Shora is an AI meditation app. It is also the only one we know of built by someone who shipped the write-it-on-the-fly version first, watched it backfire, and deleted it.
An AI meditation app uses software to pick or produce a guided meditation for your current state, instead of leaving you to browse a library. There are three kinds. Some generate a brand new script on demand. Some use AI only to recommend a pre-recorded session. Shora is the third kind: it composes a session from a fixed set of published techniques, choosing which ones you need and in what order.
That third category barely gets talked about, and the difference between the three is the whole subject of this page.
The first Shora did the obvious thing. You wrote a journal entry, and the meditation was about what you wrote. Your worries, in your session, in a calm voice. On paper that is the most personal meditation anyone could give you.
In practice I sat there with my own problems read back to me, and my mind stayed locked on them for the whole session. Which is the opposite of the point. A guided practice is trying to quiet that loop, and I had built something that fed it instead.
So the journal entry is now used for exactly one thing. Reading your mood. It picks the session. It never becomes the script.
Every app selling “a meditation about your exact situation” is selling the version we threw away.
Here is every moving part of Shora and what is actually running in it. We have not found another app in this category willing to publish this, which is probably the point.
One footnote, because leaving it out would be the same vagueness we are complaining about. AI did help build the library. While we were authoring it, some techniques were scored against our emotion model with AI assistance. That work is settled into a lookup table now, and it does not run when you press start.
Twenty techniques, all of them from published protocols. Cyclic sighing, box breathing, 4-7-8, NSDR, body scan, loving-kindness, cognitive reframing, five-senses grounding, and twelve more.
Each carries an evidence grade of Strong, Moderate or Emerging, and a link to the paper behind it. Five minutes a day of cyclic sighing improved mood and slowed breathing more than mindfulness meditation did, in a Stanford trial published in Cell Reports Medicine in 2023. You can read every source we use without taking our word for any of it.
That grading matters more than it sounds. It is the difference between a technique that has been tested and a sentence that sounded plausible to a language model.
Open an app that writes your session on demand and you will wait. In the ones we tried it was around thirty seconds before the first word, sometimes longer, staring at a loading state while a model writes a script and a voice engine reads it back.
That is not a bug anyone forgot to fix. It is what generating costs. If the words do not exist until you ask for them, something has to make them while you wait.
Shora’s audio already exists. The voices are synthetic, and because the scripts are fixed we render every line of every technique ahead of time, in each voice and language we offer. That is the trade we made on purpose: you give up a voice that was in a room with a microphone, and you get a session that starts the moment you press start, sounds the same every time, and exists in Romanian as readily as in English. Thirty seconds is a long time when you are having a panic response.
There is a real criticism of this category, and it is worth reading rather than dodging. In December 2024, three psychologists at the University of Southampton wrote in The Conversation that if it is unclear what source material an app used to generate its meditation guidance, it is going to be equally unclear how well it will help its users. They also pointed out that meditation done badly can produce negative emotions, self-criticism or hypersensitivity, which is why expert involvement matters.
We think they are right, and that criticism is exactly why Shora is built the way it is. Our source material is not a mystery. It is a list of published protocols you can read, graded by the strength of the evidence behind each one.
Two honest notes on the rest of it. Their article is specifically about eyes-open meditation apps, which is a different practice from the mostly eyes-closed protocols Shora uses. And Shora is a wellness practice, not treatment. If you are struggling, a licensed professional is the right call, and Shora can sit alongside that rather than replace it.
Probably better than a library, because the hardest part of starting is knowing what to pick, and that is the part we took off your hands. Every practice is explained as you go, with a brief refresher each time you meet it again.
It also gets quieter as you go. A looped breathing segment starts with the full instruction, then drops to “continue on your own”, then to a soft one-line nudge, then to nothing at all. Do a technique enough times and the walkthrough at the start is replaced by the short version. The teaching is meant to get out of your way, which is hard to do when a session is written fresh every time and has no idea whether you have done this before.
Anything mystical. Shora will not float you out of your body, and we are not going to pretend otherwise. You check in, you practice, and you feel a bit better. Or you don’t, and Shora learns from that too.
It also does very little if you only open it when things get bad. This works like a habit, not a rescue. The longer version of why I built it covers what else we got wrong.
The rest of the questions, on pricing, privacy and the science, are on the FAQ.
An AI meditation app uses software to pick or produce a guided meditation for your current state instead of leaving you to browse a library. There are three kinds. Some generate a brand new script on demand. Some use AI only to recommend an existing session. Shora is the third kind: it composes a session from a fixed set of published techniques, and the only place a language model runs is reading what you wrote about your day.
It depends entirely on where the words come from. The techniques we use work because they have been tested and shown to work. When a language model improvises a script, the specific words you hear have never been through a trial. Shora only uses techniques with published research, and each one carries an evidence grade of Strong, Moderate or Emerging.
Psychologists at the University of Southampton, writing about eyes-open AI meditation apps, raised a concern that applies more widely: if it is unclear what source material an app used to generate its guidance, it is equally unclear how well it will help. That is a fair criticism of scripts written on the fly. Shora answers it by publishing its sources. Shora is a wellness practice and not treatment, and it is not a substitute for a licensed professional.
People do. The scripts are written from published protocols and curated by a meditation specialist working with a small team, not generated by a model. AI is used around them, never to write them: to score drafts so changes stay coherent and nothing regresses, and to draft translations that a human then reviews against a glossary. Cyclic sighing, box breathing, NSDR, body scan, cognitive reframing and fifteen more, each linking to the research it came from on our Science page.
It is arguably better for beginners than a library is, because the hardest part of starting is knowing what to pick. Shora explains each practice as you go, then says less and less. The guidance thins out inside a session as you repeat a breath, and thins out again once you have done a technique enough times to know it.
