From your submission to a corrected page

What we take on

We take a claim to a publisher when the page is out of date or wrong, a document settles the question, and people act on what the page says. All three have to hold. A claim that fails any one of them stays with us, however interesting it is. The examples on this page are made up to show the kind of claim we mean; none of them is a real case.

The page is out of date or wrong

Most pages we see were accurate on the day they went up. Then the thing they describe changed. A product was retired, a rule or requirement changed, an approval was withdrawn, or a guideline or figure was replaced, and the page went on saying what it always said. Other pages were wrong from the start, usually because they report something their own source never said. Either way, what we look at is what the page tells a reader today.

A document settles it

We need something in writing from the organization that made the change, or the original document the page relies on. A company's own notice that it retired a product settles the question. So does an agency's published rule, a court's order, or the current version of a standard or form that the page quotes. A forum post doesn't settle it, and neither does a news story that paraphrases the change or our own sense that a page is probably out of date. The section on what counts as a primary source sets this out in full.

People act on it

We take on claims that change what people do. If believing the page could cost someone money, time, access, health or safety, it qualifies. A fee that no longer exists, a deadline that moved, a form that was replaced or a product that was recalled can each send a reader the wrong way. A wrong year in a list of trivia can't, however well the correction is sourced, so we leave it alone.

What we don't take on

We stay out of anything a document can't settle, and out of politics entirely.

• Opinions, reviews, predictions and matters of taste. No document can settle whether a product is good or a forecast was sensible, so there is nothing for us to check.
• Open scientific questions. When experts still dispute a question, we are in no position to call it settled. For health, food and safety we act only on formal steps by a regulator or official body, such as a ban, a withdrawn approval, a replaced guideline or a recall. We point to that step and leave the science to the people who did it.
• Political claims. These are out even when documents are attached. Taking them on would put us on one side of arguments we have no part in, and publishers of every leaning should be able to trust a message from us.
• Requests to remove accurate information. We don't help take down true things, however unwelcome they are. The point of the service is a more accurate web, and removing accurate pages works against it.

If a page is about you or your business, we can help when documents show the page is out of date or wrong, such as an old address or a service you stopped offering. Disagreements about how a page describes you, with no document to settle them, are outside what we do.

What makes a strong submission

The strongest submissions point to one sentence on one page, say what replaced it, and link to where that is written down. The form asks for each of these.

• The page address. The exact page that carries the claim, not the site's homepage. If the claim appears on several pages, send the one you saw and mention the others in your description.
• What's wrong, and what's true now. Quote the sentence or figure if you can. Then tell us what replaced it, and since when, if you know.
• A source, if you have one. A link to the company, agency or document that shows the change saves time. If you don't have one, we'll look.
• Your connection to the page. Whether you're a reader who noticed it or the page is about you or your business. It tells us where your knowledge of the claim comes from.

You don't need to be certain. If a page looks out of date but you can't find proof, send it anyway and tell us what made you doubt it. Finding the proof is our job.

What counts as a primary source

A primary source is the organization or document a fact comes from. In practice that means one of these:

• The organization that made the change. A company's own announcement, help page or terms; an agency's rule, notice or guidance; a court's order.
• The original document. The current version of the law, standard, form, study or policy that the page describes.
• Official records. Registers, filings and databases kept by the body responsible for them.

News coverage, blog posts and other secondhand accounts help us find primary sources, and they never replace one. When the original has disappeared and a reliable secondhand account is all that is left, we may rely on it, and we say so in our message. An archived copy of an original, such as one saved by the Internet Archive, counts as the original when its origin is clear.

Two things never count as evidence here: our own memory, however confident we are, and an answer from an AI tool.

How we check a claim

Every claim is checked against its sources twice before anyone is contacted, and the second check is done separately from the first.

1. We read the page as a reader would. We note the exact wording of the claim, and the date the page says it was published or updated, if it gives one.
2. We open and read every source. Nothing is taken from memory or from a summary.
3. We rely on primary sources. Secondhand accounts count only when the original is gone, and we say so.
4. We check the dates. The date a change was announced and the date it took effect are often different, and a page describing the old rule can still be right until the new one applies.
5. We look for anything later. Rules are sometimes delayed or reversed after they are announced, so we search for anything that came after the change and undid it.
6. We read the page fairly. A loosely worded page may still be correct. We ask for a change only where the page, read fairly, tells a reader something that is no longer true.
7. We check it again. A second, separate check reopens every source before anyone is contacted.

If we can't prove the claim, we tell you and stop there. Nothing goes to the publisher.

How we contact publishers

We write to the publisher, using contact details they list publicly, with the claim, what is true now and the sources that show it.

Every publisher gets the same message, and anyone can read it on our page for publishers. It says who we are, points to the specific sentence on the page, states what is true now and links the primary sources. It asks the publisher to update the page and leaves how to do it up to them. The editor approves every message before it is sent, and it goes out in our name. Yours stays out of it unless you ticked the box allowing us to share your email.

Nothing about a claim, a page or a publisher is published on this site, and there is no public list of the pages we contact. Publishers can opt out at any time, and the opt-out is permanent.

If a publisher disagrees or doesn't reply

If a publisher shows us we got it wrong, we say so. If they don't reply, we follow up, and if there is still no reply, we close the case.

A publisher who disagrees can reply with their own evidence. We check it the same way we checked the claim. If they're right, we tell them, we tell you, and the case is closed. If we still think the page is out of date, we explain why and leave the decision with them.

Some publishers fix the page without replying at all. When that happens, we tell you the page has changed. We can't make anyone change a page. The decision always belongs to the publisher, and a page that stays as it is after a clear, sourced request is an outcome we report to you like any other.

Our promises

These four promises apply to every claim we handle.

• We check before we contact anyone. Every claim is verified against primary sources first, and claims we can't prove are declined.
• Your details stay private. We never pass them to a publisher unless you tick the box on the form.
• We report back. You'll hear whether we take your claim on and what the publisher does.
• We never charge publishers. Nobody can pay us to leave a page alone. The service is free for the people who submit, too.

After you submit

You'll hear from us by email when we decide whether to take the claim on, and again when we know what the publisher did.

1. We read your submission and check the claim against primary sources.
2. We email you to say whether we're taking it on. If we aren't, we tell you why, for example because no document settles the question or because the claim falls outside what we take on.
3. If we take it on, we write to the publisher.
4. We tell you how it ended: the page was updated, the publisher disagreed and why, or there was no reply after our follow-up and we closed the case.

We don't promise a timeframe. A careful check can take a while, and publishers reply at their own pace. You can reply to any of our emails to ask where things stand.

Your details

Your details stay with us unless you tick the box that lets us share your email with the publisher.

The form sends what you type to our mailbox through Web3Forms, a form-delivery service. We use it only to check the claim, to contact the publisher and to tell you what happened. We contact the publisher in our own name. If you tick the box, the only thing we pass on is your email address, so the publisher can reach you directly if they want to. The privacy page has the full detail, including what this site measures.

Our use of software and AI

We use software, including AI tools, to find sources and draft messages, and a person checks everything before it is sent.

We use AI tools to search long documents and to draft routine letters. We don't use them to decide what is true. OpenAI's own researchers write that language models "sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty", and a confident wrong answer is exactly what this service exists to catch. So nothing an AI tool says counts as evidence here. The facts come from the sources themselves, every one of which is opened and read, and the editor approves every message before it goes out.