You have found something wrong in the Wikipedia article about you. Right there on the page is an edit button. Almost everyone’s first instinct is to press it. Almost everyone is better off walking away, and the reasons why are worth twenty minutes of your time, because the people who get this wrong tend to get it spectacularly wrong and the damage lands on the article they were trying to protect.
What the rules actually say
Two separate things govern this, and people confuse them.
Conflict of interest is a guideline, meaning the community wrote it and the community enforces it. It says that editing Wikipedia to advance outside interests, including your own, your employer’s, or your clients’, is strongly discouraged. That falls short of an outright ban. What it says is that you are the wrong person to make the call. You cannot assess your own article neutrally, and the community will not assume otherwise.
Paid contribution disclosure is different and much harder. It sits in the Wikimedia Foundation’s Terms of Use, which makes it a condition of using the site rather than a community norm. Anyone compensated for edits must disclose their employer, their client and their affiliation. That covers agencies, consultants, communications staff and anyone editing as part of their job. Failing to disclose is a Terms of Use violation, not an etiquette problem.
Here is the distinction that catches people. A communications manager editing their own company’s article is doing paid editing, whatever anyone called it at the time. They draw a salary, and the edit is part of the job. Disclosure is required either way, and the requirement does not soften because the person editing thinks of themselves as an employee rather than a paid editor.
Why editing anyway goes badly
Wikipedia is unusually good at spotting this. The evidence is structural rather than circumstantial, which means nobody has to prove intent or guess at motive: the pattern is sitting in public, attached to an account, and it does not go away.
Every edit is public. It stays attached to an account or an IP address forever, and every account carries a contributions list showing everything it has ever done, which means an account whose entire history is edits to one company’s article, made from an IP address that resolves to that same company, has already announced itself before anyone goes looking. Editors work on exactly these cases. There is a public noticeboard where it gets discussed.
What follows is usually worse than the original problem. Edits get reverted. Accounts get blocked, sometimes permanently, and the block notice is public. Worse, an article can end up tagged with a banner saying its neutrality is disputed, or that a major contributor has a close connection to the subject. That banner is visible to every reader, and it is a far bigger reputational fact than whatever sentence prompted the edit. In cases involving undisclosed paid editing at scale, the result has repeatedly been press coverage naming the firm involved.
There is a second-order problem too. Editing while logged out to hide the connection is sock puppetry, which Wikipedia treats as a serious offense and is often able to detect. As usual, the cover-up costs more than the thing it hides.
What you are allowed to do
The rules discourage editing. They do not require silence, and Wikipedia has built real routes for subjects of articles.
Use the talk page. Every article has one, reached from the Talk tab. Post there, say plainly who you are and what your connection is, and describe the problem with a source. Requesting an edit this way is explicitly the recommended path, and there is a template that flags a request for an uninvolved editor to review.
Bring sources, not arguments. “This is unfair” goes nowhere. It never has. “This sentence says the company was founded in 2011, the incorporation record and this trade-press article both say 2009” gets fixed, often quickly. Wikipedia responds to verifiable evidence from published sources and is deliberately unmoved by the subject’s own assertion.
Escalate seriously harmful material properly. The biographies-of-living-persons policy is strict. Contentious material about a living person that is unsourced or poorly sourced should be removed immediately, and that applies to any editor who sees it. There is a dedicated noticeboard for these cases, and unlike ordinary content disputes it is treated urgently.
Fix plain factual errors directly, carefully. The guideline does allow uncontroversial corrections, such as an obviously wrong date. Declare the connection in your edit summary, keep it to the fact, and do not touch anything involving judgment or tone.
What a good talk page request looks like
Most requests from article subjects fail for one reason. They argue about fairness. Wikipedia does not run on fairness, it runs on sources, and a request built on the wrong currency buys nothing however well it is written.
A request that works has four parts. Say who you are and what your connection is, in the first line, before anything else. Quote the sentence you are disputing, exactly as the article has it. Give the source that contradicts it, with a link. Propose the specific replacement wording.
Something close to this:
I am the subject of this article and work at the company mentioned, so I am not editing it myself. The Early career section says I founded the company in 2011. The company’s incorporation filing and this trade-press profile from 2010 both give 2009. Suggested change: “founded the company in 2009”. Sources linked below.
Four sentences. One fact, two sources, one proposed edit, and an uninvolved editor can act on it in two minutes without taking anything on trust.
Now the version that fails. It is a long post explaining that the article is unfair, that the incident described was blown out of proportion, that competitors are probably behind it, and that the whole section ought to come out. Every word of it may be true. Nobody can act on a syllable, because there is nothing in it anyone can check.
Keep to one issue at a time. Raise six problems in a post and the discussion will settle on the weakest of them, resolving none. Take your clearest factual error, get it fixed, then come back for the next one.
Nothing happening is also a possibility, and there are routes from there. An edit-request template flags your post for review. Beyond that there are topic-specific noticeboards, and for seriously harmful material about a living person, a board that moves quickly.
The one thing to avoid entirely
Do not hire someone who promises to edit the article for you without disclosure. That market exists. It is a bad trade. If they disclose, they are subject to the same scrutiny you would be, so you gained nothing but a bill. If they do not disclose, they are violating the Terms of Use on your behalf, and when it surfaces the article gets tagged, the story becomes the undisclosed editing rather than whatever you wanted changed, and your name is attached to it.
This is why we do one thing and not the other. We have never edited a Wikipedia article for a customer and we will not. Reading a public page is not editing it, so none of the exposure above applies. Knowing early is worth more than intervening anyway, and it leaves you free to use the talk page properly instead of reacting to something that has been live for a month.
Where monitoring fits
The routes above all depend on finding out. Timing decides most of it. A talk page request lands best within hours of an edit, while the editor who made it is still around, still watching the page, and still willing to discuss a change they made this morning rather than defend one they made last spring. Six weeks later, the same request is archaeology.
We watch the page and email you within 60 seconds of a change, with the diff attached, so the decision about how to respond is yours to make while it is still early. See how it works, or read why we watch and never edit.