Viral images & world events
The Wiener–Chan chatbot: who is speaking when a bot uses a person’s name?
A chatbot’s persona is not its author. If a system answers in the name of a real person, the response still needs to be attributed to the system and its operator, with any factual claims checked separately.
The reported experiment
Mission Local reported that Scott Wiener’s campaign removed an AI chatbot representing opponent Connie Chan on August 7, after launching it earlier that week. The report includes responses from the campaign and criticism of the experiment. The relevant media-literacy problem is attribution: generated dialogue attributed to a persona is not, by itself, a statement by the person being represented.
Sources: Mission Local: Local reporting on the chatbot’s removal
Check a circulating chatbot screenshot
Treat the screenshot as a claim about an interaction, not a reliable transcript of the represented person’s views. There are at least three things to establish: who ran the service, what conversation produced the response, and whether its cited facts are accurate.
- Identify the original service and operator from contemporaneous reporting or a saved disclosure. Do not infer ownership from the persona’s name, portrait or first-person language.
- Preserve the full visible prompt and response, including any earlier messages supplied with the screenshot. Note missing conversation history rather than assuming a neutral starting point.
- Separate generated assertions from linked evidence. Open the underlying records independently and check whether they support the specific assertion, not merely a related topic.
- Look for the represented person’s original statement or official record before quoting a position. Attribute the bot’s wording to the bot if that is the only available source.
- When the service is no longer available, retain that limitation. A new chatbot or a recreated prompt is not a reproduction of the earlier system’s output.
Write a correction without amplifying the persona
A concise correction can identify the operator and explain that the screenshot contains generated dialogue. Link to the source record and the original statement relevant to the claim, without turning the correction into campaign messaging. If a screenshot itself has been edited, that is an additional issue; even an unedited screenshot can accurately show a bot making a false assertion. Checking image provenance cannot decide whether that assertion is true.
Questions about this case
If a bot cites a source, can I attribute its answer to the person it represents?
No. Verify what the source actually says and who said it. A generated response remains the system’s output even when it includes a valid link.
Sources and dates
- Local reporting on the chatbot’s removal — Mission Local, 7 August 2026.
Sources checked on 27 September 2026. Linked reporting and official statements establish the attributed facts; the review steps are isGenAI’s analysis.
Our methodology explains supported checks and result limits. Found a factual error or a newer primary source? Use our corrections process and include this article’s URL.