Can I write my own Wikipedia article?

No, and attempting it is worse than doing nothing. Editing an article about your own company is a conflict of interest, it gets reverted, and the attempt is logged permanently on a public page that is itself indexed and machine readable. The prohibition is what makes a Wikipedia sentence carry weight, so removing it would remove the reason you want the page in the first place.

How does a company actually get a Wikipedia article?

By being covered independently until leaving it out looks like an omission. Editors need sources that are not yours: not a press release, not your blog, not a paid directory, but somebody unconnected writing about you in a place with an editorial process. The article is a lagging indicator of coverage rather than something you can commission, which is why buying one never works and frequently backfires.

Does Wikipedia really affect AI answers that much?

Yes, and more than most teams expect. Wikipedia and Reddit together account for more than a quarter of ChatGPT's citations, and Wikidata identifiers feed the entity layer that decides whether a model recognises your company as a distinct thing at all. A model that cannot resolve you as an entity struggles to recommend you consistently, whatever your own site says.

Wikipedia: The Entity You Are Not Allowed to Edit

Q: What can I legitimately edit?

The Wikidata item and the category pages, both of which are open and largely neglected. Wikidata holds the structured facts that search and AI systems read directly, and adding a properly sourced fact to an item is ordinary maintenance rather than self promotion. Category placement decides which set of competitors you are grouped with, and being absent from your own category list is a quiet, easily fixed loss.

¶ PLAYBOOKTHE FORBIDDEN PAGE

01The page that decides your category is one you cannot write

Wikipedia and Reddit together account for more than a quarter of what ChatGPT cites. Reddit at least lets you turn up and talk. Wikipedia does not, and the rule is not a technicality you can route around, it is the thing that makes the source worth citing in the first place. The most powerful page about your industry is one you are forbidden from writing.

Why the ban is the point

A page anyone could write about themselves would be worth nothing to a model, because every company would write the same page. The conflict of interest rules are why a Wikipedia sentence carries weight, so wishing them away is wishing away the value. The prohibition is not an obstacle to the authority. It is the source of it.

Editing your own article gets reverted, logged and remembered, and the log is public. A conflict of interest flag on your company's talk page is a permanent, indexed, machine readable record that you tried. It is difficult to imagine a worse trade for a paragraph that survives eleven minutes.

// THREE SURFACES, THREE DIFFERENT RULES~~ONLY ONE IS OFF LIMITS~~

The article

Prose about your company or category, written and policed by volunteers who owe you nothing.

access · You may not write it

The Wikidata item

A structured record of facts with a stable identifier, read directly by search and AI systems.

access · Editable, with sources

The category pages

The lists and comparison tables that place you next to your competitors.

access · Open, and mostly ignored

Most companies spend their entire effort on the column they are forbidden from touching, and ignore the two that are open.

¶ PLAYBOOKTHE OPEN DOORS

02Two surfaces nobody is fighting over

The article is one of three things, and it is the only one that is closed. The other two are open, consequential, and almost entirely neglected because everybody is staring at the first column.

The item, not the article

Wikidata holds the structured record: what kind of thing you are, when you were founded, what you are called elsewhere, which identifiers point at you. It is read directly by systems that never touch the prose, and adding a properly sourced fact to an item is ordinary maintenance rather than a conflict of interest problem. That is the same discipline as keeping one story about yourself everywhere, moved to the place machines actually read.

The category, not the company

Category and comparison pages decide which set you belong to. Being absent from the list of tools in your own category is a quieter loss than a missing article and a much easier fix, because adding a well sourced entry to a list is uncontroversial in a way that writing about yourself never is. Being in the right category beats being described well in the wrong one.

EXTRACTED — THE SENTENCE THIS ENTRY EXISTS FOR

You cannot write the page. You can make it impossible to write the page about your category without mentioning you.

¶ PLAYBOOKWHAT TO DO INSTEAD

03Earning the sentence you are not allowed to write

The only legitimate route to an article is to become difficult to leave out of one, and that is a coverage problem rather than an editing problem.

Give an editor something to cite

Wikipedia runs on independent sources. Not your press release, not your blog, not a directory listing you paid for. Somebody unconnected to you writing about you, in a place with an editorial process. If that does not exist, no amount of wanting an article produces one, and the honest reading is that the article is a lagging indicator of coverage you have not earned yet. Wikipedia is not a place you publish. It is a scoreboard for coverage somebody else already gave you.

Fix the facts you are allowed to fix

Correct the structured record, add the identifiers, make sure the category placement is right, and leave the prose alone. In practice that means a handful of unglamorous fields: the official website, the founding date, what kind of organisation this is, and the links out to whichever other databases already hold a record of you. Each one is a small, sourced, uncontroversial edit, and together they are what lets a system resolve you as one thing rather than three similarly named things. That work is permitted, it is uncontested, and it feeds the systems that decide which candidates a model considers at all before any question of description arises.

It is worth saying plainly that this is slow and mostly outside your control, which is exactly why it is worth something. The surfaces you can buy your way onto are the ones that carry the least weight. The pattern holds everywhere, which is that what you cannot control is what gets believed. My own numbers said the same thing from a different direction, which is that you do not get to pick what you are cited for either.

Tomorrow: the language you are not publishing in, and why it is the cheapest citation surface left.