AI-generated sports content has a credibility problem, and it is getting worse. In June, Press Gazette documented what happened to three football sites, She Kicks, Football Blog and Sportscasting, after a new owner replaced their reporters with AI. A report on the Women's FA Cup Final was "riddled with basic errors" before it was pulled. A story quoted a player with lines that, Press Gazette found, appear nowhere in the British Vogue interview it cited. Even the reporters' headshots on the bylines were AI-generated. Press Gazette described the business model plainly: buy reputable sites, replace the reporters, and fill the pages with casino and sports betting content.
The rise of AI slop in sports media
It isn't an isolated case. Front Office Sports this summer called sports "ground zero for AI slop." Three years ago Sports Illustrated was caught running articles under authors who didn't exist. Every few months the same story repeats with a new masthead, and every time it costs sports journalism a little more of its readers' trust.
Why AI-generated sports content gets the facts wrong
Here is the uncomfortable part for anyone building with AI, us included: the technology is not the problem. A language model is very good at structure, speed and tone. It is very bad at knowing what it doesn't know. Hand it one headline and a two-sentence summary and ask for 600 words, and it will deliver 600 words. Whatever the summary didn't say, the model fills in from memory, and memory is where wrong scores and invented quotes come from.
So the question for sports media isn't whether a story was written with AI. It's whether anyone, or anything, checked every fact in it against a real source before it was published.
How AX Sports fact-checks every AI-assisted story
That is the standard we hold AX News to, and as of this week it is enforced in code, not in a style guide:
• Every AI-assisted story says so on the byline. Readers shouldn't have to guess how a story was made.
• No story from a single source. Our news desk writes only when at least two outlets have reported a development, and the length follows what those sources actually carry.
• The outlet gets credited in the story the first time we use its reporting.
• Every name, number, date, place, quote and result is fact-checked against the sources before publishing. Anything unsupported comes out. If the headline itself isn't supported, the story doesn't run. In its first test, our AI fact-checking held a headline that claimed a result none of its sources reported.
The cost is volume. We expect to publish far fewer stories than we did last month. That's the trade, and it's the right one. A sports fan can get a hundred versions of every story for free. The only reason to read ours is that it's right.
One standard for every platform AX Technologies builds
AX Sports is our own product, but it is also the proof of concept for AX Technologies, the sports technology company behind it. We build live sports platforms, scores, video, news, AI and community in one app, for brands that want to own their audience instead of renting it. Every one of those platforms runs on the same rule: a fact is something you check against a source, not something you generate. The AI sports assistant in each of them, starting with Clutch on AX Sports, answers scores, schedules and betting lines from live data feeds rather than from what the model remembers. When it can't confirm something, like who's coaching a team this week, it says so instead of guessing. And it will never point anyone under 21 toward a sportsbook.
AI is going to write a lot of sports content from here on. The sports media companies that last will be the ones that treat a fact as something you check, not something you generate.
Mark Krause is the founder and CEO of AX Technologies (axtech.ai), the sports technology company behind AX Sports (axsports.ai), the AI sports platform for live scores, odds, news and highlights.