How AletheiaAI produces a verdict
Every verdict starts from the same question: does the market’s story about this company survive contact with what the company actually filed? Narratives are checked against primary sources — exchange disclosures, annual reports, SEBI and NCLT orders, earnings materials — and the evidence is cited on every claim.
The two scores
The two scores are deliberately separate. A headline can rest on a true fact and still be a distorted story — and a messy, unpopular narrative can be well supported by the filings. Each verdict also labels the item signal or noise, and states its confidence level explicitly.
What the verdict examines
What the primary documents say. Exchange filings (NSE/BSE), annual reports, SEBI orders, NCLT filings, and earnings materials outrank broker notes, which outrank news reports, which outrank social media. A claim supported only by unnamed sources scores low.
Whether the narrative makes a checkable claim. "Margins expanded 180bps in Q3" can be verified against the results filing. "Massive turnaround underway" cannot — vague claims are flagged as unverifiable rather than scored as true.
Whether anything in the primary record cuts against the story — auditor qualifications, contingent liabilities, related-party transactions, promoter pledge changes, resignations, or regulatory actions the headline omits.
What the narrative leaves out. A headline about record revenue that ignores a one-time gain, or a growth story that ignores receivables ballooning faster than sales, gets marked down for what's missing, not just what's wrong.
Known narrative-inflation patterns — cherry-picked time windows, percentage growth off a tiny base, conflating order pipeline with revenue, quoting market size instead of company traction. Detected patterns are listed on the verdict.
Whether the item changes the investment picture at all. Much of daily market news is noise — repackaged PR, routine disclosures, price commentary. Each item is explicitly labelled signal or noise so attention goes where it matters.
The pipeline
From headline to published verdict, every item follows the same four steps — so the output is consistent whether the company is loved, hated, or ignored by the market.
Market narratives are collected continuously — exchange announcements, corporate disclosures, regulatory orders, earnings materials, and financial news across sectors.
Each claim is traced back to the documents that could confirm or refute it: the company's own filings on NSE/BSE, its annual report, SEBI/NCLT records, and earnings call materials. If no primary source exists, the verdict says so.
The claim receives a truth-confidence score and a narrative-distortion score, each out of 100, plus a structured breakdown: what holds up, what's missing, what's misleading, and the alternative read.
The verdict is published with its evidence chain — the specific filings and sources behind it — so any reader can verify the reasoning independently rather than trust the output.
What AletheiaAI does not do
See the methodology in action
Read today’s reality check — the day’s market headlines scored against the filings behind them — or start from the screener.