When ChatGPT Gets Your Business Wrong: How to Correct What AI Says About Your Lodge or Outfit

Sooner or later, an owner types the name of their own lodge or outfit into an AI assistant and gets back a description that is wrong. Maybe it names the wrong town. Maybe it lists a trip you stopped running, or describes a different kind of business altogether. The first instinct is to hunt for a button that fixes it. As far as the vendor pages we have read show, that button does not exist.

This post explains how an assistant ends up with a description of your business, why small and new sites get misread more often, what an owner can actually do about it, and how to tell over time whether anything changed. Everything about the AI products below is as of October 13, 2026, when we read the vendors' own help pages, and we rechecked the key points in late October 2026. These products change often, so before you act on any step, open the product's current help page and confirm it still says the same thing.

First, what "ChatGPT" is actually doing when it describes you

Most of the bad advice online treats "fix the AI description" as one job. It is at least three, because an assistant can build its answer in different ways.

According to OpenAI's help article on whether ChatGPT tells the truth, the model can produce incorrect facts and fabricated citations, and a confident tone is not a sign the answer is reliable. The same article says that when search is off, the answer comes from training data up to a cutoff date. When search or deep research is on, ChatGPT can pull from web sources and cite them. It also says the model may fail to read a site because of technical issues, paywalls, or robots.txt.

So a wrong description has one of two broad origins:

• A training-data answer. The assistant is working from what it absorbed before its cutoff and is not looking at the web for that question.

• A cited-page answer. The assistant searched, found a page, and repeated or misread what was on it.

The fix is different for each, which is why this matters before you touch anything.

How a search-backed answer gets built

OpenAI's help page on searching the web with ChatGPT adds detail. It says ChatGPT may search automatically, and that the citations it shows can be incomplete, outdated, or incorrect, so the reader should open the cited page. When it uses a search partner, it rewrites your question into one or more targeted queries. It may use an approximate location based on your IP address to give local results. OpenAI also says saved memory can shape the rewritten query, and that a VPN can change the approximate location.

That explains something owners notice and cannot make sense of: the same question about the same lodge can come back with a different city, or a different set of businesses, depending on who asks and from where. It does not mean the system is random. It means the question your customer typed is not always the question that got searched.

OpenAI's October 31, 2024 announcement of ChatGPT search said it uses third-party search providers plus content from partners, and that any website or publisher can choose to appear. It does not tell owners which index sits behind it, so we will not guess at one here.

Why new and thin sites get misread

We want to be careful here. We did not find a vendor page that says "small sites are misread more often." What we can say is what follows from how the pieces above work, and we are labeling it as reasoning, not documentation.

If an assistant is answering from search and your own site says very little, there is less for it to quote. A homepage that says "welcome" and a gallery gives it nothing to ground a description on. Then the answer leans on whatever else is out there: an old directory entry, a tourism listing, a review page, a social profile written years ago. If those disagree with each other, the assistant has to pick, or blend them.

If your business is new, or changed names, or moved, the older pages tend to be the ones with the most history. The newer, correct page may simply not be what got retrieved.

None of this is a guarantee in either direction. It is the reason the fixes below focus on making the facts about you plain, consistent, and easy to find in more than one place.

What does not fix it

Before the real fixes, here are three things we would not spend a day on.

There is no correction form

Our research turned up no documented way to submit a replacement description to OpenAI. A search answer we looked at said the same: there is no direct edit of a company description. Thumbs-down feedback on an answer is feedback, not a guaranteed rewrite. If you find a page that claims otherwise, ask where in OpenAI's own help center it says so.

Blocking GPTBot is not a correction

OpenAI's crawler documentation says each robots.txt setting is independent. GPTBot is the training crawler. OAI-SearchBot is the search crawler. If you disallow GPTBot, that does not remove your site from ChatGPT search answers. And if you disallow OAI-SearchBot, OpenAI says your own pages will not be shown in ChatGPT search answers, although a navigational link can still appear. In other words, a careless robots.txt change can take away the very pages you want an assistant to read.

