The Tweet That Changed My State

Quick Answer

~Cortisol Checker~ began with a subjective observation, not a scoring system: I could read one post and feel my attention tighten, then read another and feel it settle. The first question was not whether a machine could know my biology. It was whether the language patterns preceding that change could be made visible.

What the checker does

~Cortisol Checker~ is a research tool for inspecting stress-associated patterns in a piece of text. A reader pastes a post or passage; the checker applies the same versioned rules to everyone; and it returns a score with the evidence and component breakdown that produced it.

The public analysis is anonymous and transient. It does not use identity, retain the submitted text, call a language model, or measure a hormone. A verified participant may later add their own separate impact response; that self-report does not change the content score. Run the current research beta.

Before the checker: what cortisol is

Cortisol is a steroid hormone made by the adrenal cortex. It is not simply a “bad stress chemical”: it participates in ordinary daily regulation, metabolism, immune activity, cardiovascular function, and adaptation to challenge. Cortisol also follows a circadian rhythm, generally rising toward waking and declining across the day. The NCBI overview of cortisol physiology describes these functions and the daily rhythm.

When a situation is appraised as sufficiently demanding or threatening, one route involved is the HPA axis. The simplified flow is:

In fuller terms, the hypothalamus releases CRH, which prompts the anterior pituitary to release ACTH. ACTH then signals the adrenal cortex to produce cortisol. Circulating cortisol helps mobilise resources and also feeds back to the brain and pituitary to regulate further release. This is an adaptive system, not a one-way pipe from “negative words” to a hormone. Endotext’s stress-system chapter describes the HPA axis and its regulation.

Psychological stressors do not produce one universal cortisol response. A meta-analysis of 208 laboratory studies found especially reliable responses when tasks combined uncontrollability with social-evaluative threat, while responses varied widely across tasks and people (Dickerson & Kemeny, 2004). That variation is one reason the checker analyzes language patterns rather than claiming to predict a reader’s hormone level.

The internet usually describes a post by what it says, who published it, and how people reacted. My experience suggested another layer. A post also arrives as an event in the reader's time. It can introduce urgency, threat, shame, outrage, uncertainty, evidence, nuance, or a cue to recover. Those qualities do not guarantee a particular response, but they shape the conditions under which a response becomes more or less likely.

“Cortisol” was the first intuitive word for this experience. It gave the idea a memorable handle, but it also carried a risk: software analyzing text cannot measure a hormone in a body. The name survived as the name of the expression. The claimed measurement did not. The current output is Estimated Language Stress Load, accompanied by an explicit statement that it is not medical advice or a cortisol measurement.

That distinction is foundational. A score describes matched properties of submitted language under a published deterministic rubric. It does not state what happened inside a particular reader. When the system later added a personal response, that value had to come from the participant rather than being inferred.

The first inquiry therefore split into two durable questions:

1. What stress-associated patterns are present in this text?
2. What did this text do to me, here and now?

The checker answers the first. The represented observer answers the second.