The Stars Don’t Lie (But Your Chatbot Does)
A Guide to Using AI for Astrology—And What It Teaches Us About LLM Limitations
It’s been a minute. Almost two months, actually.
I spent most of the fall out of the country—often in places with spotty internet at best—and started working with a new main client that’s taken up more bandwidth than anticipated. Both made posting difficult. But I’m back, aiming for a biweekly cadence going forward. I’d rather give you something thoughtful than churn out filler.
For now, I wanted to kick off 2026 with something fun.
I. A Confession
Astrology and woo-woo stuff have always been a guilty pleasure of mine.
I know. I’m generally a pretty rational person. I’m not religious. I’ve never made life decisions based on what Mercury was doing. But I’ve always found this stuff genuinely entertaining. The way I think of it is as personalized performance art—a framework for self-reflection wrapped in cosmic aesthetics.
Is any of it real? Almost certainly not. It’s probably complete bullshit. But it’s bullshit I enjoy.
And here’s the thing about bullshit: even when the conclusions are meaningless, the process of getting there can have real technique behind it. Astrology has actual math. Calculating planetary positions, house cusps, aspects—there’s a whole computational layer that most people never see because they just read their Sun sign horoscope in a magazine.
II. Why LLMs Are Terrible at This
So when I see people throw their birth date into ChatGPT and ask it to “read their chart,” I cringe.
Because here’s what I’ve learned from my own experiments: LLMs are genuinely terrible at astrology. Not because astrology is hard to interpret (that’s the easy part for them), but because they cannot do the underlying calculations.
I’ve written before about how these tools can’t even get basic things like astrological signs right and are bad at really basic math. What they’re good at is coding—so instead of asking LLMs to analyze anything quantitative directly, I ask them to help me write code that can do the analysis. The coding assistance is reliable; the direct calculation is not.
A. The Tokenization Problem
The problem is architectural. LLMs don’t process numbers as actual values within a coordinate system—they process them as “tokens,” linguistic fragments that get statistically predicted based on training data. When you ask ChatGPT to calculate your Moon position, it’s not querying an internal ephemeris. It’s guessing based on patterns it’s seen in text.
The Moon moves about 13 degrees per day. The Ascendant shifts roughly 1 degree every 4 minutes. The probability of a statistical model guessing these correctly? Basically zero.
B. The “Vibe Check” Problem
This is what researchers call “hallucination”—the AI sounds confident while being completely wrong. In astrology, this shows up as what I call the “Vibe Check” problem: the AI understands the vibe of a Scorpio Moon (intense, secretive, emotional) and can write beautifully about it. But it cannot reliably tell you whether you actually have a Scorpio Moon.
C. Context Drift
There’s another failure mode that matters for longer astrological analysis: context drift. As conversations get longer, LLMs lose track of variables established at the start. They might confuse a day chart for a night chart, or forget that Saturn is currently retrograde. If you’re asking for a year-ahead forecast, this compounds quickly.
III. The Right Way: Code First, Talk Later
So what’s the solution?
After putting together some research on this, I’ve landed on what I’m calling the “Centaur” approach: use deterministic software for the calculations, and AI for the interpretation. Decouple the math from the meaning.
A. Get Your “Ground Truth” Data
The first step is generating accurate chart data from tools that actually use proper astronomical calculations. These calculators exist for virtually every astrological tradition you might be interested in.
Western astrology: Astro-Seek is probably the most robust free option. The key is to use their “Full Landscape” or “Print” view to get planetary positions as plain text tables rather than just visual chart wheels. AI can parse structured text much better than it can extract degrees from an image. Astro.com (Astrodienst) is another gold standard—it runs on the Swiss Ephemeris, which is the benchmark for astronomical precision.
Chinese astrology and BaZi (Four Pillars of Destiny): Since 2026 is the Year of the Fire Horse, this is relevant. Use dedicated BaZi calculators like Joey Yap or BaZi-Calculator to get your exact Heavenly Stems and Earthly Branches. Don’t trust an LLM to figure out whether you’re a Yang Water Rat or a Yin Fire Snake.
Vedic astrology (Jyotisha): Sites like Lagna360 and Café Astrology’s Vedic Calculator can generate sidereal charts with nakshatra placements and dasha periods.
Aztec astrology (Tōnalpōhualli): The 260-day sacred calendar is still in use among some Mesoamerican communities. Calculators at AztecCalendar.com can determine your day sign (tonalli) and trecena.
