Empty Input, Full Confidence: The Silent AI Pipeline Failure That Costs Sports Newsrooms Most

Modern sports newsrooms run on three layers: gathering raw material, decomposing that material into facts and claims, and shaping those facts into something a reader can use. When all three connect, you get clean journalism. When the middle layer fails silently, you get a report that looks complete and contains no story at all. This report concerns that middle layer. In the case examined here, Stage-1 deconstruction returned an empty set. No title, no source, no article type, no core viewpoints, no information points, no entities, no time-sensitivity assessment, no source-quality grading. The only populated field in the entire output was a domain label: football. Stage-2 behaved correctly. It did not refuse, and it did not invent. It rendered all nine analytical dimensions — tactics, club finance and transfers, results and public opinion, league positioning, rules and governance, management and dressing room, risk profile, media narrative, industry transmission — and marked every cell 'insufficient information.' Zero tactical signifiers, zero financial data, zero fixtures, zero named personnel, zero rule events. The most important distinction in the whole exercise is this: unknown is not the same as low. A report that says 'no financial risk identified' can mean two entirely different things — the accounts were examined and cleared, or the club was never named. Readers cannot tell the difference. A system that renders nulls as lows converts missing evidence into false reassurance, which is the single most expensive failure mode in analytical publishing. Sports journalism is unusually exposed to this. A single sentence, a single name, a single availability line can move millions. 'Injured' and 'fit' sit three words apart. Transfer rumours, injury news and tactical claims are the three categories where fabrication does the most damage — and all three are structurally inoperable without source provenance, named entities and match-level data. The pipeline itself offers a diagnostic clue. Domain classification succeeded while entity extraction failed. That asymmetry almost certainly means a split-input bug: the classifier consumed metadata, URL or a search snippet; the extractor consumed the article body, which was empty — paywall, bot-block, or a page that was never an article to begin with. The fix is a gate before extraction: if body length falls below a floor, stop and record an explicit empty-input state. One rule eliminates an entire failure class at a single point instead of at nine downstream dimensions. Three concrete controls follow. First, a machine-readable NULL_INPUT flag attached to any analysis built on an empty information set, so no downstream consumer mistakes absence of evidence for evidence of absence. Second, publication gating — the flag must block automated summarisation and publishing, because nobody reads technical warnings. Third, batch monitoring: an empty-input rate above a small threshold signals a systematic ingestion defect, not an isolated exception. Separately, make source quality and time sensitivity mandatory, non-nullable Stage-1 fields; without them, a tabloid rumour and a governing-body statement are indistinguishable, and evergreen analysis cannot be dated. The case is notable for one reason. It had nothing publishable, and the system said so, loudly and in nine places. That is not success — the real failure happened one layer up and should be fixed. But it is honesty. The day sports newsrooms treat 'this space is empty' as a first-class output is the day their reporting becomes trustworthy again.