OpenAI also lists a third agent, ChatGPT-User, which fetches pages for some user actions. It says that agent is not used to decide whether content can appear in search, and that robots.txt may not apply because the fetch is user-initiated. Perplexity documents a similar split between PerplexityBot, which surfaces and links sites in its search results, and Perplexity-User, which fetches a page when a user asks and generally ignores robots.txt because the user requested it.

If you are thinking about whether to block any of these, that is its own decision with its own tradeoffs. For now, the point is that blocking is not a way to correct a description.

Schema is not a switch

Google's Organization structured-data guide says the markup can help Google understand details about your organization and tell it apart from others. It also says Google does not guarantee that a structured-data feature will show. Google's guide to AI features and your website says there are no extra technical requirements for AI Overviews and AI Mode beyond a page that can be indexed and shown with a snippet, and that no special schema.org type is required for them.

You will see advice online that structured data makes your site directly readable to the crawlers that feed every AI answer. The pages above do not say that, and we are not going to repeat it. Good markup is worth having for Google's understanding of who you are. Do not expect it to overwrite a sentence in a chat window.

The fixes that match how the system works

Here is the practical part. The goal is to make the correct facts about your business easy to find, consistent, and corroborated, so that whichever route an assistant takes, it lands on the right description.

Step 1: Work out which route produced the wrong answer

Ask the question with search clearly on, and look at what it cites. If it cites a page, open that page. If the wrong detail is on that page, you have found a cited-page error, and you can fix the source. If the answer cites nothing, or cites pages that are correct while the sentence is still wrong, you may be looking at a training-data answer. You cannot edit that one directly.

Do not skip opening the cited page. OpenAI itself tells readers to do this because citations can be incomplete, outdated, or incorrect.

Step 2: Fix the page it cited

If the cited page is yours, correct it and make the key facts plain in the visible text: what you are, what you offer, where you operate, and who runs it. If the cited page belongs to someone else, such as a directory or a listing, ask them to correct it. Our post on association and directory listings as entity verification signals walks through the name, address, phone, and URL consistency audit, so we will not repeat it here. The new point for this post is the order of operations: start with the specific page the assistant actually used.

Step 3: Put one correct description everywhere you control

Pick one short, accurate description of your business and use the same wording, with only light changes for length, on your website, your Google Business Profile, your social profiles, and any directory you control. When your own profiles agree, an assistant has less to reconcile.

Google gives a useful example of how this works in its own product. In Business Profile, you can write the description yourself or click Suggest description. Google says that suggestion is written from your Business Profile and other sources such as your website, and that you should review it. Google also says profile edits usually show in about 10 minutes and can take up to 30 days. That is a different product from ChatGPT with different rules, but the lesson carries over: the tool is reading what you have already published.

Step 4: Add structured data for the right reason

For a local business, Google's Organization guide recommends the most specific LocalBusiness subtype, and its AI features guide says structured data should match the visible text and that Business Profile information should be up to date. Treat markup as a clarity aid for Google, not a ChatGPT override.

Step 5: Build real third-party mentions

Your own site describing you is one voice. A trade publication, a conservation group, a regional tourism page, or a guest review that describes you correctly is another. Our post on why an AI Overview trusts third-party citations over self-published claims is clear that this is an observed pattern, not a published formula, and we will hold to the same standard here. The sensible move is to earn accurate mentions from people who genuinely know your operation and let the description follow from something true.

Step 6: Make sure the crawlers you want can read you

OpenAI's search help page says placement is not guaranteed, and that eligibility means allowing OAI-SearchBot and letting your host or CDN accept OpenAI's published search-bot IP ranges. Check your robots.txt and ask your host or CDN provider whether those ranges are being blocked. For Perplexity, read its crawler documentation for the current settings. Both vendors say a robots.txt change can take about 24 hours to show up, but confirm that on their current pages, because those numbers can change.

The three-clock retest card

Here is the one tool we would give an owner who wants to track progress. It is a simple log, and it keeps three kinds of answer on separate lines so you do not mix up what you are measuring.

Each time you test, write down:

• The date.

