AI is making sports fans lazy - and hockey writing worse
Stats aren't opinions. They're countable. Yet I keep seeing fans and even writers punt the work to a chatbot, then defend a wrong number with confidence. Example: "35% of NHL games go to overtime." A quick check of recent seasons puts it closer to 20-25%. Two minutes on the standings page would have saved the take.
If you write for a living, this isn't a small problem. It's your reputation, your process, and your craft.
The contamination loop: credible site → AI → more sites → readers
Here's the bigger issue. AI models learn from "credible" sources. When those sources publish AI-made mistakes, the models ingest the errors and repeat them everywhere. Recently, a major Canadian outlet stated that Logan Cooley won the Art Ross Trophy in 2023-24. He didn't. He tied for 167th in points.
For hours, models echoed the wrong claim because the source looked legit. The correction came later, but the loop did the damage: a falsehood, amplified by speed and authority labels.
A chat with a bot isn't a source
I tested a model on a basic hockey question: Where did Logan Cooley finish in 2023-24 scoring? It hedged, gave ranges, then produced a precise rank pulled from a table that mixed in playoffs. That's not "intelligent." That's a guess dressed up as certainty.
It took longer to wrangle the model than to confirm the answer using a primary source. The model's own disclaimer spells it out: it can be wrong. Treat it like a calculator with a speech impediment - helpful, but never authoritative.
What disciplined writers do instead
- Start with the claim. Write it down before you search.
- Go to a primary source first: NHL standings, NHL stats, or a full-season table.
- Pull the exact number and the context (season, game type, filters).
- Cite the source and the retrieval date in your draft notes.
- Use a second independent source if the number matters to your thesis.
- If you don't have the full dataset, don't publish an exact rank - use a range and explain the limitation.
A lightweight verification workflow you can run on every piece
- Two-source rule for any stat that drives your argument.
- Copy the raw table (or take a screenshot) into your notes. Keep an audit trail.
- Quote the methodology: "Regular season only, 2023-24, skaters, points, no playoffs."
- Delay publication by five minutes to re-check the top three facts and all proper nouns.
- If you use AI, use it to format tables, draft outlines, or summarize - never to assert a number you haven't verified.
How to use AI without getting lazy
- Ask for sources, not answers. Click through. Verify there.
- Have the model explain the calculation steps. Reproduce them yourself.
- Use it to generate questions you should ask before publishing. Not conclusions.
- Keep a "source-of-truth" spreadsheet for recurring stats you cite often. Update it weekly.
Practical templates you can steal
- Stat line: "[Claim]. Source: [Link], [Date]. Filters: [Season/Game Type/Category]. Screenshot saved in [Folder]."
- Rank language (when uncertain): "Cooley finished around 160-170th in points among NHL skaters in 2023-24 based on regular-season totals."
- Correction policy: If a reader flags a number, verify within 15 minutes, add a note in-line, update the audit trail.
Fans are copying you
If your copy leans on guesses, readers will too. If your work links to primary data, they'll learn to check it. The standard you set in your draft becomes the standard of the conversation around your work.
The point
Speed doesn't beat trust. AI won't save a sloppy process - it will just spread it faster. Be the writer who clicks the source, runs the math, and signs their name under a claim they can back up.
Want to sharpen the way you work with AI (without blurring facts)?
Learn workflows that keep you fast and accurate: prompts that demand sources, verification steps, and tooling that supports editorial discipline - not shortcuts. Explore practical options for your role here: AI courses by job.
Your membership also unlocks:
