I’ve spent the past two years building NeoSearch with a simple premise: search got bad not because the web got worse, but because the incentives did. Ads, profiling, and SEO all pull results away from what I actually wanted.
NeoSearch is my attempt to fix that. It’s private by design (no cookies, no IP logging, no profiles, no ads), and it uses AI to improve results rather than replace them: independent sources get ranked above affiliate listicles, titles and snippets are rewritten to say what a page actually is, duplicate results get merged into one enriched entry, and “Lenses” let you flip between meanings — search “eagles” and choose the band or the football team.
Currently it takes the top results from Google’s index as raw material, then re-ranks, rewrites, and enriches them so heavily that Google’s #20 result can land at #1. That’s stage one, to prove the experience.
Stage two is crawling the parts of the web mainstream engines ignore.
The end goal is an open, distributed search index anyone can build their own engine on. I don’t want to be the next Google, I want search to stop needing one.
As of this week the whole thing is open source (Apache 2.0) on GitHub, so you can audit it or run your own.
It’s free, I’m funding it personally, and there will never be ads. I’d genuinely love your feedback — especially the searches where it does worse than Google. Those are the ones I learn from. Ask me anything!
Every result gets assessed by AI
Every title and snippet is rewritten from scratch
I am looking for a search engine with less hallucinated nonsense so I assume this isn’t for me
Doing some more digging some serious redflags are coming up. Several different accounts on Reddit are all claiming to have invented or be working on Neosearch. From what I can see it’s run by one person on Gitlab with no other contributors. Even more odd all the accounts across all platforms are brand new. The ones one Reddit are marked “18 and up only “ content.
Something really phishy is going on. And it’s a shame because the sites leave nothing that I can see uBlock origin is detecting. Which normally to me would be a wonderful thing
Until more information can be surfaced or explained I’d highly suggest ignoring this …
I’m not sure what you’re talking about here, these are our Reddit accounts: https://www.reddit.com/user/BartJellema/ (11 year old account) https://www.reddit.com/user/PhillSebben/ (5 year old account)
There are some other things out there called NeoSearch, maybe that tripped you up? And yes, I’m the only developer so far. It’s been open source for 2 days!
Also, this is not us. Looks indeed like a brand new account. Probably karma farming or something. https://www.reddit.com/r/foss/comments/1v97ge5/neosearch_is_now_open_source/
Also, I totally support being a bit cynical these days… trust me, in the process of building NeoSearch and finding out how many things really work, I’ve also become very cynical. This is partially why I didn’t promote NeoSearch for the last 2 years. It was nice to just have a handful of users. As an example, I was looking into adding movie showtimes. Turns out that movie showtimes is a multi-million dollar business where theaters sell their showtimes (yes, you read that correctly, they SELL their movie schedule that is publicly listed on their website) to data brokers and to get access you need to pay those data brokers. One of too many examples.
Do you run both Reddit accounts ?
I just tried looking for this https://neosearch.org/?q=Golf+4+1.4+MAP+sensor+normal+pressure+readings
Results are stunning compared to google’s BS, nevermind the lack of ads and ACTUAL RESULTS and information as opposed to “are you looking for this?” https://www.google.com/search?q=Golf+4+1.4+MAP+sensor+normal+pressure+readings
Amazing! I’m bookmarking this and will be recommending it to everyone!
Thanks! You made my day… nice to get this type of response after 2 years of mostly working in a vacuum :)
Dude this is fucking amazing. Another great example imo:
The answer: The package is in the local warehouse (as in, not still in transit from whatever other state or city) and is awaiting final delivery.
Neosearch: https://neosearch.org/?q=What+does+“Out+for+delivery”+mean+on+Etsy%3F The literal first thing is exactly what I’m looking to know. Bingo, right away. But I also get a nice organized list of results if I really am a maniac about delivery terminology.
Startpage (basically google): https://www.startpage.com/do/search?pl=opensearch&t=device&sc=665xzJlcgGieM8M81frOBRrTLcSGlMZztqKrtYtY4PXzpNrRR1rC2XHV4e1WDfWDmrrcwu5PCvbv1hMbH&lui=english&language=english&query=What+does+"Out+for+delivery"+mean+on+Etsy%3F&cat=web&fm=70ab7cbd10f407df1f79e683dd6a3dda&fmt=ONMTSIORUTN&sp_grace=1 2 completely irrelevant first results from Reddit and Facebook. 3rd one is the right one but the answer is buried in the article.
Google: https://www.google.com/search?q=What+does+“Out+for+delivery”+mean+on+Etsy%3F&iflsig=ABILxe8AAAAAamkSMJqFYh9pvQeswC98HPVB4gunEUUx Does pretty good actually. First highlighted result is what I’m looking for, but the AI overview omits the most important part (the “local” part): “the courier has your package and is in the process of delivering it to you”. What the hell is that link though?! More tracking than searching.
Chatgpt: https://chatgpt.com/share/6a6904c6-78c0-83eb-9068-704b9cbe09c6 Answers properly and the first paragraph contains the most important bit, but as always, makes a word salad.
Idk man if this project of yours is financially viable I’m all for it. How much would one user need to pay monthly to pay for their own usage, approximately of course? Is it viable? To me it seems awesome so far.
