The people, scenes and conversations used as illustrations in this book are imagined, including Tess and the other named characters; they are not reported encounters. Research findings are identified in the text and source notes. The small numerical model uses invented records and audiences; it describes no actual incident. No interviews or field research are claimed.
Tess has one boot on when her phone lights up. A customer has sent a picture of her face beside a sentence she has never said. The picture is hers. The name is hers. The account that posted it is not.
The customer has written a short question: Is this you?
She sits on the edge of the stairs. She is an electrician. There are tools to load, a part to collect, a kitchen without power waiting across town. She can answer the question in four words. I never said that. Yet those words do not finish the job. The customer wants to know how the picture got there. Tess would like to know that too.
The image claims that she boasts about adding costs to a bill after customers have agreed to the work. It looks like a quotation taken from an interview. There is no interview link. Beneath it, someone has added that people ought to know who they are letting into their homes.
Tess opens her calendar. The customer has not cancelled. The customer is deciding.
This is where the problem becomes larger than a screen. An unverified sentence can enter a real decision. A picture beside a name can change who gets hired, heard or invited before anyone asks where the sentence began. No advanced technology is needed for this particular deception. A photograph, a text box and a wrong attribution are enough.
We could begin with the most convincing artificial voice or the most startling fake video. There is a place for those questions. But the danger starts earlier. Somebody has attached a claim to a person. Somebody else is using it. The important work is to find out whether that attachment is justified and, if it is wrong, to undo more than the first post.
Tess can delete nothing from the customer’s memory. She cannot reach into the town’s phones. She cannot make everyone who saw the image subscribe to her next message. Even if the original vanishes, a saved copy may still arrive at a kitchen table where someone is choosing a tradesperson.
I think we ask too little of the people who pass such things along, and too much of the person trying to put them right. We behave as though making an allegation costs one click, while answering it properly may take an unlimited portion of somebody else’s life. A fair system would make repair part of the act of publishing. If you helped carry a claim out, you should help carry the correction back.
The wrong question can waste the day
A friend tells Tess to prove the image was made with AI. Another tells her it could not be AI because the hands in the photograph look normal. They are arguing about a photograph that Tess took herself.
The false part is the sentence beside it.
That seems obvious once we slow down. In the first rush, however, the loudest question can replace the necessary one. Was the file edited? Was a voice cloned? Is the person in the image real? These are separate questions. A genuine photograph may bear a false caption. A real recording may be assigned to the wrong event. An accurate transcription may identify the wrong speaker. A fabricated image could, by coincidence, illustrate something that really happened. We need to locate the actual claim before testing it.
Tess begins with the sentence. Did she say those words? If somebody claims she did, what recording, document or witness supports that claim? Who first connected them with her? These questions will not answer themselves, but they give the search a shape.
There is a difference between saying that no source has been produced and proving that no source exists anywhere. She should not claim the second merely because she has established the first. Nor should the customer treat the missing source as Tess’s failure to supply one. The person making a specific accusation owes an account of its basis. A demand that Tess disprove every possible recording would give an unsupported statement an unfair advantage.
Notice how easily that demand can grow. Perhaps the quotation is not exact. Perhaps it expresses what she meant. Perhaps it sounds like something she would say. Each retreat changes the claim while keeping the suspicion alive. A correction needs a fixed target. Otherwise, the person answering is sent from one accusation to the next while the accuser keeps the benefit of the first.
Tess can say what she knows: the words are not hers; the photograph came from her business page; the post supplies no interview; she is asking the publisher to identify its source and remove the false attribution. She need not explain the entire history of synthetic media before the customer can judge whether a quotation has been supported.
Precision is a form of kindness here. It saves time for the accused person, the reader and the person who made an honest mistake. It gives everyone something they can check.
What belongs to you
The subtitle asks who owns your reputation. Strictly speaking, the answer cannot be you alone. Other people have memories of you, opinions about your work, and freedom to describe their experiences. They may dislike you. They may tell the truth about something you wish they would forget. A decent account of reputation must leave room for criticism, disagreement and justified warnings.
That does not give others a right to invent what you said.
Think of a customer who had a poor experience with Tess. The customer may report that experience and challenge her account of it. A complaint can be painful and still deserve to be heard. Calling every unwelcome statement a reputational attack would let the powerful turn embarrassment into a reason for silence. The question is whether the particular claim has a sound basis, not whether its subject enjoys reading it.
