results start at the top. Each result lists the URL first, then the title and a short description, followed by details that help you judge the page before you visit it:
- An info link that opens what the engine knows about the domain.
- A visualization of where your search terms appear on the page.
- Notes when a page uses affiliate links, tracking or JavaScript.
Because ranking rewards plain, content-rich pages, Marginalia is weak at product searches, local queries and breaking news. That is by design: it is a discovery tool, not a general answer engine.
How to use Marginalia Search
Marginalia Search works best with keywords that match the text you expect on the page. The developer has long noted that the engine is not built to answer questions phrased in natural language.
- Open marginalia-search.com. No account or cookie prompt is needed.
- Type a short keyword query. Search for "chrome youtube crash" instead of "why does Chrome keep crashing on YouTube".
- Pick a filter to narrow the type of site, then refine with the advanced settings if results are noisy.
- Click a result's info link to check the domain before you visit it.
The filters change which part of the web you search. Options shown on the search page include:
- Blogs, Small Web, Vintage and Wikis for personal and older sites.
- Academia and Forums for research pages and discussions.
- Recipes for cooking pages without the usual ad-heavy clutter.
Advanced settings let you search in titles only, prefer recent results, remove JavaScript-heavy pages, reduce adtech, reduce NSFW results and choose a language.
Two extra tools are worth a look. Explore shows a random selection of websites, each with a link to find similar sites. Sites lets you inspect a domain, including its link graph and backlinks.
Small search engines compared
Small search engines are independent services that index a narrow slice of the web, usually personal and non-commercial sites, instead of trying to cover everything. Marginalia is the broadest of them, but each one has a different focus.
| Search engine | What it finds | Best for |
|---|---|---|
| Marginalia Search | Text-heavy, non-commercial sites from its own crawler, with filters for blogs, forums, academia and more | General discovery and research away from SEO content |
| Wiby | Lightweight personal, hobbyist and academic pages in the style of the early web | Nostalgia browsing and vintage computers |
| Mwmbl | An independent, open source, non-profit index that users can help build | People who want to contribute to an open search index |
| searchmysite.net | Vetted personal websites and digital gardens, with ad-free pages ranked higher | Finding personal sites and adding search to your own site |
| Kagi Small Web | Recent posts from personal blogs, independent YouTube channels, webcomics and code projects | Browsing fresh posts rather than searching |
Kagi Small Web is a discovery feed, not a keyword search engine, and it only shows posts from roughly the last week. Wiby and Marginalia are the closest to a classic search box.
When to use a small search engine instead of Google
A small search engine is the better tool when the big engines return the same commercial pages over and over. Dan Luu's comparison of search results from Google, Bing, Marginalia, Kagi, Mwmbl and ChatGPT is a good read on why many users look for alternatives.
Reach for Marginalia or its peers for:
- Personal experience write-ups, hobby projects and niche technical notes.
- Older documentation and archived personal pages that rarely rank today.
- Forum threads and blog posts on topics dominated by affiliate content.
- Browsing for serendipity when you do not have a precise question.
Stick with a general engine for shopping, local businesses, maps, current news and anything that needs a fresh, complete index. Many users keep Marginalia as a second search engine in the browser and switch to it when Google's first page feels like advertising. Google's own option to try results without personalization is another way to see a less tailored first page.
Marginalia Search for site owners and developers
Site owners can check whether Marginalia knows their site with the Sites tool, and request inclusion through the "submit-site-to-marginalia-search" repository described in the crawler documentation. The crawler identifies itself as search.marginalia.nu, which is also the name to use in robots.txt, and the project kept that user agent after the domain change so existing rules keep working.
Developers can query the index through the Marginalia Search API. A shared key named "public" is available for development, but it cannot create custom filters and often hits its rate limit. Free non-commercial keys are available on request by email, and their results are licensed under CC-BY-NC-SA 4.0. Metered commercial keys remove the usage restriction and the attribution requirement.
The software can also be self-hosted, either as a copy of Marginalia Search or as a white-label search engine for your own data. The repository's readme says a full instance will run on a 32 GB machine, possibly smaller, while a local developer setup needs much less.
FAQ
Does Marginalia Search track users?
Marginalia states that it collects no personal information. Access logs with IP addresses are kept for up to 24 hours, according to its privacy statement.
Is Marginalia Search free?
Yes. Searching is free, needs no account and shows no ads. The project asks for donations to cover running costs, and only commercial API use is paid.
Can I make Marginalia my default search engine?
Yes, in browsers that accept custom search engines. Add a new engine with the search URL https://marginalia-search.com/search?query=%s, then set it as default or give it a shortcut keyword, as described in our guide to changing the default search engine in Microsoft Edge.
Is Marginalia Search Russian?
No. Marginalia is developed and run from Sweden. The old .nu domain caused that confusion, which is one reason the search engine moved to marginalia-search.com.
Related: Chrome's choose your search engine prompt, ClearURLs removes tracking elements from links
A good article about 11 best Private Search Engines. My 2 favorites from the list are Brave Search and Mojeek.
https://restoreprivacy.com/private-search-engine/
https://techrights.org/o/2020/07/02/ddg-privacy-abuser-in-disguise/
Why People Should Never Ever Use DuckDuckGo
About DDG: https://techrights.org/o/2020/07/02/ddg-privacy-abuser-in-disguise/
Thanks Martin, Marginalia is certainly worth a try. Very interesting concept of searching.
