Can you change what ChatGPT says about your brand? Yes, but not by editing ChatGPT. The model cannot be petitioned, paid or logged into. What can be changed is the evidence it writes from: the search results, review pages, forum threads, news articles and entity records that ChatGPT retrieves and summarises. Change those and the answer changes with them. This is the method we use in ChatGPT reputation management engagements, laid out step by step so you can run the first parts yourself.
Step 1: Before you change what ChatGPT says about your brand, what does it say now?
Do not rely on one screenshot. ChatGPT answers slightly differently every time, and differently again with browsing on or off. Build a prompt set that matches how people actually ask about you:
- your brand name alone, and your name with your city or industry;
- “is it legit”, “is it safe”, “is it a scam”;
- “reviews”, “complaints”, “lawsuit”, “controversy”;
- comparisons with your two or three closest competitors;
- the awkward questions your sales and hiring teams already hear.
Run each prompt in the plain mode and with browsing enabled. Record the full answer and every citation. This is your baseline, and everything that follows is measured against it.
Step 2: How do you trace a bad sentence to its source?
Every damaging claim came from somewhere specific. With browsing on, ChatGPT usually cites the page. Without browsing, search the exact claim on Google and Bing and you will normally find the page that seeded it. In our experience ChatGPT’s retrieval leans on Bing, so check both indexes.
You end up with a mapped list: each offending sentence, the URL behind it, the platform that hosts it, and an honest read on whether it is removable, suppressible or simply outdated. Our post on what sources ChatGPT trusts explains which source types tend to sit at the top of that list.
Step 3: What can be removed outright?
False, fake, defamatory and policy-violating content goes through the channel that fits it.
- Reviews that break a platform’s own rules, such as fake engagement, conflict of interest or harassment, are reported to the platform with evidence.
- Forum and complaint-board posts are handled through the site’s grievance or moderation process, and through legal grounds where they genuinely apply.
- News articles are approached through the publisher, with a correction, an update or a request to de-index, depending on what the piece got wrong.
- Personal data is handled through the privacy request channels the platform or search engine offers.
Removal is the cleanest fix because it changes the answer for both of ChatGPT’s layers at once. Our guide to removing negative links from Google covers the search-engine side in more depth.
Step 4: What do you do with content that cannot be removed?
Outrank it. Where a page is accurate but unflattering, or simply cannot be taken down, the goal is to make sure the pages an assistant is most likely to read are the ones that tell the truth about you. That means building and promoting credible material, on your own site and on third-party platforms, that ranks above the negative source for the queries ChatGPT runs. In a Google result this pushes the bad link off page one. In a ChatGPT answer it changes which pages are retrieved at all, which is the only thing that matters when there is no page two.
Step 5: How do you give ChatGPT a version of the facts it can use?
Your website is the one source you fully control, and it is often the weakest signal you have, because the facts a model needs are buried in design. Fix that directly.
1. Publish clear factual pages that state who you are, what you do, where you operate and who runs the company, in plain sentences a machine can quote. 2. Add structured data so those facts are unambiguous, and make the site crawler-friendly, including an llms.txt file. 3. Use the same company name, founding date, address and leadership across your site, LinkedIn, registries, directories, Wikidata and any knowledge panel. Inconsistency is how a model merges you with someone else.
This is the foundation of generative engine optimization, and it does double duty: it repairs a negative answer and it makes you easier to recommend once the answer is clean.
Step 6: How do you stop the answer from drifting back?
ChatGPT’s frozen layer only updates when OpenAI ships a model, and its live layer changes whenever the web does. Re-run your prompt set on a fixed schedule and compare against the baseline. If a new thread or review appears, you catch it before it becomes a sentence. This is what AI brand monitoring does as a standing service, and it is the difference between a fix that lasts and one that quietly relapses.
What should you not do?
Do not flood the web with thin, keyword-stuffed pages about yourself. Engines treat that as noise, and it can push the genuine material further down. Do not post fake positive reviews to balance the fake negative ones; that breaks the same platform rules you are relying on for removal. Do not threaten a forum poster or a journalist. A screenshot of a legal threat becomes a new source, and it is usually a worse one than the original complaint. Every step above works on the evidence, not on the people who wrote it.
How long does this take?
It depends on which layer the damage lives in. Changes to live sources show up in browsing-enabled answers as soon as the pages themselves change and are re-crawled, usually days to weeks. Changes to the frozen layer wait for the next model update, on OpenAI’s schedule. Our post on how long online reputation management takes sets out realistic timelines for removal and suppression. For the mechanics of browsing and citations, see OpenAI’s help page on ChatGPT search.
If you want the audit and the fix run for you, call +91 8700 450 757 or contact the team and we will come back to you privately with what is driving the answer.