• The assistant and, if it shows one, the product mode.

• Whether search was on.

• The exact question you asked.

• The answer, copied as text.

• Each cited URL.

• Which clock you are waiting on.

Then sort the result into one of three lines:

• Training-data answer. No documented date exists for when a training-data description is replaced. Log it, but do not expect to move it with a page edit.

• Cited-page answer. This follows the pages the assistant reads. OpenAI and Perplexity both describe about 24 hours for a robots.txt change. Google says it can take several days to find and crawl a new page. Retest after those windows, not before.

• Google Business Profile description. Google says edits usually show in about 10 minutes and can take up to 30 days.

The reason for separate lines is that these clocks run at different speeds and apply to different things. If you retest the day after you edit a page, see no change, and conclude nothing works, you may simply have checked too early. If you edit your Business Profile and expect ChatGPT's training-data answer to change, you will wait a long time for something that was never connected.

Test from the same location, with the same account settings where you can, and keep memory and VPN differences in mind, since OpenAI says both can change the query that is actually searched. Run each question more than once, and run it on a few assistants, not just one.

Our own before and after

The plan for this post called for a dated before-and-after from our own site's AI visibility reports. We do not have an exported, dated capture ready to publish, so we are not going to describe one from memory. When we have a clean comparison with dates and the assistant and mode noted, we will add it here. In the meantime, use the retest card above to build your own record. A log you kept yourself is worth more than any story we could tell.

Before you act on any of this

Check each vendor's current help page before you rely on any detail above. That includes OpenAI's help article on searching the web with ChatGPT, OpenAI's crawler documentation, Perplexity's crawler documentation, and Google's pages on AI features and structured data. For more on how Google's own AI answers work, see our guide to Google AI Overviews for outfitters, and for the wider checklist see how to show up in ChatGPT as a Southeastern outfitter.

Work with Pine & Marsh

If you would like a second set of eyes on what the major assistants are saying about your lodge, guide service, or charter, we offer a free AI visibility check. We will tell you what we find, what is a page problem you can fix, and what is out of anyone's hands. What you've built deserves to be found. Start at pineandmarsh.com/contact.

Frequently Asked Questions

Can a lodge file a correction that replaces the sentence ChatGPT uses?

The pages we reviewed in October 2026 describe no form for submitting a new business description. Thumbs-down feedback on an answer is feedback, not a guaranteed rewrite. What an owner can do is correct the pages an answer cites and keep their own profiles accurate. Check OpenAI's current help center in case that has changed.

If search is off, where does the description come from?

According to OpenAI, without search the answer comes from training data up to a cutoff. That means a page edit will not change it directly. Log it as a training-data answer and focus your effort on the cited-page route.

Does blocking GPTBot remove a wrong description from ChatGPT search?

No. OpenAI says each robots.txt setting is independent. GPTBot is the training crawler and OAI-SearchBot is the search crawler. Disallowing OAI-SearchBot can keep your own pages out of ChatGPT search answers, though a navigational link can remain. Blocking is not a correction.

Why can the same lodge question come back with a different city?

OpenAI says ChatGPT may rewrite your question into targeted queries and may use an approximate location based on your IP address. Saved memory can shape the rewritten query, and a VPN can change the approximate location. These behaviors can change, so check OpenAI's current help page.

Does Organization schema make ChatGPT replace a wrong description?

Nothing we found says so. Google says Organization markup can help Google understand and tell your organization apart from others, with no guarantee a feature will show. Its AI features guide says no special schema type is required for AI Overviews or AI Mode.

Is Google's Suggest description the same control as ChatGPT?

No. It is a Google Business Profile tool. Google says the suggestion is written from your profile and other sources such as your website, that you should review it, and that edits usually show in about 10 minutes and can take up to 30 days.

How long before I retest?

It depends on what you changed. OpenAI and Perplexity both describe about 24 hours for a robots.txt change, Google says a new page can take several days to be found and crawled, and a Business Profile edit can take up to 30 days. No page we found gives a date for a training-data description to change. Check each vendor's current page, since these windows can change.