Human Design: This newer synthesis of the I Ching, Kabbalah, the chakra system, and astrology has its own calculators at Jovian Archive and myBodyGraph. It’ll tell you your “type” (Generator, Manifestor, Projector, or Reflector) and your decision-making strategy.
B. Feed It to the AI for Interpretation
Once you have verified data, then you bring in the AI. Different models have different strengths.
Claude works well for psychological analysis and synthesizing contradictions in a chart—though it can be verbose and has guardrails that might block doom-laden predictions. Think of it as the “therapist astrologer.”
ChatGPT (currently version 5.2) is stronger at logical reasoning and processing large datasets of transits. It has a tendency toward repetitive structures (”In conclusion...”), so you may need to prompt it to vary its output.
Gemini has deep research capabilities and massive context windows—over 2 million tokens—which makes it useful for mundane astrology (world events) and historical correlation research. The tradeoff is that it can pull in low-quality sources if you don’t curate carefully.
Perplexity Pro is useful for a different purpose: fact-checking specific astronomical events in real-time. If you need to verify an ingress date or check eclipse visibility, it gives short answers with citations.
C. NotebookLM vs. Standard Chatbots
The best approach I’ve found for interpretation is using NotebookLM as a kind of personal research assistant—but it works differently from a standard chatbot, and understanding that difference matters.
With NotebookLM: You upload PDFs of actual astrological texts, and the AI is “grounded” exclusively in those sources. It won’t pull from the open internet’s sea of pop astrology. This creates what’s called a RAG (Retrieval-Augmented Generation) system—the AI retrieves from your documents before generating responses. When you ask NotebookLM to interpret your chart, it pulls directly from the books you’ve uploaded and cites them. The tradeoff is that it won’t know anything outside those sources—if your uploaded texts don’t cover a specific transit, NotebookLM won’t make something up.
The workflow looks like this:
Collect PDFs of authoritative astrological texts
Upload them to a specific “Notebook”
Upload a text file with your 2026 transit data (from Astro-Seek)
Query: “Based only on the uploaded sources, interpret the transit of Saturn conjunct Neptune in my 10th House”
The AI generates an interpretation using your uploaded authors’ voice and philosophy, citing page numbers
With standard chatbots (Claude, ChatGPT, Gemini): These draw on their entire training data, which includes both serious astrological literature and countless listicles about “Top 10 Signs You’re Dating a Scorpio.” You get more breadth but less control over quality.
For NotebookLM, I recommend uploading canonical texts from whatever tradition interests you. For Western astrology, Robert Hand’s Planets in Transit and Richard Tarnas’s Cosmos and Psyche are foundational. The Internet Archive has an extensive collection of free, downloadable astrology books covering transits, horary astrology, and multiple traditions.
D. A Sample Prompt Structure
Here’s the kind of system prompt that works:
ROLE: You are a master Astrologer with 30+ years of experience. You are expert at synthesizing contradictory data into a coherent narrative.
CONSTRAINTS (CRITICAL):
NO CALCULATIONS: Do NOT attempt to calculate any planetary positions. Use ONLY the degrees and signs I provide below.
Be specific about areas of life (Houses) affected. Avoid “Barnum Statements” (vague generalizations that apply to everyone).
DATA: [Paste your text export from Astro-Seek here]
ANALYSIS REQUIRED: [Your specific questions]
The explicit “NO CALCULATIONS” constraint is non-negotiable. Without it, the AI will try to be helpful by computing things itself—and get them wrong.
IV. Advanced Mode: Building an Astrological Pipeline
For those comfortable with code—or willing to learn—there’s an even more powerful approach. Instead of manually copying and pasting between astrology websites and chatbots, you can build automated pipelines that do the heavy lifting for you.
A. What AI Can Actually Do For You
AI tools can identify all your transits for a given time period (sites like Astro-Seek offer personal transit calendars that list every aspect transiting planets make to your natal chart), generate interpretations for each transit, and create calendar events with transit meanings.
B. Model Context Protocol: The 2026 Development
The most significant recent development is the adoption of Model Context Protocol (MCP). This technology bridges the gap between AI’s linguistic capabilities and the deterministic nature of astrological calculation.
MCP allows AI models like Claude or ChatGPT to connect directly to external “servers” or tools without complex custom code. Developers have released specific astrology MCP servers. Instead of pasting data, you install an “Ephemeris MCP.” When you ask the AI “Where is Mars right now?”, it doesn’t guess—it sends a query to the local MCP server, which runs Swiss Ephemeris code and returns the exact degree (e.g., “Mars is at 15° Leo”). The AI then uses this verified fact to generate its response.
