How our data and articles are made
Our weekly data views display published player projections, rankings and matchup fields. The lineup view uses half-PPR projections, with a separate defense board. The waiver view joins Sleeper trending adds with the week's projections; trending adds measure recent adds, not the percentage of leagues in which a player is available. These views sort, filter and format data. They do not call an AI model or make reviewed start/sit recommendations. The displayed time comes from the data's source or publication time, rather than the time someone opens the page.
Article writing has a separate review path. The current process uses an AI-assisted writer, an automated claim checker called Jev, and a separate automated review. Jev checks whether statements follow from the supplied evidence. The independent reviewer works without the writer's private conversation or self-assessment and cannot edit the draft. Code also checks source freshness, player identity and factual consistency. These checks reduce mistakes; they are not a guarantee, and they do not mean every historical article received this review or was read by a named human.
Missing, stale or unsupported evidence can hold an article back. Our rule is: no article is better than a wrong article. A hold can leave a scheduled slot empty while the required evidence or review is unavailable.
If something looks wrong, use our contact page and include the page link, the statement or number, and any supporting source. Incorrect or unsupported material can be withheld or withdrawn; a correction needs evidence too.
1. AI-assisted features
Some FCT features use automated systems that may combine Fantasy Chalk Talk data, application logic, and third-party model providers. Depending on routing and availability, those providers may include OpenRouter, Anthropic, Google, and DeepSeek.
FCT does not currently offer an AI chat assistant. The FantasyGPT chat product was retired in August 2026; chat records created while it operated are handled as described in the Privacy Policy.
Automated data views, such as rankings and projection tables, are generated from FCT data and are not individually reviewed by a human before they are shown to you.
Automated responses can make mistakes, miss context, or return outdated information. Use them as one input into your fantasy decision-making, not as your only source.
2. FCT Site Content and Automation
Fantasy Chalk Talk uses automated systems and AI-assisted workflows in parts of its content and publishing stack. Depending on the workflow, that may include research support, structuring, drafting, rewriting, formatting, QA checks, social packaging, and publication steps.
Some FCT content flows through review queues and editing steps. Human review may occur before publication, after publication, or both, depending on the workflow and content type.
Unless we explicitly label a piece as staff-written, expert-reviewed, or otherwise manually reviewed to a specific standard, you should not assume that every line of every published item was individually reviewed by a named human before publication.
3. Data and Model Limits
- Automated systems can misread context, summarize badly, or propagate stale information.
- Model outputs may be confident and still wrong.
- Fantasy football outcomes are uncertain even when the data is correct.
- External providers, upstream data delays, and product bugs can affect output quality.
4. What We Aim To Do
We aim to use verifiable data, keep unsupported claims out of product copy, and maintain workflows that reduce fabrication and stale-source problems. Even with those controls, no automated or AI-assisted system is error-free.
5. Your Responsibility
Use FCT content and AI-assisted output as informational tools. Verify important decisions with multiple sources, especially when lineup moves, waiver priorities, trades, betting, or paid decisions are involved.
6. Questions or Corrections
If you believe something is wrong or misleading, contact us through Fantasy Chalk Talk support.