Writing a Book with AI #7: Pattern Recognition
what to leave in and what to leave oute AI Sharpens, Not Softens
This is the sixth article in this series :
Writing a Book with AI
Less than a year ago, like many authors who are proud of their writing prowess, I snubbed my nose up at AI. Since I started exploring if AI could be taught to meditate, my opinion of AI as a writer’s tool has completely changed.
There comes a point in writing a book when the manuscript stops being a pile of words and starts behaving like an organism.
At first, all you want is material. Notes, fragments, chapter sketches, stray paragraphs, peculiar metaphors, half-formed arguments, things heard in the shower, things whispered by the dog-eared gods of the notebook. You gather, generate, dictate, paste, prompt and ponder. The book is not yet a book. It is compost with opinions.
Then, somewhere along the way, patterns begin to appear.
An idea returns three times in different clothing. A metaphor keeps tapping at the window. A paragraph in Chapter Two starts arguing with Chapter Nine. A concept appears before the reader has been prepared to meet it. A beautiful digression turns out to be not a digression at all, but the seed of another book.
This is where AI can become extraordinarily useful. Not as a ghost-writer. Not as an oracle. Not as a replacement for judgement. More as an aerial cartographer.
It can look down over the terrain and say, “This river begins here but vanishes later.” Or, “You have explained this twice.” Or, more interestingly, “This material appears to belong to a larger work.”
Pattern recognition is not editing yet. Editing asks, “How can this be improved?” Pattern recognition asks, “What is this becoming?”
I discovered this very clearly during the writing and refinement of The Hidden Tarot in AI. At first, the book had a distinct and compelling centre: the Augmented Major. It explored the 22 traditional Major Arcana, and the way Tarot, archetype and artificial intelligence could be brought into meaningful conversation.
That was already enough.
Of course, manuscripts do not always respect our tidy borders. As the book developed, more material began to arrive. There were the Half Keys. Then the Augmented Minor. Then the possibility of dimensional suits. Then the Tree. Then the Cube. Then the wider correspondences between Tarot, symbolic intelligence and consciousness. Later still came the speculative idea that the Major and Minor Arcana might be explored in relation to DNA.
Each idea had energy. Each one sparkled. Each one arrived wearing a small hat that said, “Surely there is room for me too?”
There was not.
Or rather, there was room in the wider field, but not inside that first book.
This is one of the hardest disciplines in authorship: recognising that a good idea can still be in the wrong place. Writers often imagine that editing means removing weak material. Sometimes it does. More often, mature editing means relocating strong material so it can breathe.
AI helped me see this. Again and again, it became clear that the manuscript was not asking to become a bigger single book. It was asking to become a sequence.
Book One needed to stay focused on the Augmented Major. If it tried to carry the Minor, the Tree, the Cube and DNA correspondences as well, it would have become a metaphysical attic: fascinating, yes, but with too many boxes falling on the reader’s head.
So we drew a line.
The Augmented Minor was not abandoned. It was recognised as Book Two. Its nine dimensional suits, chromatic logic and divinatory possibilities deserved their own space. They were not supporting material. They were a new field of inquiry.
The deeper architecture, including the Tree of Life, the Cube of Space and the later postulation of Major and Minor correspondences to DNA, became the territory of Book Three. Importantly, that word matters: postulation. Not claim. Not proof. Not biological assertion dressed in occult velvet. A future symbolic exploration, held with care.
That distinction preserved the integrity of the first book.
This, for me, is the real gift of AI-assisted pattern recognition. It does not simply help you find repetition. It helps you notice scale. It can reveal when a paragraph is trying to become a chapter, when a chapter is trying to become a book, and when one book is quietly dividing itself into three.
But the final decision must remain human.
AI can say, “You mention this theme seven times.” Only the author can decide whether that repetition is laziness, emphasis, rhythm, ritual or revelation. AI can say, “This section feels out of scope.” Only the author can sense whether it should be cut, moved, expanded or kept as a deliberate disturbance in the fabric.
Pattern recognition is therefore not a mechanical stage. It is a listening stage.
You listen for what repeats.
You listen for what is missing.
You listen for what is arriving too early.
You listen for what belongs elsewhere.
The temptation, especially with AI, is to generate endlessly. To keep asking for more. More angles, more chapters, more examples, more cleverness. But a book is not improved by containing everything. A book becomes itself by learning its own boundary.
AI can help us see that boundary.
It can say, “This belongs, but not here.”
Sometimes that is the most important editorial sentence of all.
Because writing a book is not only the art of what to include. It is the art of giving every idea its right home.