prophiq.

Prophecy × IQ. The intelligent way to forecast what's next.

How Prophiq forecasts the future.

Prophiq has a conversation with you to understand exactly what you are forecasting, then grounds the answer in live data and the agreement of multiple AI models, calibrated against reality.


Three layers feed every forecast

The foundation underneath every answer.

Prophiq doesn't just ask an AI model a question. Three layers of intelligence feed every forecast, in a process we've refined over thousands of events.

01
Our own forecasting database

Built and refined over thousands of forecasts, with continuous event discovery, classification, and curation tuned for each domain. Every resolved forecast feeds back into our calibration system.

02
Live data from authoritative sources

Real-time feeds from official data providers, government statistical agencies, prediction market platforms, and industry-standard databases. The exact mix depends on the domain, so forecasts always rest on domain-appropriate evidence.

03
Multi-model AI reasoning

Multiple frontier AI reasoning models analyze the assembled evidence independently. Their outputs are combined via a consensus algorithm so no single model's quirks dominate the answer.

04
Honest data grounding

Every forecast shows how it is grounded. When live feed data covers an event - real runners, real tournament fields, real prices - the forecast is built on it directly. When no feed exists, live research grounds it instead. Prophiq never invents a favourite or fakes a number, and when a field is not yet set, it says so.


The four domains

What Prophiq covers.

Prophiq covers four domains. Each gets data sources specifically relevant to that field, so the underlying evidence is always domain-appropriate rather than generic.

Sport

Match outcomes, championship winners, player milestones, tournament progression.

How we ground it

Live data covering horse racing across the UK, Ireland and North America, golf across all the major professional tours, football, and more - real runners, real fields, real competitors as the outcomes.

Politics

Elections, leadership changes, policy outcomes, geopolitical events.

How we ground it

Real-time prices from leading prediction market platforms, providing crowd consensus alongside our AI analysis.

Markets

Economic indicators, earnings, central bank decisions, macroeconomic releases.

How we ground it

Official government economic data sources covering leading indicators, inflation, employment, interest rates and GDP, plus live market data feeds.

Entertainment

Award winners, release performance, cultural milestones.

How we ground it

Industry-standard entertainment databases and leading music industry data sources for cultural signals.


The forecasting engine

A four-step pipeline.

Every forecast goes through a four-step process. We don't just ask one AI model and hope for the best.

01
Understanding what you mean
You ask in plain language. If your question is clear, Prophiq forecasts straight away. If it could mean more than one thing - "who wins the US Open" could be tennis or golf, across several tours - Prophiq asks, in natural conversation, until it knows exactly which event you mean. It never guesses, and it never dead-ends. Only once it is sure does it move on.
02
Live research and structured data
Real-time AI research pulls the latest context from the moment your forecast is requested. In parallel, the relevant domain-specific data sources are queried, so a forecast about a central bank rate decision sees the actual current interest rate from authoritative macro sources, and a forecast about a match or tournament sees the latest real runners, real fields, real competitors and prices from established sports data feeds. All evidence is assembled and combined with Prophiq's own historical forecasting data before reasoning begins.
03
Multi-model consensus
Multiple frontier AI reasoning models analyze the evidence independently. Each produces its own probability estimate. When models disagree, that disagreement itself is signal worth examining.
04
Calibrated consensus
A consensus algorithm combines the models' outputs into a single forecast. Confidence labels reflect how strongly the models agree with each other, where high confidence means the models converged on similar probabilities, and lower confidence means they split.

Why this matters

What makes Prophiq different.

Three reasons Prophiq forecasts differently from asking an AI chatbot "what's the probability of X?"

01It understands before it answers

Most tools take your words literally and guess at the rest. Prophiq talks it through with you - like a sharp analyst would - so the forecast answers the question you actually meant, not a near-miss of it.

02Live data, not just training data

Foundation models know what they learned during training, which can be months or years stale. Prophiq queries live data sources at forecast time: real-time prediction market prices, the most recent economic data prints, current sports standings, latest entertainment industry signals. Your forecast reflects the world as it is now, not as it was when the model was trained.

03Multiple models check each other

Single-model forecasts are confidently wrong too often. By running the same question through multiple independent frontier AI models, then combining their answers via a consensus algorithm, we surface real signal and dampen individual model quirks. When the models agree, that's meaningful. When they disagree, that disagreement is itself information.

04Track record, not vibes

Every forecast is logged. Every resolution is recorded. We compute calibration metrics so the accuracy of Prophiq's forecasts is measurable, not just claimed. Forecasting tools that don't publish their calibration shouldn't be trusted.


The daily lock

Editorial curation, every morning.

Every day at 06:00 UTC, Prophiq selects the six most consequential forecasts across all domains, one per slot. These are the day's headline forecasts, surfaced to all users. Below the headlines, each domain gets a "lead forecast" highlighted on its dedicated page.

Users want curation, not 400 events to sift through. The daily lock is our editorial signal: of everything happening today, here's what's most worth your attention.


What we don't do

Honest expectations.

Honest expectations matter for a forecasting tool. Here's what Prophiq is not.

  • We don't predict markets in seconds.
    This isn't day trading or high-frequency. Our forecasts are for events that resolve over hours, days, or months.
  • We don't claim certainty.
    Every forecast has a confidence range. A 70% forecast means we expect that outcome 70% of the time, not 100%.
  • We don't cover every event.
    We focus on the four domains where structured data exists and where forecasts have meaning. We won't forecast lottery numbers, the weather in five days, or your love life.
  • We're not a betting service.
    We provide calibrated forecasts. What you do with them is your decision.

Got a question? Ask Prophiq.

Ask something →