AI reputation for executives and public figures: what ChatGPT, Gemini and Meta AI say about you

Before a meeting, an interview, an investment or an election, someone asks AI about a person. The questions usually look like these:

• "Who is this person and what do they do?"
• "What experience do they have in their role?"
• "How is their track record in the job perceived?"
• "Have they been involved in any controversy?"
• "Is this person trustworthy?"

The answer blends what the model learned with what it finds on the web at that moment. It can be out of date, mix up two people with the same name, or repeat a low-quality source. It is worth knowing it before it becomes the version everyone reads.

When the subject is a person, Refractia switches from a brand framing to a reputation framing:

• AI Reputation: how each engine describes the person. Sentiment from 0 to 100 (70 or more is positive; 40 to 69, mixed; under 40, negative), the adjectives that keep coming up, and whether the attributes the person wants to convey are actually associated with them.
• Performance by Issue: each reputation issue measured on its own, for example management, credibility, transparency or track record. It shows on which issues AI names the person and on which the person disappears.
• Presence vs. Opponents: how often AI names the person compared with others in the same field, whether peers, rivals or opponents.
• War Room: the Strategy Room in reputation mode, with a plan generated from the scan.

In addition, Influence Sources shows which sites the engines cite when they talk about the person, and the identity profile checks entity signals: whether a Wikidata item exists and whether Google's knowledge panel matches that person. As everywhere in the platform, chat and AI search are reported separately, and every rate comes with its sample ("k of n").

What an engine says about a person comes from what it learned and from what it finds on the web. The legitimate way to correct it is to work on those sources:

1. 01

   Find where the error comes from

   Influence Sources shows which domains the engines cite when they talk about the person. A wrong fact usually has a specific origin.

2. 02

   Publish the correct version where it can be verified

   An up-to-date biography on the official website, with roles, dates and track record in plain text, and public profiles consistent with it.

3. 03

   Request the correction at the source

   If the error is in a media outlet, use its corrections channel. On collaborative sites such as Wikidata, follow their rules on sourcing and conflicts of interest.

4. 04

   Measure again

   AI search can reflect the change sooner, because it checks the web as it answers; chat depends on what the model learned. A new scan shows what changed.

What is off the table: inventing quotes or facts, investigating or attacking other people, or recommending inauthentic amplification. The War Room works under those fixed rules and on how the person is perceived, not on facts for or against anyone.

• The method is the same for any person, regardless of sector, role or political position. Questions are phrased as perception questions ("how is their track record perceived?"), not as statements.
• An engine that didn't answer is not a negative opinion. That question stays marked as not measured.
• It is not real-time monitoring. Each scan is started by a person; in a crisis, you measure when you decide to measure.
• Engines measured: ChatGPT, Claude, Gemini, Google AI Overviews and Muse AI (Meta AI in WhatsApp, Instagram and Facebook). We do not measure Perplexity, Copilot, Grok or DeepSeek.
• Measuring doesn't change what AI says. Change comes from work on the sources.