What happens when AI gets a fact about a program wrong?

Where the wrong fact comes from

An engine does not invent a tuition figure. It quotes something. Usually that something is a third-party listing that was right two intakes ago, or your own site, which puts the fee on the bursar's page, the format on the department's and the duration in a PDF brochure, and lets the engine pick whichever it found first. A line like "expensive for the outcomes" is a third case: a judgement assembled out of a ranking and a thread, not a fact anyone published.

How we find it and trace it

The platform keeps the full answer and every source cited, for each question, engine and country, so nothing depends on somebody happening to screenshot it. A person on our team reads those answers and flags each claim that is wrong, outdated or unfavorable, then opens the cited page to see where the claim came from. The same bad fact often shows up in one market and not another, which is why the reading is done per country and per persona rather than once for the brand.

What happens next

The roadmap splits the flagged claims in two. The ones sourced from a page you own become a rewrite with an owner: one page stating cost, format, duration and outcomes together. The rest become a named source for comms to work on. We will not put a date on when an engine stops repeating an old fact, because we do not control the recrawl or the model. You watch the claim thin out in the tracking, and the monthly briefing says whether it moved.