All of them involve some level of self-hosting, either running a stand-alone local search engine or deploying them to the Internet. Where I've landed is that I think a very compelling system could be composed of heterogeneous implementations linked together by an easy-to-implement protocol using a "follower" network model that allows a single PSE to "follow" others to share searches and results. You setup your PSE, you follow a few other PSE's that index the kind of pages you want to search and you start to build a nice corpus of resources. It's easy to see how a network like this spreads, especially if the protocol makes it easy for nodes to share a list of nodes they trust.
But this all hinges on each person running a literal instance of their own search engine. This is, of course, kind of the point of a personal search engine, but it's not a trivial ask, especially given that these instances need to be deployed to routable Internet addresses.
While there may be ways to avoid this routeability requirement, all of the alternatives I've explored come with heavy side-effects.
Deploying a piece of software to the Internet and maintaining it are not something I would expect every person to do, in fact it's even a lot to ask of myself, so I don't see this as a path toward any sort of mass adoption of PSE's. That might be fine, people who have the means and really care about how they find things on the web might tolerate this or collaborate with others to get it done, but I couldn't help but feel like there might be another approach that could even work in tandem with the do-it-all-yourself option.
One option I've considered is a PSE hosting service akin to Fediverse node hosting services who would take care of the work for a regular fee. This might close the gap somewhat, but it's still a barrier and it can't completely eliminate the maintenance burden.
This morning it hit me and it was so obvious that I can't believe it didn't occur to me earlier: why not make it work just like a social media site? This is a model that most humans are very familiar with, and adopting it might make PSEs accessible to a much wider audience.
Imagine it like this: You sign-up and add a few URL's you'd like in your personal search index (maybe just a few you paste-in, or upload a bookmark file or RSS reader dump, etc.). The site then gives you a search page at a unique URL where you can search against your personal index created by a crawl of the URL's you provided. The page makes it easy to add this personal search URL to your browser as your primary search engine, and it will work from any of your devices because it's on a public Internet server.
Indexing could be enhanced with a browser plugin or some other means of automatically indexing bookmarks, etc.
Where this gets interesting is the fact that there's other people using the system who are creating their own indexes as well, and the site makes it easy for them to expand their own personal search indexes by following one-another like any other social media application. At the most basic level this could take the form of people directly sharing their unique URL with one-another, or the site could provide some "discovery" tools to find other people who might be curating an index you would like to include in your searches.
For me personally where this would be very powerful is to have a way to search the web without getting results that point to LLM-generated websites. So in that case I could curate my personal index so that no such content is included, and maybe I could "tag" that index with a "no-ai" tag. I could convince my friends who also want to search without AI to join and as they curate their own ai-free indexes I can follow them and expand the range of sites in my search results without risking contamination. Others could benefit from this as well by discovering my index via the "no-ai" tag, and that relationship could work both ways (always optional and opt-in of course).
A system like this could also interact with independent stand-alone PSE's by adopting the same protocol as I originally proposed for stand-alone nodes, creating a sort of hybrid. Under the hood, even the "centralized" system described here would likely be implemented as a federated system, but the user experience would be designed to emulate the behavior of a "traditional" centralized social media site.
Aside from streamlining the adoption and operation of PSE's, there's operational benefits to this approach that make building and operating it attractive as well. Some of the concerns about duplication, network storms and abuse become more manageable in a system like this and having the option to migrate in and out of it to or from a stand-alone instance helps abate the nastier aspects of a centralized system.
To a "normal" person, it's just another website, not a piece of software and equipment that you have to put a lot of effort into or maintain ownership of, but functionally it has all of the advantages of PSE's I've discussed before. The missing ingredient is how to make it sustainable. With a stand-alone PSE, the cost of operating it falls on the person who owns it; but how to cover the costs of operating the shared PSE system described here?
The obvious answers are fees or subscriptions but I don't like how that unfairly impacts people based on income, etc. Advertising is how the social media people do it but of course I have no interest in that, and I won't even begin to talk about covering costs through the evils of personal data peddling and selling surveillance. The idea I like best so far is simply voluntary community support and transparent books, so that the people who use the system can see what it costs to operate it and contribute what they can if the service is valuable. I think a combination of this, along with credit to people who contribute constructively to the quality and value of the system through curation of indexes, support, etc. is the right place to start and honestly a well-written system shouldn't cost a lot to operate based on my experience.
Jason J. Gullickson, 2026