AI disclosure
Top Tier Newswire uses AI models across most of the site. This page says exactly where, so you can weigh what you are reading.
Where AI is used
Sentiment scores (0-100). Every headline that reaches the wire is scored by a language model running on our own hardware. 50 is neutral, above is bullish, below is bearish. The score reflects the tone and market implication of the headline and its summary — nothing else. It is not a price target, a forecast, or a recommendation.
Ticker tagging. The same model decides which listed companies a headline materially refers to, backed by a deterministic ticker-extraction pass. A headline mentioning a company in passing should not be tagged to it; when we get that wrong, the item shows up under a ticker it does not belong to.
Category labels. Earnings, Fed/CB, Macro, M&A, POTUS, Geo and the rest are model-assigned.
The "why is it moving" explainers. These articles are written end to end by a language model from the wire coverage, price action and filings data we already hold, then published without a human editing each one. They are labelled on the page as AI-assisted analysis composed from newswire data. The model is instructed to use only the data it is given; it is not browsing, and it is not sourcing anything we have not already ingested.
AI Top Trades. The long/short board in the paid product is model-generated, with its conviction scoring and evidence inputs described in the methodology.
Where AI is not used
Prices, volumes, insider filings, congressional filings, earnings dates and short interest are ingested as data from their sources and are not model-generated. When one of those numbers is wrong, it is wrong because the upstream source or our parser is wrong, not because a model invented it.
What this means for accuracy
Language models make confident mistakes. A sentiment score can misread sarcasm or an ambiguous headline. A ticker tag can attach to the wrong company where names collide. An explainer can attribute a move to the most prominent story in the coverage when the real driver was not in the news at all. We publish the method rather than a claim of accuracy, and we correct errors when they are reported: see the corrections policy.
Nothing on this site is financial advice. Every AI-derived figure here is a research signal to be checked, not a conclusion to act on.