How Your SBTI AI Report Is Generated — Inside the Pipeline

The report does not start with your type code. It starts with your 15-dimension shape — the actual Low, Medium, or High rating you scored on every single dimension of the test. Your primary type (say, The Controller, or The Possum) is the headline, but the underlying L/M/H pattern across Self-Esteem, Self-Clarity, Core Values, Attachment Security, and twelve others is what the AI actually reads.

That distinction matters more than it sounds. Two people who both land on the same SBTI type can still have very different dimension shapes underneath. One Possum might have Low Self-Esteem and High Self-Clarity. Another might have the opposite. Those two people deserve different reports, and our pipeline generates different reports for them. A generic "describe a Possum" prompt would collapse both into the same writeup.

What the AI actually receives is a structured prompt with five slots, one per report section, each fed a specific subset of dimensions that matters most for that section. Career pulls from the Action Drive and Social models. Love pulls from Emotion and Attachment. Growth pulls from wherever your lowest dimensions live. The prompt is long, specific, and numerically anchored — nothing like a one-line "you are THINK-R, write about them."

The generation model is Google Gemini. It reads your dimension shape, the locked section structure, and the SBTI voice guide — a style document that teaches it to sound playful but honest, mirror-like but never mean, and specifically not clinical. The first pass produces a full report in one shot across all five sections.

The second pass is a safety review. This is the part most AI content products skip and the part we care most about. A second Gemini call reads the draft report back and asks: does this cross from "playful mirror" into "amateur diagnosis"? Does any paragraph read like it's telling someone they're depressed, attachment-avoidant, or in a bad relationship? Does any sentence feel like it could hurt the reader if they were having a rough day?

When the safety pass flags something, the pipeline rewrites that section with a softer, more entertainment-first frame before any email goes out. You never see the flagged version. This is invisible work, and it is why refund rates stay low, why we do not wake up to angry emails, and why we feel comfortable charging for the report in the first place.

The entire process is asynchronous. You pay, you see a "your report is baking" confirmation page, and the report arrives in your inbox when it is done. For most people that is between one and three minutes. Sometimes it is faster. Occasionally, if the queue is warm from a traffic spike, it is slightly slower.

You might wonder why we do not just serve the report instantly. We could. Using a faster cheaper model, skipping the safety pass, and returning raw output on the same page is technically simpler and feels snappier. We tried that early and the reports were noticeably worse — generic-sounding, occasionally tone-deaf, and not worth $4.99. The current two-pass pipeline is the trade-off that produced reports people actually screenshot and share.

If you have not taken the test yet and want to see what dimensions it actually measures before deciding whether a premium report is for you, start free at sbti-tests.app. For the full explainer of what the 15 dimensions and 27 types mean, see What Is the SBTI Personality Test?. If you want to see exactly what ends up in the report, the companion article Inside the 5 Sections of Your SBTI Premium Report walks through each section.