The 72 Spirits of Goetia

Working with the seventy-two is easier once you stop treating the list as a menu of monsters and start treating it as a filing system. That is what it was built to be. The material below takes the question apart in the order a working actually needs it answered: which entry fits the question, what the rank means in practice, how a seal is constructed, what the sequence of a working consists of, and what to write down afterwards. It is written for someone who has decided to try this properly rather than browse it, and who wants the structure clear enough that a result, or a failure, can be attributed to the right cause. Nothing here requires belief about what the entries are, and the sequence works the same whether you regard them as independent figures or as a useful catalogue of psychological patterns.

The list as a classification problem

The seventy-two are organised by rank and grouped into families, and almost every practical difficulty with the material comes from treating the grouping as decorative. It is not decorative. It is the index. An entry's rank tells you its standing, its family tells you its neighbours, and its stated office tells you what it is for.

Because the offices overlap heavily, a person asking a narrow question will usually have several defensible candidates and no single correct one. Deciding between two entries whose descriptions both cover your situation is a judgement about your question, not a lookup. If you want a single right answer you will be disappointed. If you want a defensible one you will have several.

The failure mode is choosing by recognition instead. Names travel further than descriptions, and a widely known entry gets chosen for a narrow domestic question that a mid-rank functional entry was written for. The working then does something reasonable and unrelated, and gets logged as a failure of the entry rather than a mismatch in the selection.

What rank actually changes in practice

Rank matters less than beginners expect and in a specific way. It does not determine power, because the classical text makes no such claim and the modern brand invented one. What it determines is how much of the entry's description is administrative detail and how much is a distinctive office.

The top rank entries have broad, glamorous-sounding domains. The middle ranks have narrow, workmanlike functions. The lower ranks tend to describe specific operations rather than spheres of influence. If your question is a sphere of influence, a top-rank entry is a reasonable default. If it is an operation, the narrower entry is usually a better fit and far less likely to be selected for its reputation.

The practical advice is to select on the wording of the office, not on the rank and never on fame. Write the question down first, find the entries whose description actually contains it, and only then consider rank. That order of operations prevents most of the mismatches that make a working look like it produced nothing.

Where the material came from, and what that limits

The lineage runs back through two channels, a conjuration text printed in the seventeenth century and the older Lemuria. A good deal of the name material goes further still, to a category of angelic being recorded in Jewish and Christian demonological writings including the Pseudepigrapha Book of Enoch. The compilation as an inventory of seventy-two ranked entities is a comparatively late arrangement.

That matters for two reasons. The list is a synthesis rather than a survival, so its internal structure tells you about the compiler and his period rather than about whatever the names originally referred to. And the modern branded form, dominated by video essays, commentary and apps, is a further layer on top of that, added in this century.

None of this makes the material useless. It makes it legible. A text built as a classification will behave like a classification, and expecting anything else from it is how practitioners end up with a working that technically succeeded and practically did nothing.

Step one, write the question before the name

This is the step that decides the outcome and the one most often skipped. Before choosing an entry, write the question you are actually asking, in one sentence, with the kind of outcome named. Not a state of mind. A specific situation, named, with a timescale.

The sentence does not have to mention an entity at all. If the sentence cannot be written, the question is not yet specific enough to work with, and any entry you pick will be a guess. If the sentence names a person, stop and read the ethics section of whatever you are about to do, because that is a different category of working.

A good sentence is testable in the sense that someone else could read it and say whether it had been achieved. If nobody else could say, you will not be able to say either, in six weeks, which is when the memory has become unreliable.

1. Write the question as one sentence with a named outcome and a timescale.
2. Check whether the sentence names a person, because that changes the category entirely.
3. Keep the sentence somewhere you will find it again in six weeks.

Step two, select on office rather than reputation

Take the sentence you wrote and read the stated descriptions of candidate entries until you find ones whose wording contains your situation. Ignore how well known any of them are. Note which two or three fit best and keep those notes, because you will want the rejected options later.

Then read the rejected options honestly. The reason you rejected each one is the same reason it would be a poor choice, and that reasoning is worth more than the entry you did select. It is the part of the process that generalises to the next working.

If only one entry fits, take it and move on. If five fit, you have a vague question and should go back to the first step rather than picking the most famous of the five. The number of matches is a useful diagnostic and it costs nothing to compute.

Step three, drawing the seal

The seal is a piece of conventional graphic apparatus. Its job is to bind a name to a particular occasion and prevent the general name from being invoked every time you say it. That is its whole function, and it is a practical one rather than a mystical one.

Draw it yourself if you can. A hand-drawn seal carries something a generated one does not: mainly the state you were in as you drew it, which is useful later when you read the log. If you generate it, generate it once and then keep the file, because a seal you cannot reproduce is not a record of anything.

Whatever you do, do not skip it and do not improvise it per working. A seal is a fixed object tied to a fixed question. Rebuilding it each time makes every working a new working, which removes the possibility of comparison and with it most of the learning.

Step four, the contact itself

The classical sequence is long and its stages are not optional to the tradition, but they collapse cleanly into three parts: an opening that establishes the terms, a call that states the question and requests the office, and a closing that ends the contact and dismisses it properly.

The opening is about containment rather than ceremony. You are setting boundaries on the contact, and the classical form is verbose because a verbose frame is hard to argue with later. A short modern version works, provided it still says what you want, what you are offering, and how long you intend to be in contact.

The call itself is the part practitioners vary most and understand least. It should name the office you selected and repeat your sentence nearly verbatim. Do not improve the wording at this stage. A request that has drifted from the logged question is a request you cannot evaluate afterwards.

Step five, close and record in the same sitting

The closing is not decoration. Leaving a contact open is the single most common way a working turns into something the practitioner later has to work around. The classical material is emphatic about this, and the emphasis is practical rather than doctrinal.

Record the entry the same day, while the details are still available, and include the things you would otherwise forget: the time, the state you were in, the physical sensation if any, and the first thing that came to mind afterwards. Write the null details too, not just the notable one. A log that records only the strange entries will tell you nothing a year later.

Then set the review date before you stand up. Not the date you will review it, which will be never, but a date in a fixed number of weeks when you will look at what you wrote and ask whether the question got answered. If nothing was recorded, the null is the answer.

The classical office, and why the media framing misleads

In the original scheme these were administrative or disciplinary officials with assigned jurisdictions. That is a much less dramatic picture than the one that circulates, and it is a more useful one, because an official is something you can make a request of in ordinary language and hold to a stated function.

The media framing pushed them out of that context into the role of torturer or adversary, and a practitioner working inside that framing will build an entirely different apparatus than the one the text supports. Protection rites with heavy defensive language, contracts framed as threats, and an expectation of resistance are all consequences of the wrong reading.

If you intend to work with this material, read at least one source that treats the entries as offices rather than as adversaries, and notice how much of your planned ritual changes when you do. In my own practice the change was immediate and it made everything shorter, quieter and easier to record.

What a first series actually produces

Most first workings produce very little that could be called contact. What they reliably produce is a sharpened question, which is not nothing. Writing a narrow sentence about a specific situation forces you to notice that you had not decided what you wanted.

Some people report a sense of presence, an image, a change in mood, or nothing at all. All four of those belong in the log with equal weight, because the log is a record of what happened rather than of what was meaningful. Filtering at the point of recording is how a log becomes an argument rather than a record.

The working that reliably produces a usable result is the second or third one, on a question small enough to notice an answer to. The first working is nearly always spent learning the sequence, and treating that as a failure is how people conclude the material does not work.

A selection sheet to fill in beforehand

Before any working, fill this in and keep it with the log. It takes five minutes and it converts an operation that feels vague into one that can be evaluated.

• The question, in one sentence, with a named outcome and a timescale.
• The three candidate entries and the office wording that matched each one.
• The one entry selected, and the one sentence naming the reason the others were rejected.
• The terms of the contact, including the duration and what you are offering.
• The date the review happens, written before the working rather than after it.

The stopping rule, and what to write at the end

Stop when the review date arrives, not when you feel you have got what you came for. The second rule produces a different log every time, and a log that is shaped by how you felt cannot be compared with the last one.

Reduce the intensity rather than stopping outright once the routine begins to cost you work or relationships. Escalating in response to a disappointing result is the most common and the most avoidable pattern in this material, and a shorter daily practice is almost always the better response than a longer one.

If two consecutive cycles produce nothing you can describe, the problem is almost never the entry. It is the question, the timing, or the fact that you are treating an event as a relationship in which something can be asked. Audit the selection first, then the question, and only then the entity.

The parts worth keeping for a year

What survives a year of this is smaller than people expect and more useful than what they hoped for. It is a set of written questions that are sharper than the ones you started with, a log that can be read without embarrassment, and a working method you can run without inventing anything.

What does not survive is the sense that the entries are powerful in a way the text does not claim. That impression comes from the framing around the material rather than from the material, and it makes the practice harder to keep doing honestly.

The other thing worth keeping is the ability to change your mind. The list is a map, not a territory, and a map that has stopped being updated becomes a way of confirming what you already believed. Reading one new source per quarter is enough to keep it honest.

Frequently Asked Questions

How do I pick a spirit when the ranks seem arbitrary?

Ignore rank for selection and use the wording of the stated office. Write your question as one sentence with a named outcome and a timescale, then find the entries whose description actually contains that situation. Rank tells you how much of the entry is administrative detail and how much is a distinctive sphere, which is useful afterwards but is not the selector.

What are the nine Presidents actually for?

Their stated domains are broad spheres of influence rather than operations, which makes them a reasonable default when your question concerns rulership, teaching, discovery or investigation in general terms. If your question is a narrow operation, a mid-rank functional entry will usually be a better fit and far less likely to be chosen for its reputation.

Are these entities demons in any real sense?

Not in the way the popular material implies. The names reach the grimoire through a seventeenth-century conjuration text and the Lemuria, with much of the older name material drawn from Jewish and Christian demonological sources. In the classical scheme they held office with assigned domains, so calling them demons describes where the material came from.

What actually happens on a first working?

Most first workings produce very little that could be called contact. What they reliably produce is a sharper question, because writing a narrow sentence about a specific situation forces you to notice that you had not decided what you wanted. The working that usually produces something usable is the second or the third.

What am I paying for when I use a grimoire app here?

The print and manuscript material is public domain and the reasoning in this piece is free. A paid browser earns its place on logistics: sealed entries that keep the working question and the rejected candidates on one screen, the selection sheet, the review date that holds you to your own stopping rule, and a saved history you can compare a second cycle against.