Coca-Cola's personalised can tool blocks Christian phrases like 'Jesus is King' while cheerfully printing 'Satan is good', proof that the company censors one faith and not its opposite

The part of the complaint that stands up

Start with what the objector gets right, because it is more than the usual nothing. Coca-Cola's personalisation tool does block religious content. Its own published guidelines say the service filters names of religious figures, political slogans and potentially offensive material, and when Snopes tested it in 2024 the word Jesus was refused. So was Pope.

If your complaint is that a soft-drink company has decided your faith is a category of risk to be managed, that complaint does not need a conspiracy attached to it and it is not answered by pointing out that other faiths are treated the same way. A person who wanted their can to say something about their religion was told no by an automated system, and being told no by an automated system is a recognisable modern insult.

There is also a legitimate observation about inconsistency, which is the framing Fox Business used. A moderation system that refuses a phrase at one stage and displays its opposite at another is functioning badly, whatever its intent, and a company that sells personalisation is selling a promise it then polices in ways it does not explain. None of that requires anybody at Coca-Cola to have an opinion about Jesus.

The filter has no favourites, and it has been tested twice

The claim needs the filter to have a preference. Independent testing says it does not.

Snopes ran the tool through the official Coca-Cola Store in 2024 and reported that religious names including Jesus, Satan and Allah were blocked, along with numerous Biblical names. Political slogans went the same way: it found Trump 2024 refused, and in the 2026 write-up reported a Harris slogan blocked alongside it.

That list is the answer. A system that refuses Jesus, Satan, Allah, the Pope, Trump and Harris is not adjudicating between them. It is declining to print any of it, which is the least interesting and most likely thing a beverage company would do with a product that carries its own logo next to whatever you typed.

And here is the detail that a single fact-check cannot show, because it takes two. The 2024 version of this claim alleged that the filter favoured the Democratic ticket. The 2026 version alleges it favours Satanism. The same test defeats both. When a rule is accused of opposite biases two years apart and refuses every term in both pairs, the recurring element is not the rule.

The video stops one step before the answer

The 2026 footage may well be authentic, and it still does not show what it is taken to show, because it ends at the preview screen.

Coca-Cola told Snopes by email that a technical issue affected the personalisation preview tool, so that variations of restricted terms were visible in preview mode, and that those phrases would not have been approved for production through its review and moderation process. It said it had disabled the preview feature and paused new online personalised orders while it took corrective action.

Its site describes the step the video never reaches. After an order is placed it enters an In Review Pending status and a trained member of staff reviews it manually; anything carrying religious, political, trademarked, inappropriate or otherwise restricted language is cancelled and refunded. Moderation happens after checkout, not at the keyboard.

Which leaves the one piece of evidence that would settle this entirely, and which nobody in either wave has produced: a can. Not a screenshot of a preview, not an error message, an actual printed object with one of the disputed phrases on it. Snopes could not even attempt the test, because by the time it tried, the tool was already switched off and the store was listing personalised products as out of stock.

Why a screen recording beats a fact-check

The format is doing most of the work. A person filming their own screen reads as testimony rather than as argument, and it carries an implicit claim of completeness: you saw what I saw. What you actually saw was one stage of a two-stage process, but nothing on the screen tells you that a second stage exists.

It also begins with something true. The refusal was genuine. A claim that opens with a verified fact and then asks you to accept an inference about motive is far more durable than one that opens with a fabrication, because the correction has to work against something the audience has already confirmed for itself.

Then the pairing supplies the outrage. A company that blocks all religious phrases is a dull fact about content moderation and nobody shares it. A company that blocks yours and permits its opposite is a statement about you, and that version travels. In 2024 the pairing was partisan and in 2026 it was theological, which suggests the pairing is chosen for the audience rather than found in the tool.

The Bud Light comparison was made explicitly in one of the 2024 posts, and it matters. A ready-made template exists for a brand that has supposedly picked a side, complete with a script for what happens next, and a new instance does not have to establish its premise. It only has to supply a logo.