Abstract,close up of Mainboard Electronic computer background. (artificial intelligence, ai) Credit: Canva When someone asks ChatGPT about you, they don’t get a page of links to scroll through. They get a one-paragraph summary instead. That answer is fast becoming your first impression, and you can’t scroll past it or push it to page two. The good news is that you can influence it.
AI tools don’t invent your reputation. They repeat what they find about you online. Fix those sources, and you fix the answer. That is the work that separates the best AI reputation management firms from agencies still focused only on Google’s blue links, and it’s where Reputation Resolutions leads the industry. “Your Google and AI search results are your first impression to the world,” says Anthony Will, co-founder and CEO of Reputation Resolutions and the industry’s leading expert in AI reputation management.
The stakes are high. ChatGPT reached 900 million weekly active users in February 2026, and Google’s Gemini app passed 750 million monthly users by the end of 2025. This guide draws on Reputation Resolutions‘ work correcting what AI tools say about its clients. It covers where these tools get their information, why it matters, and what to do when the answer is wrong.
How ChatGPT, Perplexity and Gemini decide what to say about you AI tools repeat what the web says about you. There is no profile you can log in and edit. Every answer comes from two places. What the model learned in training. Each model learned from a frozen snapshot of the web.
If an old lawsuit, a harsh review or a namesake’s arrest was widely published back then, the model may repeat it long after the facts changed. What it finds when you ask. Most tools also search the web in real time. But each platform searches differently and trusts different kinds of sites, so the same question can get three different answers. ChatGPT searches when a question needs current information, rewriting it into targeted queries.
It leans on encyclopedias and news: Wikipedia made up 47.9% of its top-10 cited sources. Perplexity searches for every answer and shows numbered citations. It trusts online communities most, with Reddit at 46.7% of its top-10 cited sources. Google Gemini and AI Overviews run several related searches at once across subtopics and draw on a mix of social, video and professional sites.
In AI Overviews, Reddit (21%), YouTube (18.8%) and LinkedIn (13%) are among the top-10 cited sources. Those preferences aren’t fixed, either. Which sites a platform trusts can change quickly and without warning. In September 2025, a Semrush study of 230,000+ prompts found ChatGPT citing Reddit in about 10% of answers, down from about 60% just weeks earlier, while Forbes, Medium and PR Newswire gained ground.
The takeaway: you don’t need to chase each chatbot separately. They all read the same open web. The pages that rank well, get cited often and state clear facts about you are what AI repeats. Fix those, and you improve the answer everywhere. “There’s no back door into ChatGPT,” Will says. “Every answer traces back to something on the open web.
Find that page and you’ve found the lever.” Why the AI answer matters more than the search results In search results, a bad link is one of ten. In an AI answer, it can become the whole story. Four things make that risky. Most people take the answer and move on. Few readers click through to check what an AI summary says.
When Google showed an AI summary, people clicked a search result on just 8% of visits, versus 15% without one, according to a Pew Research Center study. They clicked a source inside the summary only 1% of the time. AI can get the facts wrong and still sound certain. These tools sometimes invent claims or mix up details. ChatGPT falsely told a journalist that radio host Mark Walters had been accused of embezzlement.
In Australia, it said a mayor had been jailed for bribery. He had been the whistleblower and was never charged. Suing rarely fixes it. Walters took OpenAI to court and lost when a Georgia judge ruled for OpenAI in May 2025. For most people and businesses, correcting the sources is faster and more reliable than a lawsuit. “Suing the AI company is the slowest route to a fix,” Will says. “The answer changes when the source changes.“ Sharing a name can hurt you. If someone with your name made the news, AI may merge the two of you into one answer.
How to run an AI reputation audit When we asked Will how his team starts with a new AI reputation management client, his answer was simple: get a baseline before you change anything. “We need to know exactly what each AI is saying today,” he says. “So we start by asking the same questions your customers are already asking.” That means searches like “Who is [name]?”, “Is [company] legit?”, “[Company] reviews”, “[Name] lawsuit” and “Best [service] in [city].” His team runs those questions on every major platform: ChatGPT, Perplexity, Gemini, Google AI Overviews and Claude. Each pulls from different sources, so the answers rarely match. They use fresh conversations each time and note the date and whether the tool searched the web. Then they save everything. “We screenshot the exact statement and record every link it cites,” Will says.
Perplexity numbers its sources, and ChatGPT and Gemini show links when they search. Those links become the map of what to fix, but Will cautions against taking them at face value. “Open every one. A page can be cited even when the answer gets it wrong.” From there, each claim gets sorted as accurate, outdated, false, or about someone else, and anything with no source behind it gets flagged. False allegations and mistaken identity come first.
The audit doesn’t end there. “We rerun the same questions every month,” Will says. “Answers change as the web changes and as the models update, so we track whether that specific claim is still showing up.” If your site allows OpenAI’s crawler, ChatGPT also adds utm_source=chatgpt.com to referral links. That lets you see in your analytics which pages ChatGPT sends people to. Five ways to shape what AI says about you You can’t edit AI directly. You can change what it reads.
Fix the source, not the chatbot. Telling ChatGPT it’s wrong only changes that one conversation. Removing or correcting the article, review, or record it cites changes the answer for everyone. Publish clear, accurate content you control. A well-structured About page, leadership bios and FAQ pages give AI tools a plain statement of the facts. Google says there are no special files or markup needed to appear in AI Overviews.
The page just has to be indexed and useful. Show up where each AI looks. Perplexity and AI Overviews lean on Reddit, YouTube and LinkedIn. ChatGPT leans on Wikipedia and major news. Earned media, active professional profiles and real community discussion all feed the answers.
Keep your facts consistent. Use the same name, title, location and affiliations on every profile. Consistency helps AI tell you apart from a namesake. Don’t block the crawlers you want. OpenAI advises sites not to block OAI-SearchBot if they want to appear in ChatGPT search. Expect weeks to months, not days.
Answers built on live search usually improve first. Answers from training data catch up as models are retrained. What to do when an AI answer is false or harmful Trace it. Find the page the AI is drawing from. In most cases, a false AI answer points back to a real article, review, court listing or forum post.
Remove or correct the source. Many publishers and platforms will update or take down content that is false, outdated or violates their policies. Court records can sometimes be sealed or expunged. This is the most lasting fix, because every AI tool reads the same web. Report claims with no source. Sometimes there is no page to fix,
Leave a Reply