When ChatGPT gets a plastic surgery practice wrong, the fix is to correct the pages the assistant is reading, not the assistant. A wrong credential, address, or retired procedure almost always traces to an outdated listing or a conflicting page, and a practice can repair those in 1 to 2 weeks, then measure over 4 to 8.
AEO, GEO, AI visibility, and AI SEO are four names for the same work: getting a practice named, and named accurately, in ChatGPT, Perplexity, Gemini, and Google AI Overviews. This post uses AEO and covers corrections only. Where the engines source their picks is in [where each AI engine gets surgeon recommendations](/blog/chatgpt-perplexity-gemini-ai-overviews-plastic-surgeon-recommendations).
- A wrong ChatGPT answer about a surgeon is usually a wrong or conflicting source upstream, such as an old directory listing or a retired procedure page. Fixing the source is the lever; asking the assistant to change is not.
- A correction process has 5 steps: record the wrong answer, trace it to a source, fix the source, align every listing to one fact sheet, and re-run the same prompts after 4 to 8 weeks.
- Board certification claims carry the highest risk. A wrong or unverifiable certification claim should be corrected at the source and checked against the certifying board's public lookup.
- No correction is guaranteed to change an AI answer, and answers vary by run. Success is measured as a lower share of wrong answers across 100 or more runs. These are ClinicAds working practices, not guarantees.
Why does ChatGPT get a surgeon's details wrong?
ChatGPT states a wrong detail when its sources disagree, when a source is out of date, or when no source answers the question and the model fills the gap. A surgeon who moved offices, added a partner, or stopped offering a procedure leaves old facts on directories and press pages that the assistant can still read.
The error is therefore a data problem on the open web first. ClinicAds treats each wrong answer as a pointer to a specific source that needs attention.
Which errors matter most for a plastic surgery practice?
Credentials matter most, because a wrong board certification claim is a patient-safety and compliance issue, not only a marketing one. Location and phone errors come next because they send patients to the wrong office. Retired procedures and wrong surgeon names rank after those.
Priority should follow patient risk, not how annoying the error is. Fix credentials the same day the wrong claim is confirmed.
| Wrong detail | Patient risk | First place to check | Typical fix window |
|---|---|---|---|
| Board certification | High | Certifying board lookup, state medical board, directory profiles | Same day for the source |
| Office address or phone | High | Google Business Profile, directories, old press pages | 1 to 2 weeks |
| Retired procedure offered | Medium | Old service pages, third-party listings | 1 to 2 weeks |
| Wrong surgeon name or partner | Medium | About page, staff pages, press mentions | 2 to 4 weeks |
| Wrong prices or promotions | Low to medium | Old blog posts, deal and coupon sites | 2 to 4 weeks |
How do you correct a wrong ChatGPT answer step by step?
Correction is a 5-step source repair followed by a re-test. No step edits ChatGPT itself, because a practice cannot rewrite the model. Each step changes something a retrieval system reads, so the fix compounds across ChatGPT, Perplexity, Gemini, and Google AI Overviews at once.
- 1. Record the wrong answer: the exact prompt, the engine, the date, and a screenshot
- 2. Trace it: search the wrong claim in quotes and find which page or listing states it
- 3. Fix the source: update or request removal, starting with pages the practice controls
- 4. Align everything to one fact sheet: name, credentials, address, phone, procedures offered
- 5. Re-run the same prompts 5 times each after 4 to 8 weeks and compare the share of wrong answers
Can a practice report a wrong answer to OpenAI or Google?
Yes, ChatGPT and Google AI Overviews both offer feedback controls on an answer, and a practice can use them for a factual error. ClinicAds treats them as a supplement. A feedback report is not a documented route to a specific correction, and no independent source found while writing this post shows how often one changes an answer.
Defamation or serious misstatement about credentials is a different case. Those go to the practice's attorney, with the screenshot record from step 1 as evidence.
How long until the answer changes?
Engines that retrieve live pages can reflect a source fix within days to a few weeks, depending on when the page is next crawled. Answers drawn mostly from the model's training can lag much longer, and some will not change until a new model version.
Because answers vary by run, a single clean answer proves little. ClinicAds measures the share of wrong answers across 100 or more runs, and a falling share over 2 consecutive checks is the signal that the fix is working.
How do you keep it from happening again?
Keep one fact sheet and a change checklist. Any move, new partner, retired procedure, or credential update triggers the same 5 updates: the website, Google Business Profile, the 3 to 5 directories patients use, state and board profiles, and the practice's schema markup.
A quarterly prompt check of 20 procedure-plus-city prompts catches drift before a patient does. Paid media at $80 to $150 per booked consult is planned separately, and these are agency averages, not guarantees.
Why is ChatGPT saying the wrong thing about my practice?
Usually a source it reads is out of date or conflicts with another. Search the wrong claim in quotes to find the page, then correct that page and any matching listings.
Can I ask ChatGPT to remember the correct information?
Telling the assistant in a chat changes nothing for other users. Correcting the pages and listings it reads is the durable fix.
What should I do about a wrong board certification claim?
Fix the source the same day, confirm the correct status on the certifying board's public lookup, and keep a screenshot. A serious misstatement should go to the practice's attorney.
Does ClinicAds guarantee a corrected AI answer?
No. ClinicAds reports figures as agency averages, not guarantees, and AI answers vary by engine and by run.