Stop Searching by Keyword, Just Ask Your Saved Stuff a Question

"I have used WhatsApp but searching is a pain." "Even when I KNEW it was in Evernote, I could not access it." These are real people describing the same quiet defeat: they saved something, they know it is in there somewhere, and the search box hands them nothing useful.

The usual conclusion is that you tagged it wrong, or filed it in the wrong place, or picked the wrong app. But the real issue is smaller and more human than that. The word you would search with today is almost never the word you saved it under. You remember the gist, the moment, the reason it mattered. You do not remember the exact phrase from the title. And keyword search only rewards the exact phrase. This piece is about closing that gap: why finding things by description in your own words works when keyword search does not, and how to set things up so you can just ask.

You are not searching, you are re-finding

Most of the searching you do is not discovery. It is going back to something you have already seen. In an analysis of a full year of real search logs from 114 users, plus a controlled study of 119 more, researchers found that as many as 40 percent of all queries are re-finding queries: searches that lead you to click a result you had clicked before. For some people in the logs, more than 80 percent of their searches were re-finds (Teevan, Adar, Jones and Potts, 2007, Proceedings of SIGIR).

That number reframes the whole thing. A huge share of the effort you spend in a search box is spent trying to get back to something you already found once. And here is the detail that matters most for anyone with a pile of saved stuff: the same study showed that the query you use to re-find something is often different from the query you used the first time. You do not re-run the original search. You describe the thing again, fresh, from whatever you happen to remember this time.

So the system is fighting you at both ends. You saved the thing with one set of words, you go looking for it with another, and a keyword index quietly requires the two to match.

Why keyword search keeps coming up empty

A keyword search is a string-matching exercise wearing the costume of a question. You type a word, and it returns the items that literally contain that word. If you saved an article titled something dry and corporate, but you remember it as "the one about why teams burn out on Fridays," the match never happens. The information is right there. The words are not.

This is why so many people describe their own saved archives as unusable. "The search is broken and I can't find anything like I used to." "It would NOT show info within notes that I was searching for." The complaint is rarely that the thing is gone. It is that the thing exists but refuses to surface under the words a normal person would actually use to call it back.

The other half of the problem is scale. You do not have one search box, you have a dozen. As one person put it, the thing that kills every system is that they save across too many different apps and there is no single place to search all of them at once. So even when you do remember a good keyword, you have to guess which silo it landed in, and search each one separately, hoping the index in that particular app is the one that works.

The alternative: find it by description, not by keyword

There is a different shape for this. Instead of matching exact words, you describe what you are looking for the way you would ask a friend who was paying attention. Not "invoice_template_v3," but "that spreadsheet a coworker sent me for tracking expenses." Not a remembered tag, but "the apartment listing with the weird kitchen, somewhere downtown." You hand over the gist, and the gist is enough.

This matters because the gist is exactly what your memory actually keeps. We are good at remembering meaning and context and bad at remembering precise strings. A retrieval system that meets you at the level of meaning is working with what your brain stored, rather than demanding the one thing your brain threw away.

It also dissolves the scrolling. The thing that sends people to a search box in the first place is often a visual pile: hundreds of near-identical thumbnails, screenshots that all look the same, a camera roll that has become, in one person's words, a graveyard. Describing the one you want by what it actually was beats scrolling hundreds of look-alikes by hand, every time.

And because you are describing meaning rather than matching a string, you get a second chance every time you forget. If the first description does not land, you try a different angle: the person who sent it, the week it happened, what it was about. With keyword search, a missed word is a dead end. With description, you are just rephrasing the way you would in a normal conversation, and each rephrase narrows in instead of starting over.

How to actually ask your saved stuff a question

This is what dEssence is built to do. You save anything from anywhere: a page, a screenshot, a PDF, a voice note, an article, forwarded straight from Telegram, your browser, or the web app. You do not tag it, file it, or name it for future-you. You just keep it.

Later, you ask in plain language. "The thread where someone explained the visa thing." "That recipe with miso, not the other one." "The article about why teams burn out." dEssence matches on what you mean, not on whether you guessed the exact title, and hands the thing back. Because so much of searching is really re-finding, it also resurfaces saved things on its own, so the item you cared about comes back to you instead of sinking out of reach.

And it solves the dozen-silos problem by being the one place those silos point to. Everything you save lands in the same searchable space, so there is a single question to ask instead of a dozen boxes to check. It also works wherever you already think, so you can pull a saved thing into a conversation with ChatGPT, Claude, or Gemini without going to dig for it first.

What changes when asking is enough

The shift is quiet but total. You stop bracing for the search box to fail. You stop saving things under careful keywords you hope future-you will reconstruct. You stop running the same re-find search three different ways before giving up and re-Googling the whole thing.

The 40 percent of your searching that is really re-finding stops being a chore. The thing you saved was never the problem. The exact-word gate in front of it was. Take that gate away, let a description be enough, and the archive you built finally answers when you ask it.

FAQ

Why does keyword search fail even when I know the item is there? Because keyword search matches exact words, and you rarely remember the exact words you saved with. Research on real search logs found the query people use to re-find something is usually different from the one they first used, which is exactly where keyword matching breaks.

How is asking in plain language different from a normal search? A normal search returns items that literally contain your word. Asking by description matches on meaning, so "the spreadsheet for tracking expenses" finds the file even if its actual name was something you would never type from memory.

Does this work across different apps and file types? Yes. Pages, screenshots, PDFs, articles, and voice notes all save into the same place and come back the same way, by asking, so you are not searching a dozen separate boxes one at a time.