Can a Correction Page Change What ChatGPT Says About a Professional Firm?

When does a correction page work?

It helps when:

• The error is factual and specific, such as a wrong practice area, a partner who has left, a closed office, or a merger that did not happen.
• The page answers the question in the words people use, for example "Does [firm] handle personal injury cases?" or "Is [partner] still at [firm]?"
• Other sources agree with it. If your practice area pages, bios and directory listings say the same thing, assistants have a consistent picture.
• ChatGPT searches for the answer, which is common for questions about specific firms.

It helps less when:

• The error comes from many third-party sources, such as directories, old articles and review sites, which outnumber your one page.
• The answer comes from model memory, without a search.
• The page is hard to find, such as a PDF, a page buried in a news archive, or a page with an unclear title.

Where does the fix really belong?

Negative statements are useful when they are factual and specific. "We do not handle personal injury cases" helps assistants exclude you from questions you should not be answering, which also saves your intake team time.

How should a correction or facts page be written?

If a dedicated page makes sense, such as a "Facts about [firm]" or "Common questions about [firm]" page, write it carefully:

• Lead with correct facts, not with the error. "[Firm] is an employment law practice for employers, with offices in [city] and [city]" works better than "Contrary to reports, we do not do personal injury."
• Use question headings that match how people ask: "Does [firm] represent employees?", "Which offices does [firm] have?"
• Keep it short and specific, with links to the detailed pages.
• Date material facts, such as "Opened our [city] office in 2025" or "Closed our [city] office in March 2026".
• Link to it from the about page and footer, so it can be found.

Avoid restating damaging false claims in headings or opening lines. Assistants learn associations from text that appears together, and "not involved in [scandal]" still pairs your name with the scandal.

Does fact-check markup help?

Not in Google any more. In 2025 Google announced it is phasing out support in Search for ClaimReview, the structured data used for fact-check results. Some assistants also read only the visible text of a page. Whether a landing page with fact-check markup can correct AI answers is covered in whether a landing page with FactCheck schema can correct outdated AI answers. Put the effort into clear visible text and consistent sources.

What rules apply to professional firms?

Professional rules usually govern what firms say about themselves. For lawyers in the US, for example, the ABA's model rule on communications about a lawyer's services prohibits false or misleading statements, and states have their own versions. A correction page is a communication like any other, so keep it factual and verifiable. Accountants, financial advisers and medical practices have their own rules. If the error concerns a regulatory or disciplinary matter, involve your compliance or ethics adviser before publishing.

What else should you do?

• Correct third-party sources: legal and professional directories, rating sites, chamber listings and old press releases. Asking publishers for corrections is covered in correction requests versus takedown requests for inaccurate pages.
• Report the answer through ChatGPT's feedback options, linking to the correct page.
• Use privacy rights where an error concerns an individual lawyer or partner personally, as covered in what the right to rectification means when ChatGPT gets business facts wrong.

Who inside the firm should own the correction?

Errors about professional firms often touch several people: the marketing team that runs the website, the partner or lawyer concerned, the practice group leader and, for anything touching conduct or regulation, the general counsel or ethics partner. Agree one owner who keeps the evidence, coordinates the page changes and directory corrections, and runs the follow-up tests. Brief the intake or reception team too. They hear the error first, from callers who read it in an AI answer, and can log each instance so the firm knows whether the fix is working.

How do you test whether it worked?

Before publishing, record answers to five to ten questions about the error, several runs each, with and without search. After publishing and fixing sources, repeat every two weeks. Check whether search-based answers change and whether your page is cited. Search-based answers can change within weeks once the main sources are fixed; if they have not changed after two or three test rounds, look at which sources are still cited. Memory-based answers can take much longer.

What might an illustrative case look like?

Consider an illustrative employment law firm for employers that ChatGPT kept describing as "representing employees in discrimination claims". Clients' employees kept calling for representation. The error came from two legal directories listing the wrong practice areas and an old article about a case the firm had defended. The firm corrected both directories, and added "We represent employers only" to its homepage and practice pages. It also published a short "Questions about [firm]" page with a heading asking whether the firm represents employees. Within a month, search-based answers described the firm correctly and cited its practice pages, and misdirected calls fell.

What do people ask about correction pages?

Should the correction page be called "corrections"?

Usually not. A title such as "Facts about [firm]" or "Common questions about [firm]" matches how people ask, and avoids drawing attention to the error.

Will the correction page rank for the firm's name?

Possibly, but it does not need to. It needs to be found by assistants answering the specific question.

How long should the page stay up?

As long as the error keeps appearing in AI answers or search results, and at least a few months after it stops.