The best part is you get the results too not just AI hallucinations and made up walls of text. You get links, media, different sections with nice labelling, it’s fucking incredible. The very few times I’ve tried it, it’s also been accurate too.
Looks interesting! Will keep trying it over the next few days.
- I’ve already changed the “Continue your search” settings to disable Google & Chat GPT; enabled DDG. Nice to have the option.
- Maybe add Marginalia support for small web / indie web results?
- I searched for a film title & was pleasantly surprised to see results for streaming sources. Is there a way to get it to return streaming results available in my country?
- Yeah, for sure. I actually just found out about it yesterday. I’ve tried contacting the founder to see if I can include some of those search results into the mix. In fact, just added it in the code so it will go live next time I update.
- Yeah, good call. This is probably best added with a plugin. There are APIs out there that provide this data, so I can at least start with that to provide the streaming services that show the series or movie in your country. I’ve just open sourced the code and first focus will be to make it easier for others to install and run the code so that anyone can add these kind of extensions. They can even be vibe coded pretty easily because the task is pretty constrained. This has actually been very much my focus during development, to add a lot of extensibility that is structured in a way that AI can semi automatically write hundreds of plugins.
Thanks for the feedback!
Cool, looking forward to those features
where do i find those settings?
The gear icon down at the bottom, under “Continue your search…”
thank you
I swapped to it on my phone browser to give it a try. So far I’m liking it. Reminds me when Google used to be good.
I wonder if there’s a way to rank fully LLM generated websites lower than ones with content written by a person.
This is certainly the goal. I’ve started crawling and analyzing about 16M domains and I’m planning to take screenshots of all to have AI analyze what it actually looks like. Does it have a lot of ads and other crap or is it clean and easy to navigate, etc. And also to analyze the content to see if it looks like useful original content or AI slop. The great thing about being small is that everybody is optimizing for Google, not for NeoSearch. The idea is to ditch the traditional SEO pagerank, etc way of ranking because now we have AI that can actually read and understand the content, so it can determine the quality and how well it matches searches, etc. Anyway, early days, but this is the goal for sure. I’m sooo sick of almost all Google results coming from the same 100 big platforms.
I don’t really search often but I had to try this. I instantly changed it as the default search engine on my phone where I do most of my searches.
Thank you!! Really well done.
I like the idea but it suffers the same issue I see on a lot of engines like this. The results are just a bit too busy. It do a lot for readability to break up the search segments into different sized text.
I’d also on mobile ensure the search bar has some pill outline around it. It took me to long to figure out how to search
Thanks you, that’s good feedback. This puts theme support higher on the todo list.
Just tried it out! Happy with the results, but I agree some UI tweaking could go a long way. Where to type your search, and the results are too “clean” to the point where it’s a bit confusing.
Great work!
I keep using the search I am impressed it’s so good I am slightly skeptical about tracking.
This will be my daily driver I think …
I love the idea
Looks like Google Custom Search API is deprecated? https://developers.google.com/custom-search/v1/overview
It’s fast, polished, really quite impressive. Maybe too AI-centric for my taste, and yet I like how the results are presented, and they were quite relevant at times… and bad too. For example I was searching for chatons.org (software) and it only gave me results for kittens* (chatons in french). While recognizing I was in France but giving all answers in English.
I wonder though, if I want to search something simple, say I forgot the exact URL of a website, how much ressources are being wasted explaining what the website does instead of simply giving me a link?
I like how Kagi does it, when I need a short AI answer I add a questionmark to my query.
I’ll keep testing this, it looks very promising. Best of luck to you!
*And the AI saying kittens are hairless at birth? I don’t use google so I assume this is the type of answers it gives, man that’s terrible
I must ask because I always ask: why Apache2 and not GPLv3?
Like many devs, this may not be your priority, but GPL protects the users and developers, no?
Yeah, you got it, it’s not a priority. The way I understand it is that with Apache2, anyone can pretty much do anything with it. With GPL, any derivative work must also be open sourced under GPL. Which one is better for the open source community is something I have no opinion on, I guess there are good and bad arguments either way. So I just picked for less restrictions, because I don’t like restrictions. Also, and I might be wrong about this, but once we’re GPL and contributions are added under GPL, it would be an uphill battle to switch to Apache2. But we can switch to GPL quite easily from Apache2 in the future. But yeah, mostly, it’s just not something I have a strong opinion about, so I picked Apache2 because it’s easy.
Loving it so far! It’s very fast and I love all the search engine options, I’ll definitely be giving his a try! Thank you so much!
Maybe I’m missing it though, is there a way to turn off the ai overview completely?
You’re the second person asking for this today, so this will be an option soon.
I usually do some test searches using basic terms (e.g., “dog”), and although it doesn’t show many results, the ones it does show are the most logical. I’ve seen links that make me wonder why they don’t appear in other search engines, because they’re really good.
Hard to notice when you first use NeoSearch, but at the top there are “Lenses”. There are more results hiding in those. But most of the time, what you’re looking for will be in the first lens.
and it uses AI to improve results
Aaaaaaaand this is where I stopped reading
Every major search engine is using it even Kagi. It’s going to be hard to avoid
So why contribute to the slop? The other companies have way more funding to do the normal thing. If it’s an indie project it’s better to differentiate yourself
Yeah. What do people think a “search engine” is.
Im no expert. But I believe there can be a difference between running a proprietary algorithm and using an external LLM for your indexing.
