Can an Astrology App Pass a Blind-Reading Test? Our AI Experiment, Results, and What They Mean

We ran a blind astrology experiment: give an AI only our app’s structured engine output—no names, no biographies—and ask for a predictive synthesis and a best-guess identity. Could it find real signal?

TL;DR

We tested whether our app’s engine (which uses Morin-style determinations) produces signals strong enough for an AI to: (1) write a concise, predictive reading from the structured output alone, and (2) guess the identity behind the chart.
Across three trials, the AI produced accurate, house-level predictions and guessed correctly twice (user-confirmed) and missed once—the miss was later revealed as Adolf Hitler. Even so, the reading still matched core biographical themes.
The results show face validity and early discriminant power for both astrology and our engine: the structured outputs consistently led to distinctive, testable narratives.

Why a blind test?

Most skepticism claims astrology “fits anyone” because readers know the subject. We removed that bias: the AI received only the engine’s structured output—no name, no biography, no hints—then produced a reading and a guess. If the reading aligns with real life and the guesses sometimes hit, we’re seeing genuine signal.

The setup (simple and fair)

Inputs to the AI:

Planetary locations by house (highest priority)
Rulership routes (e.g., Ruler of 10 in 7)
Aspects to house cusps with phase (applying/separating), partile strength (≤1°), and dexter/sinister notes
State coloring: dignity/debility, combustion/under the beams, retrograde, angularity
Receptions (mutual/mixed), plus planet strength tables and top traits

What the AI had to do:

Produce a concise, predictive synthesis prioritizing Location → Rulership → Aspectual (the same logic our engine uses).
Offer a best-guess identity based solely on the output.
Accept confirmation/denial from the user.

Controls: No web searches. No biographies. No “leading” hints.

What the AI actually saw (tiny excerpt)

{
  "planet_strengths": {
    "top": [
      {"planet": "Sun", "house_position": 5, "total": 26.41},
      {"planet": "Moon", "house_position": 0.5, "total": 22.6},
      {"planet": "Mars", "house_position": 5, "total": 21.84}
    ]
  },
  "houses": [
    {
      "house": 7,
      "synthesis": [
        "Sun presses on House 7...",
        "Mercury presses on House 7..."
      ],
      "basic_analysis": {
        "location": [
          {"planet": "Sun", "summary": "Sun in House 7... [afflicted]"},
          {"planet": "Mercury", "summary": "Mercury in House 7... [Combustion, afflicted]"},
          {"planet": "Venus", "summary": "Venus in House 7... [dignified, Under the Beams]"}
        ]
      }
    },
    {
      "house": 10,
      "determinators": {
        "aspect": {
          "planet": "Sun",
          "aspect": "Square",
          "phase": "separating",
          "phrase": "waning; partile (very strong)"
        }
      }
    }
  ]
}

Results (with confirmed identities and why the engine pointed there)

Trial 1 — Donald J. Trump (correct)

Why the engine pointed there (highlights):

Location first: Sun occupying H10 (career/status) with the North Node also in H10 → public life is the life. Mars applying to Cusp 1 → combative, force-forward presence. Moon in H4 + Saturn trines to angle cusps (partile) → heavy family/legacy tone with austere framing.
Rulership: Ruler of 1 in 10 → identity fused to office/reputation; ruler chains to H11 → patrons/networks elevate.
Aspectual: Sun square Cusp 2 (partile, applying) → money/values conflict powering public stance.

Trial 2 — Benjamin Netanyahu (correct)

Why the engine pointed there (highlights):

Location: Saturn in H10 (career) + Moon and Mercury also pressing the 10th/11th axis → institutional role, public communication, and networked influence. Mars in H9 → ideological/foreign‑affairs conflict signature.
Rulership: Ruler of 3 in 10 and Ruler of 9 in 11 → public messaging plus allies/patrons in higher‑learning/law/religion.
Aspectual: Mercury conjunct Cusp 11 (applying) → agenda framed through groups/media; benefic trines to social houses steady the coalition theme.

Trial 3 — Adolf Hitler (incorrect guess; later revealed by user)

Why the engine still matched the biography:

Location: Sun + combust Mercury in H7 with dignified Venus under the beams also in H7 → life routed through alliances, pacts, and open enemies; relationship theater as public stage. Saturn in H10 (career) → rigid, punitive office; Mars conjunct Cusp 8 (applying) → escalation via shared resources/crisis/war.
Rulership: Ruler of 10 in 7 → career via partnerships/legal arrangements; Ruler of 8 in 7 → entangled resources through unions and conflict.
Aspectual: Sun square MC (partile, separating) → chronic friction between image and authority; Moon/Jupiter activity in H3 → volatile propaganda/overpromising.

Even when the name guess missed, the structured reading tracked the biography—exactly what we want in a blind test.

Why this supports validity (in plain language)

Face validity: Independent interpretation (the AI) turned our engine’s output into specific, testable claims—not vague platitudes. Users could verify those claims against real lives.
Construct alignment: The same structural levers our engine encodes—location, rulership routes, cusp aspects with phase and partility, plus state coloring—consistently shaped the AI’s predictions.
Discriminant power: Charts with 7th-house routes read nothing like 12th-house routes; a 10th ruled by the 7th behaves differently than a self-ruled 10th. This separation helps explain the two hits and the biographically coherent miss.
Reproducibility: Because the engine output is standardized, any competent reader (human or AI) can apply the same priority and reach comparable narratives.

We’re not claiming “final scientific proof.” We are showing that when an interpreter sees only our app’s output, the results frequently match real-world lives—evidence of meaningful signal.

Limitations & next steps

Sample size (N=3) is small. Next: scale to N≥50 blind trials.
Pre-register rules (one guess, top-3 guesses, time-boxed reading) to prevent post-hoc massaging.
Compare against baselines (random famous-person guesser; non-structured readings).
Add scoring: top-1/top-3 accuracy, plus blind rater coherence (how well a reading matches a life without knowing the name).

Conclusion

In a blind setup, our app’s engine gave an AI enough structured, prioritized signals to produce crisp, house-level predictions—yielding two correct IDs and one miss that still fit the life narrative (Adolf Hitler). That’s early but meaningful evidence that astrology—rendered structurally—and our engine—implemented consistently—carry real, testable signal.