Some options to explore:
ScottChronicity Ephemeris MCP: Provides precise Swiss Ephemeris calculations for planetary positions
MCPIztro: A specialized MCP server for Zi Wei Dou Shu (Chinese Astrology), allowing accurate calculation without hallucination
Hermes Astrology Agent: A pre-packaged agent on HuggingFace that combines traditional Hellenistic techniques with transit analysis
This is advanced territory, and I won’t pretend it’s plug-and-play. But for those who want maximum control over their astrological tracking, the tools exist.
V. What 2026 Looks Like (If You’re Into This Sort of Thing)
Since we’re starting a new year, here’s the cosmic context that makes 2026 interesting—and a decent test case for this approach.
I’m not saying any of this means anything. But astrologers who follow mundane cycles (world events, not individual horoscopes) have been pointing at 2026 for years as a convergence point. And when you look at what they’re actually predicting, it maps surprisingly well onto dynamics already visible in AI and tech.
Make of that what you will.
A. Uranus in Gemini: The Large Language Transit
The transit that speaks most directly to AI is Uranus in Gemini, which runs from July 2025 through May 2026.
Uranus is the planet of sudden technological disruption—electricity, revolution, the Promethean spark. Gemini is the sign of language, communication, data, and duality. Put them together and you get what astrologers are calling the “Neural Natal Era” or the moment when AI shifts from tool to conversational entity.
The timing is almost too on-the-nose. Uranus enters the sign of language and communication right as Large Language Models become the dominant interface for how people interact with computers. Astrologer Jessica Adams predicts a “Big Reset” for creative industries—as AI places pressure on words and images, Uranus (the liberator) will “overturn” existing structures, potentially through new copyright frameworks or entirely new ways of valuing human creativity.
Chani Nicholas notes that Gemini is the “trickster” archetype. Her prediction: a massive rise in “digital doubles”—AI agents that act on our behalf—alongside a cultural backlash where in-person conversation becomes a luxury good or even a form of resistance. The duality baked into Gemini suggests we’ll be living with both simultaneously: more AI intermediaries in our communication and a growing premium on the unmediated.
B. Saturn-Neptune in Aries: The Reality Check
On February 20, 2026, Saturn (structure, reality, limitations) meets Neptune (dreams, dissolution, illusion) at 0° Aries—the very first degree of the zodiac, sometimes called the “World Axis.”
This is a rare, epochal conjunction. Saturn demands proof; Neptune offers a mirage. Meeting in Aries (the sign of self and new beginnings), this suggests a crisis of identity—personal and collective.
If you were designing an astrological metaphor for the AI bubble, you couldn’t do better. Neptune is the planet of things that seem real but aren’t: projected revenues, capabilities perpetually six months away, valuations based on what the technology might become. Saturn is the spreadsheet that doesn’t balance, the data center that can’t get permitted, the customer who asks “but what does it actually do for us?”
Astrologer Austin Coppock describes this energy as “the head’s overwhelming hunger for the dream.” He suggests it marks a shift from collective consensus to individual combat-readiness—people having to fight to define what’s real in a landscape of synthetic media and contested facts.
Financial astrologer Ray Merriman, who’s tracked market cycles for decades, views 2026 as a critical turning point. He forecasts continued expansion driven by a “Tech Tonic” but warns of a “bearish Nasdaq divergence” and bubble conditions. The Saturn-Neptune aspect, in his reading, threatens the stability of financial systems built on speculative value—which could apply to AI company valuations as easily as crypto.
Maurice Fernandez connects the Saturn-Neptune cycle to the rise and fall of empires and their currencies. His warning: a potential collapse of “fake” value systems. The generation born during the last Saturn-Neptune conjunction (1989) is now in their mid-thirties, building the systems that will either deliver on the dream or reveal it as vapor.
There’s an optimistic read too. Some astrologers interpret this conjunction as the birth of the “Empathetic Machine”—Saturn as hardware, Neptune as the simulation of soul. Whether that’s exciting or dystopian probably depends on your priors.
C. Fire Horse and Rahu: Speed, Obsession, Disputes
Eastern systems tell a complementary story.
In Chinese metaphysics, 2026 is a Yang Fire Horse year (Bing Wu). The Horse is speed, stamina, unbridled passion. Fire adds volatility and visibility. Previous Fire Horse years include 1966 and 1906—neither known for calm.
