Fix the Source, Not the Model
When an AI answer gets a fact about you wrong, the error almost always comes from a page the engine read. AI reputation management starts by finding that page, not by arguing with the model.
Personal reputation & personal branding for the AI era · Founder, ExcBrand.AI
5 min read
In one paragraph
In AI reputation management, a wrong fact in an AI answer almost always traces back to a page the engine read. The model cannot be edited directly. Find the page that carries the error, correct it if you can, outweigh it with an accurate page you own if you cannot, then ask the same question again.
When someone first shows me an AI answer that describes them wrongly, the instinct is to argue with the engine: rephrase the question, press the feedback button, ask it to try again. I understand the instinct. I have done it about my own name. It is the wrong place to start, because the engine did not invent the sentence. It read it somewhere.
The model repeats what it read
An engine asked about a person gathers what it can find, weighs it, and writes one paragraph. If an old speaker page lists a role the person left years ago, or a directory merged two people with the same name, that error is now a source like any other. The paragraph that comes back is a summary of the pages behind it, and a wrong line in the summary usually has a wrong line somewhere behind it.
How I find the page
- Write down the wrong sentence exactly as the engine gave it, and ask the same question two or three ways to see whether it comes back.
- Read the sources the engine shows, where it shows them. The page is often there in plain sight.
- Search the wrong phrase itself, in quotation marks, with the name. A distinctive error is easy to find once you stop searching for the person and start searching for the mistake.
- Check the pages people forget: old bios on event sites, directory listings, archived team pages, profiles on platforms the person stopped using.
The page that carries the same wording, or the same mistake, is almost always the one to fix. Sometimes there are several, copied from each other, and the oldest is the one the rest repeat.
Three kinds of source, three kinds of fix
- A page you own: correct it the same day. It is the fastest fix there is, and it becomes the reference for everything else.
- A page someone else runs, such as a directory, an event site or an outlet: ask for a correction and point to the page you own as the reference. Most will fix a plain factual error.
- A page nobody will change: publish the accurate version on a page you own, make it easy for a machine to read, and link to it from the profiles you control, so the engines have a better source to weigh.
When the fix involves more than one page, the order matters. When the Sources Disagree sets it out: the page you own first, then the pages that repeat the old fact, then a fresh reading of the questions that matter.
The engine is the messenger. The page is the message.
What I do not do
I do not ask an engine to forget a fair report, and I do not publish pages whose only job is to bury one. A wrong title, a wrong company or a merged identity is an error and deserves a correction. An accurate story someone would rather not see is not an error, and this method is not for it.
The terms this piece uses
- Source page
- The page an AI engine read that carries a fact, right or wrong, which then appears in its answer.
- Trace-back
- Following a wrong sentence in an AI answer back to the page or pages that carry the same wording or the same mistake.
- Outweighing
- Publishing an accurate, machine-readable page you own when the page with the error cannot be changed, so the engines have a better source to weigh.
What to take from this
- A wrong fact in an AI answer usually comes from a page the engine read.
- Arguing with the model is the wrong place to start.
- Search for the mistake, not only for the name.
- Who controls the page decides the fix: edit it, ask for a correction, or outweigh it.
- Correcting an error is not the same as hiding a fair report.
Questions people ask
- Can I ask ChatGPT or another AI engine to correct a fact about me?
- You can leave feedback, but it does not change the pages the next answer is written from. The lasting fix is to find the page that carries the error and correct it, or to publish an accurate page you own.
- How do I find where an AI answer got a wrong fact?
- Read the sources the engine shows, then search the wrong phrase itself in quotation marks with the name. Old bios, directory listings and archived team pages are the usual places.
- What if the site with the error will not change it?
- Publish the accurate version on a page you own, make it easy for a machine to read, and link to it from your other profiles, so the engines have a better source to weigh.
Written with the help of AI and reviewed by our team.
More in Method
When the Sources Disagree
A profile and its placements are read together. When they describe the same person differently, an engine has to choose between them, and it does not always choose the page you own.
4 min read
The Search That Pairs Your Name
The people who matter rarely type a name alone. They type it next to a company, a role or another name, and the search box finishes the thought for them. That pairing is the real unit of reputation work.
6 min read
Ninety Days to a Different Answer
A quarter is long enough to change what the engines say about a person and short enough to be held to it. This is how the ninety days are spent, and what is counted at the end.
5 min read