The ideas are mine. The grammar had help.

I have written funding proposals worth tens of millions of dollars at two in the morning, and the hardest part was never the argument. It was the articles. A, an, the. The small machinery of English that native speakers install before they are five and never think about again.

English is my second language. I work in an international system that runs almost entirely in it, and I have spent nine years watching what that does to people like me.

Here is what it does. In a meeting, two people can offer the same idea. The one who says it in fluent, idiomatic English is described afterwards as strategic. The one who says it in careful, slightly formal, slightly accented English is described as hardworking. I have been both people in the same week. The idea did not change. The packaging did.

So when the current argument about artificial intelligence and writing reached me, I recognised it immediately, because the assumption underneath it is one I have been living with my whole career: that the words on the page and the thinking behind them are the same thing.

They are not. They never were.

What I actually do

Let me be specific, because vagueness is how this conversation usually stays unresolved.

When I write something now, the argument is mine. The evidence is mine, usually from data I gathered or programmes I ran. The judgement about what matters, what is dangerous to say, what a donor will ignore and what a ministry will react to, is mine, and it comes from being in those rooms rather than reading about them. What I hand to a machine is language. Tighten this. Is this clear. This paragraph says the same thing twice. Give me three ways to open, none of which I will use as written.

Then I take it back and I go through it line by line, and anything that says something I do not believe comes out. That last step is the whole job. It is also the step nobody in this debate seems to talk about, and it is where authorship actually lives.

A tool that helps me say what I think is not a tool that thinks for me. The distance between those two sentences is the entire argument, and it is not a subtle distance.

The double standard nobody names

Ministers have speechwriters. Chief executives have communications teams. Every serious newspaper puts every column through an editor whose job is to fix exactly the things I ask a machine to fix. Academics have proofreaders. Nobody has ever said that a minister did not mean her speech because someone else smoothed the sentences.

Those supports have always existed. They were simply distributed to people who already had institutional power. A senior official in London has a team. A programme director in Kabul has a dictionary and a long night. When the same assistance becomes available at no cost to the second person, we suddenly discover a concern about authenticity.

I want to be careful here, because I am not accusing anyone of bad faith. Most people raising the concern are worried about something real. But I would ask them to notice who the standard has historically protected, and who it has quietly filtered out.

Because it has been a filter. The requirement to write flawless idiomatic English has never measured intelligence, rigour, or judgement. It measures where you were born and which language your mother spoke to you in. It has cost the international system an enormous amount of thinking from people who understood their contexts better than anyone flown in to assess them, and who were never quite fluent enough to be listened to at full volume.

Where the line is

None of that makes every use of these tools legitimate, and I would not want to be read as arguing that.

There is a real difference between a person who has something to say and needs help saying it in a second language, and a person who types a topic into a box and publishes whatever comes out under their name. The second one is not writing. It is not even outsourcing, because there is nothing to outsource. Nothing originated with them, so nothing is theirs to sign.

The test I use is simple. Did I have this thought before I opened the tool? Could I defend every sentence of this in a room full of people who know the subject, without the page in front of me? If the answer is no, it does not go out under my name, no matter how well written it is.

That is not a technological standard. It is the same standard that has always separated authorship from decoration. We are only noticing it now because the decoration got cheap.

What I would say to someone in my position

If you work in a language that is not your first, and you have been holding back a piece of writing because you were not sure the English would carry it: send it. Use whatever help gets it to the standard the room expects, and then own every line of it in public, the way I am owning these.

Your ideas were never the weak part. They were the whole point. The rest is typing, and typing is the least valuable thing any of us does.

I would rather live in a version of this field where the best idea in the room wins, regardless of whether the person holding it learned English at four or at twenty-four. We are closer to that than we have ever been. It would be a strange time to decide that the accent was the qualification.