Joey Yap’s analysis emphasizes the “Second Act” quality of this energy. Old industries and careers may burn down rapidly, forcing fast pivots. AI won’t stay in the backend; it becomes flashy, consumer-facing, highly competitive. It’s also a “Dispute Star” year, suggesting legal battles over AI regulation and copyright will be fierce and explosive rather than slow-walked through committees.
For AI, think labor disputes and regulatory fights. Writers and actors already struck over AI. Artists are suing over training data. Data center projects face local resistance over water and power. Fire Horse energy suggests these conflicts escalate. It’s a year for walkouts, not white papers.
In Vedic astrology, Rahu (the North Node) transits Aquarius throughout 2026. Rahu represents obsession, insatiable hunger, and alien or foreign influences. Aquarius represents the future, networks, and technology. Combined, this creates what Vedic astrologers call the “Year of the Agent”—autonomous AI systems taking over complex tasks in logistics, finance, and daily life, driven by a collective hunger for the next thing that may override ethical caution.
The speed cuts both ways. Fire Horse years bring rapid change and unexpected developments—breakthrough applications, sudden company collapses, regulatory shifts that seemed impossible months earlier. Rahu’s obsessive quality suggests society won’t pump the brakes even when it probably should.
D. Mercury Retrograde in Water: Revisiting Decisions
Mercury retrogrades exclusively in Water signs this year (Pisces, Cancer, Scorpio). In traditional astrology, Mercury retrograde is associated with miscommunication, technical failures, and—most relevantly—revisiting past decisions. Water signs emphasize emotion, intuition, and the stuff that doesn’t fit into spreadsheets.
The “rational” approach to AI governance has mostly meant cost-benefit analysis, risk frameworks, and technical standards. Mercury retrograde in Water suggests a year when the emotional and relational dimensions become harder to ignore. How do people feel about algorithmic decision-making in their lives? What got lost when we optimized for efficiency? Which communities were told their concerns were irrational?
This is also classically a time for reconsidering decisions made in haste. A lot of AI was deployed fast over the past two years—into hiring, healthcare, education, criminal justice. Mercury retrograde is the auditor going back through the files and asking uncomfortable questions about what we agreed to and why.
E. Eclipses: What Comes to Light
The eclipses fall on:
February 17: Solar Eclipse at 28° Aquarius
March 3: Lunar Eclipse at 12° Virgo
August 12: Solar Eclipse at 19° Leo
August 27: Lunar Eclipse at 4° Pisces
Eclipses are traditionally associated with revelations—things hidden coming to light, sudden shifts in visibility. The February eclipses land right around the Saturn-Neptune conjunction, concentrating a lot of “veil-lifting” energy in a short window.
For AI, the theme is transparency. What’s actually in the training data? How do the algorithms make decisions? What did companies know about risks, and when? The past year has seen more algorithmic audits, more FOIA requests about government AI use, more whistleblowers willing to talk. Eclipse seasons tend to accelerate whatever was already trying to surface.
The Leo-Aquarius axis (February and August) deals with individual versus collective, leadership versus community. The Virgo-Pisces axis deals with systems and analysis versus intuition and care. Both feel relevant to the current fights over who AI is for and who gets to decide.
Again: I’m not claiming the planets cause any of this. But astrology has always been a framework for thinking about cycles and themes—a language for pattern recognition, even if the causal mechanism is nonsense. The 2026 themes are weirdly well-suited for the AI moment we’re in.
If nothing else, it beats reading your horoscope in a magazine.
VI. What This Teaches Us About AI More Broadly
Beyond the astrology itself, this exercise illustrates something I keep returning to in my writing: LLMs are poets, not calculators. Historians, not astronomers.
They’re brilliant at pattern-matching, narrative construction, and semantic synthesis. They’re terrible at tasks requiring deterministic logic or precise computation. Understanding this distinction—and designing workflows that play to AI’s strengths while compensating for its weaknesses—is the key to using these tools effectively.
The “Centaur” approach isn’t just for astrology. It’s a model for any domain where you need both calculation and interpretation: financial analysis, scientific research, and legal reasoning. Let the machines that are good at math do the math. Let the language models do language.
Whether you’re trying to understand your Saturn return or your quarterly projections, the principle is the same: verify your data first, then interpret.
Happy 2026. May your Mercury retrogrades be gentle and your hallucinations be minimal.