There is also a space between fact and taste. Someone can find Tess abrupt. Another can find her direct manner reassuring. No correction notice will resolve that difference. But whether she charged a stated amount, attended a stated event or used a stated sentence may be open to evidence. We should resist mixing these types of judgment. A dislike does not authenticate a quotation. A false quotation does not prove that every criticism of its subject is unfair.
So the aim cannot be complete ownership of how others see us. Such control would require controlling their minds and speech. What we can reasonably demand is more modest and, in daily life, more valuable: care in attaching claims to people, a chance to answer material errors, and a serious effort to repair a false account when the evidence changes.
A reputation includes more than a favorable score. It includes the chance to be judged for what you actually did. That chance matters even when the resulting judgment is unkind. If Tess has made a real mistake at work, a fair process should be able to find it. It should not need an invented quotation to do so.
This gives the book a firm boundary. Defending a person against false attribution does not mean defending every act of that person. Correcting a claim does not grant innocence on all other questions. We can be exact without becoming either a publicist or a prosecutor.
The audience has moved
Suppose the account owner replies. The image was copied from somewhere else. They cannot now find where. They take it down and tell Tess that the matter is settled.
One part is settled. The original post on that account is no longer visible. Yet the customer still has the screenshot. So does the neighbor who sent it to the customer. A local discussion page has posted a cropped version without the account name. The account owner’s quiet deletion has not supplied these people with a reason to change their view.
A statement has a route. It travels from a source to a post, from the post to a group, from a group to a private conversation. The route can branch. Along the way it can lose the name of its source and gain the appearance of agreement. Three people may repeat the same screenshot, and a fourth may describe the result as several reports.
Repair must pay attention to that route. Who received the claim? Which decisions may have depended on it? Where can a correction be joined to the surviving copies? These questions make correction a delivery problem as well as a writing problem.
They also reveal why a large audience for a correction can be misleading. If Tess writes a clear response and a thousand supportive strangers read it, that may help. It tells us nothing by itself about whether the original customer saw it. The people reached by the accusation and the people reached by the correction may overlap only partly. Counting all the new readers as repaired readers would confuse attention with repair.
We will work through a small example later, using invented numbers so that every person and every route can be followed. The point is to make a question visible, not to pretend that a model knows the reach of a real post. Before celebrating the size of a response, ask whether it returned to the people who needed it.
There is a further difficulty. Receiving a correction is not the same as opening it. Opening it is not the same as accepting it. Acceptance need not undo a lost appointment or a decision already made. These steps are related, but they do not collapse into one another. A system can do well at sending notices and still leave harm unaddressed.
That is a reason to measure the steps honestly. It is not a reason to give up.
A fair share of the work
By noon, Tess has answered five messages about the image. She has spent the morning explaining a sentence she did not write. The damaged kitchen circuit remains damaged.
Repair takes time that someone also needs for work and ordinary life. Someone has to trace the copies, answer the messages and check the replies. When we think only about whether the statement is true, we can overlook who is being asked to clear up the error.
The first useful help might be a friend checking the source chain. It might be a customer forwarding the correction to the group where the image appeared. It might be the account owner doing more than deleting the post: explaining the mistake to the same followers, attaching the explanation where possible, and telling people who copied it. These are actions within reach of the people who took part.
Some cases require specialist help. A threat, an impersonated account used for fraud, a serious employment decision or a repeated campaign may bring questions that this book cannot settle. Appropriate professional or local support may be needed. We should not expect a general reader to become an investigator, security expert and lawyer at the exact moment they are shaken by an accusation.
Most of the book concerns a more basic duty shared by all of us. Pause before converting another person’s name into a conclusion. Ask what the evidence bears. If you helped distribute an error, help correct the record you helped create. A person who notices a mistake and repairs it has done something better than protecting an appearance of infallibility.
There will be unresolved cases. There will be people who deny true statements and people whose correct denials are not believed. There will be evidence too incomplete to produce a clean verdict. We need procedures that can say what is known, what remains open and what response is fair in the meantime. Certainty cannot be our only tool. It is often unavailable when a decision must be made.
Tess finally calls the customer. She asks where the screenshot came from. She sends the original business photograph and her short correction. The customer agrees to forward it to the person who sent the image.
The morning is not restored. But the direction of the work has changed. Tess is no longer trying to argue with an undefined internet. There is a claim, a source to be asked for, a person who received it and a route by which an answer can return.
Her other boot is still beside the stairs. She puts it on.