About DuckDuckGo: Why People Should Never Ever Use DuckDuckGo: https://techrights.org/o/2020/07/02/ddg-privacy-abuser-in-disguise/
It is an extensive list, I do not know if everything is true. But the first point certainly is:
“DDG’s founder (Gabriel Weinberg) has a history of privacy abuse, starting with his founding of Names DB, a surveillance capitalist service designed to coerce naive users to submit sensitive information about their friends”. (2006)
“Marginalia” ?
Sounds like something you scrape across toast.
I use SearXNG which takes all its results from Google, but without all the associated telemetry crap and ads: https://searx.be/
You can add it to Firefox with this extension by Tom Schuster: https://addons.mozilla.org/en-US/firefox/addon/add-custom-search-engine/
My default for years was StartPage, whose results were mostly just Google run through a 3rd party for privacy. I say “mostly”, because Google direct still gives slightly better results.
But I had to abandon SP because it began throwing continuous unfixable “are you a robot?” messages at me. My default is now DDG, which (for me) isn’t as good, results-wise, but it’s good enough, and much less irritating.
Am I alone in noticing the increasing amount of apparently-AI-generated crap results searches are delivering lately?
I am glad to have found this article. Will definitely be giving Marginalia a try. My frustration with search engines has always been about results, or rather the lack thereof. The problem has gotten worse over the years. It started back when Google took over the internet. Currently I stick to DDG and StartPage but neither are much more helpful than Google.
My first “Marginalia” search only revealed how amateurish and undeveloped the project is: enter a well-known, well-documented, widespread search term–zero results, even with “Academia” or “no” filter mode.
“Gran Abuelo” (Great-Grandfather) tree in chile
The same search phrase entered in DDG = numerous articles, one of the first being from the “Smithsonian”:
https://www.smithsonianmag.com/smart-news/a-new-candidate-for-oldest-tree-in-the-world-is-discovered-in-chile-180980167/
Another from “The Guardian”:
https://www.theguardian.com/environment/2022/sep/23/gran-abuelo-chile-world-oldest-living-tree-alerce
It’s not like articles on the oldest tree on Earth that has witnessed more new years than 100 crores of human rebirths are irrelevant or so mysteriously-mystic that they would be hard to find.
Another search for major Hindu texts only provided a number of “encyclopedia” type links–nothing noteworthy. I should think the “Ramayana” would have appeared somewhere–nothing.
Not a search engine for my needs.
Another one to check out:
https://www.retifo.com/
Read the “Terms” which are enough to put you off from ever using that search engine for life.
Thank you
I have 12 search engines listed of which 6 “major ones”, and DuckDuckGo is presently the one set as default.
Brave [https://search.brave.com/]
DuckDuckGo [https://duckduckgo.com/]
eTools [https://www.etools.ch/]
Mojeek [https://www.mojeek.com/]
Qwant [https://www.qwant.com/]
SearXNG : an instance among [https://searx.space/]
–
MetaGer [https://metager.org/]
QuackQuackGo [https://quackquackgo.net/]
Startpage [https://www.startpage.com/]
Swisscows [https://swisscows.com/en]
Wilby [https://wiby.me/]
Yep [https://yep.com/]
I’m off to discover ‘Marginalia’.
—
Happy New Year 2024 to all :)
I like Kagi the best these days, but DDG is still pretty good too.
I just tried Marginalia, but it seems to only work with English languages queries at the moment. That’s a bit too restrictive for my use-case, but it does look very promising and I really like the design, so I’m definitely keeping it in mind! :)
I signed up for a subscription with Kagi as well. It offers perks like FastGPT which is pretty good. And I downloaded the Kagi plugin and it can summarize any page as well as keep me logged in to my account if I delete all cookies in my browser.
I used to use Startpage and DuckDuckGo but they both serve adds on their homepage now which turned me off on using them as homepages for my browsers. These engines are free so I get it. But to pay a search engine for an add free experience is the way to go. The search results with Kagi and the perks that come with it makes the subscription well worth it.
What turns me off using the Marginalia search engine is the fact that the front page is ugly. Who would want to use that for a homepage?
Tom, happy new year to you as well (and everyone else). Let me know your thoughts after your exploration of Marginalia ;)
Martin, ‘Marginalia Search’ is indeed worth being checked out and, IMO, worth being used.
As you note it in your article, “The developer notes that the search engine “is not well equipped to answering queries posed like questions”. Instead, users should search for text that matches the intent.”
This is relevant of the way search queries have changed over time : users increasingly search within what I’d call an AI driven approach, that is questioning the engine rather than searching with text, keywords. Given I personally search the Web in the old fashioned way, no problem here with favoring text to questions!
Marginalia is fast, you’ll find answers not found elsewhere, several links lead to .htm rather than .html pages and I wonder if this is relevant of some less known sites/pages of older times (not sure of this!).
As you point it out, the tips displayed on the homepage should be fully read to take full advantage of ‘Marginalia Search’.
Cosmetics :
– Results are deployed within infinite scroll, a pleasure
– Page is not centered, this css allows to fine tune the display:
// Page width (I like ’em wide!)
body {min-width:75% !important;}
// Center the page
body {margin:0; position:absolute; left:50%; margin-right: -50%; transform: translate(-50%)}
// Results’ titles colors
.title {color:#FFDD00 !important; font-size:18px !important;}
Makes my last day of 2023. Nice find, Martin.
The page you mention as being the source of your discovery of ‘Marginalia Search’ [https://danluu.com/seo-spam/} is as well most interesting.
Thanks Tom, was almost sure that you would find a liking to it :)
I use DDG.
Home page:
https://start.duckduckgo.com/
Because it is the most private search engine that returns me correct results with my searches
It’s cute that you still trust them after everything.
DDG sucks now. I switched to https://www.startpage.com/?t=dark