What ChatGPT Says About You and How to Change It
For years the default reputation check was a Google search. Now more people are running that same check through ChatGPT, and they’re getting something very different back. Not ten links to weigh. One paragraph that sounds like it already decided. That shift is driving a lot of uncomfortable conversations lately, and it’s why ChatGPT reputation management has jumped from a niche SEO topic to a live concern for executives, founders, and anyone whose name does work.
The scale here is bigger than most people realize
ChatGPT is processing somewhere north of 2 billion queries a day, with roughly 800 million weekly active users as of late 2025. Those are OpenAI’s own figures. Researchers at Elon University ran a national survey on AI adoption and found that about 18 percent of users had already asked an AI chatbot what it says about them personally. Another 23 percent had looked up people they know. Gartner expects traditional search volume to drop around 25 percent by 2026 as queries move toward AI.
There’s a phrase going around in reputation circles that captures the shift pretty well: people aren’t just Googling themselves anymore. They’re ChatGPTing themselves.
And here’s the part that actually matters. ChatGPT doesn’t show you options. It synthesizes. When someone types “Is [your company] any good?” they don’t get a ranked list of sources to weigh for themselves. They get one confident-sounding answer that reads like a recommendation or a warning. Users treat that answer more like advice from a colleague than like a search result. Which is a meaningfully different thing than what Google does to perception.
Where ChatGPT actually gets its information
If you want to change what ChatGPT says about you, you have to understand where it’s pulling from. The model isn’t scanning the live web for every question. It’s working from two different layers.
The first is training data. OpenAI has confirmed its models are trained on publicly available internet content, licensed third-party datasets, and user-provided input, with filtering applied to strip out certain categories. That dataset is huge, and it’s compiled up to whatever the model’s knowledge cutoff happens to be.
The second layer is live retrieval. When browsing is turned on, ChatGPT can look things up in real time before answering, which is why some responses feel current and others feel weirdly outdated.
The interesting question is what sources the model leans on the most. Profound ran an analysis of 30 million AI citations between August 2024 and June 2025, and the results were revealing. Wikipedia came out as ChatGPT’s single biggest source, accounting for around 43 percent of citations. Reddit was next at roughly 11 to 12 percent. YouTube was around 5 percent. Commerce queries shifted that mix: Wikipedia dropped to 22 percent, while Amazon jumped to 19 percent and Reddit climbed higher.
The takeaway is blunt. If Wikipedia and a handful of other high-authority sources describe you one way, ChatGPT will describe you that same way. If those sources are thin, outdated, or hostile, ChatGPT inherits that framing and hands it out with total confidence.
One extra wrinkle worth knowing. Analysts at Five Blocks have documented cases where ChatGPT surfaced information from Wikipedia pages that had been deleted or were never indexed by Google in the first place. Models retain snapshots from their training runs, so they can “remember” content that no longer exists on the live web. Your reputation isn’t only shaped by what’s out there now. It can also be shaped by what was out there when the model was trained.
Running your own audit
Before you try to change anything, find out what the AIs are actually saying. A decent self-audit takes under an hour and you should repeat it at least quarterly.
Pick five or six prompts someone might realistically run. Things like:
- Who is [your full name]?
- What does [your company] do?
- Is [your company] a good [industry] firm?
- How does [your company] compare to [a known competitor]?
- Has [you or your company] been involved in any controversies?
- Best [industry] firms in 2026
Run each of those through ChatGPT. Then run them through Claude, Gemini, and Perplexity, because each one has different sourcing behavior and you want the full picture. For every response, write down three things: what it got factually wrong, how the sentiment felt (favorable, neutral, hedged, hostile), and which specific sources got cited if any did.
That document is your baseline. Every move you make from here gets measured against it.
You can’t edit ChatGPT. You can edit what it reads.
One of the more persistent myths we run into is that there’s a hidden button inside ChatGPT where you can correct what it says about you. There isn’t. Status Labs put it cleanly in their recent guidance: unlike a search result, you can’t just delete information from an AI model’s training data.
OpenAI does have a privacy-related removal process for individuals in certain jurisdictions, mostly tied to GDPR. It’s narrow, it’s slow, and it’s not really built for everyday reputation issues like an outdated title or unflattering framing.
