On Tuesday, Slate published an analysis spotlighting twelve Politico articles in a recent three week period where notable revisions, overwhelmingly to correct errors of fact, were made without a formal correction notice.

The article, by Jeremy Singer-Vine, a 23 year old contributor, sought to answer a simple question: How often does Politico delete or change information in its articles without notifying readers?

The question came to his mind three weeks ago, when Politico, writing on the fallout from Michael Hastings’s Rolling Stone profile of General McChrystal, ran a story containing the claim that Hastings, a freelancer, would be more likely to write a damaging article than an access-dependent Pentagon beat reporter.

Rereading Politico’s article the next day, I pointed out that those sentences had disappeared, without any notice to readers addressing the change.

Politico declined to comment on why the sentences went missing, which naturally led to speculation (including in a widely noted post by NYU journalism professor Jay Rosen) as to what laid behind the deletion. Tim Grieve, Politico’s deputy managing editor, soon wrote Rosen to say that the sentences had been removed as the article had been updated simply because the article was rewritten many times as the quick-moving story evolved.

That episode, and Singer-Vine’s article inspired by it, point to a very live issue in online journalism. The internet, where words are set not in paper and ink but in mutable zeros and ones, has exploded the possibilities for journalists needing to correct mistakes, rework words, or update conclusions. But there are vastly different ideas and procedures in practice that grapple with how to carry into the digital age traditions of correction developed by the print press, and how and when to present notice or preserve a record—for accuracy, transparency and accountability’s sake—of fixed errors or other changes made after publication.

Singer-Vine’s analysis of Politico’s unacknowledged corrections relied on a clever computer program of his design, which, for the last three weeks, had been crawling Politico’s main page archiving copies of recently posted articles, comparing versions across time and tracking changes.

The script didn’t look for unmade corrections or standing errors; it wasn’t a fact checker. All it did was note whether the article had changed since publication. Singer-Vine then culled through the articles that had such changes, put aside routine evolving news updates, smoothed over grammar flubs and the like to determine where the changes had been significant, and had been made without a written corrections notice.

A handy sidebar spotlights 12 articles—what Slate identified as “The Best of Politico’s Sneaky Edits.” Among them are a series of revisions to Robert Byrd’s obituary that took four tries to correctly peg the dates he chaired the Senate’s appropriations committee, a piece where an author got a congressman’s age wrong, and so on.

The piece wraps up with this cutting kicker:

Politico seems to have rewritten the old wire-service motto. It’s no longer, “Get it first, but first get it right.” For Politico, it’s more like, “Get it first, and if you don’t get it quite right, quietly change it later.”

And here’s where things turn for the worse. What wasn’t spotlighted in Slate’s article was Slate’s own correction policy, which was disclosed only in a pop-up box to readers who let their mouse hover over a small grey plus sign, well over a dozen paragraphs into the story. An excerpt:

Unless we notice the error ourselves and the article is less than 24 hours old, we correct and acknowledge all factual errors.

Let’s rephrase that: Slate’s correction policy allows the site to make corrections without telling readers that anything was wrong, as long as they do so within 24 hours, and as long as Slate notices the error themselves.

And, yes, only one of the highlighted revisions caught by Singer-Vine’s script likely took place later than 24 hours after its original posting. Assuming Politico’s staffers noted the errors themselves, revising them without notifying readers with a printed correction would have been perfectly by the book—by Slate’s book, that is.

While Singer-Vine quoted Politico editor in chief John Harris saying that he wasn’t sure Politico needs to have a “black-and-white [corrections] policy,” Grieve said that there was in fact a simple rule. And in one of this tale’s many ironies, such post-publication corrections without notification are not, in fact, acceptable by that standard.

“When we make factual mistakes, our policy is to make corrections and note those mistakes,” says Tim Grieve. “And that’s not Slate’s policy.”

“There’s sort of an existential question here. What’s worse? To have the right policy and fail to live up to it, or to have the wrong policy and tweak someone else for failing to live up to [their] own policy,” asks Grieve.

Clint Hendler is the managing editor of Mother Jones, and a former deputy editor of CJR.