So you don’t edit the model. You edit the web the model reads. That’s the whole game. Change the underlying sources, and over time, as training data refreshes and live retrieval pulls from updated content, ChatGPT’s answer about you shifts too.
What actually moves the needle
Fix Wikipedia first. It’s the single biggest surface you can work on, just because of how much ChatGPT pulls from it. If you or your company already has a Wikipedia page, go through it carefully. Check the facts, make sure every claim traces back to a credible third-party source, and clean up anything that’s aged out of relevance. If you don’t have a page, don’t rush out to create one. Wikipedia’s notability bar is high, and pages clearly spun up by interested parties tend to get flagged or removed by editors. The path is to earn the kind of independent coverage that eventually supports a neutral page someone else decides to write. Keep an eye on the page after that too. Small edits to phrasing can ripple into AI output months later.
Make your owned content pull its weight. ChatGPT likes structure, clarity, and consistency. Marketing fluff doesn’t help it place you. “Award-winning marketing leader” is noise. “Director of Growth at TheBestReputation, leading brand growth and content strategy” is something the model can actually use. Your site, your bios, your press room, and your leadership pages should tell the same story in the same language. If any of them drift, the AI hedges.
Third-party coverage is where most of the real lifting happens. Trade publications, established news outlets, serious industry blogs, credible thought-leadership placements, and reputable Q&A communities are the sources ChatGPT tends to weight most heavily. Reddit deserves its own callout here. It’s one of the top non-Wikipedia sources ChatGPT cites, and it’s also one of the hardest places to influence without getting caught trying. Genuine participation wins. Promotional behavior gets flattened.
Kill misinformation at the source. If ChatGPT is repeating something inaccurate about you, that claim almost always traces back to a specific page. An old article. A scraped data broker profile. A forum thread. A piece you forgot existed. Fix it at the source by correcting, replacing, or pursuing formal removal where it applies. This is classic ORM work, but it does double duty now, because every bad input you pull down is one less thing the next training run will ingest.
Consistency across platforms is probably the cheapest win available. LLMs reward it heavily. When LinkedIn, your company site, Crunchbase, your Wikipedia, and your press coverage all describe you the same way, ChatGPT converges on that description. When they contradict each other, the model hedges, softens, or falls back to generic language. Aligning executive bios and category language across every platform you control tends to show up in AI output within a couple of months. Most companies never get around to it.
When to bring someone in
Plenty of this you can do in-house if you’ve got editorial discipline and time. A lot of executives don’t. The content load, the placement work, the relationship-building, the monitoring across four or five AI platforms at once, the SEO work underneath all of it, it adds up fast.
If you’re going to hire help, the firm should have a real track record in online reputation management, not just general SEO or PR. They should run an actual content and placement operation. And they should be honest about what AI reputation work really is. Any firm promising to “remove” something from ChatGPT is selling you something that doesn’t exist. What a serious firm does is change the underlying web so the model’s output shifts over time.
TheBestReputation (TBR) is built around exactly that kind of program. Inc. 5000 recognized, with an integrated approach that ties content, authoritative third-party placements, search optimization, and active monitoring into one coordinated operation. That matters for ChatGPT reputation management because you can’t move AI output by pulling one lever. You have to move several surfaces at once and keep moving them.
Other reputable firms in the space, Status Labs, Reputation X, and Five Blocks among them, run credible programs too and have published useful public research on AI reputation. Shopping a few firms is fair. What you’re really looking for is a team that treats AI visibility as an ongoing discipline, not a one-time cleanup.
The bottom line
ChatGPT isn’t going to stop shaping how you’re perceived, and it isn’t adding a “correct this about me” button any time soon. The people whose reputations hold up in this environment are going to be the ones who understand the mechanics. ChatGPT describes you based on the strongest, most consistent, most authoritative material it can find about you on the open web. Change that material and you change the answer.
ChatGPT reputation management isn’t about gaming an algorithm. It’s about making sure the most accurate and current version of your story is the one the AI can actually reach. The work is doable. It just has to happen on purpose.
If you’d like a structured read on what ChatGPT, Claude, and Gemini are currently saying about you or your company, and a concrete plan to shift it, the team at TheBestReputation can walk you through it.
