AI Search

Where Does ChatGPT Get Its Information? What Plumbers and Other Home-Services Owners Need to Know

Andy Harris · November 16, 2026 · 10 min read

ChatGPT gets its information from four places: its training data, live web searches it runs when a question needs current facts, content from publishers OpenAI licenses, and whatever the user gives it. For a plumber, live search is the one you can influence fastest, so your site has to be findable and crawlable.

I run into this a lot. Owners ask me how to "get into ChatGPT" as if it were one switch. It isn't. It is four different pipes, and only some of them connect to your business.

This article leans on a clear breakdown by Si Quan Ong at Ahrefs, Where Does ChatGPT Get Its Information? 4 Sources, Explained. I'll cover what it says and what it means if you fix leaks and install water heaters for a living.

Where does ChatGPT get its information?

According to Ahrefs, ChatGPT draws on training data, live web search, licensed publisher content, and the user's own context, such as prompts, files, memory and connected apps. Which source it uses depends on the question. A homeowner asking how a water heater works gets training knowledge. A homeowner asking for a plumber near them is more likely to trigger a search.

Here is the short version of what each source means for you:

  1. Training data: your website can influence it, but only slowly and indirectly.
  2. Live web search: your website can influence it directly. This is the big one.
  3. Licensed publisher content: you can only influence it through coverage in those publications.
  4. The user's own context: you only show up if the user brings you in, for example by pasting a link.

For most home-services companies, source number two is where the work is.

What is ChatGPT's training data, and does it matter for a plumber?

Training data is the huge pile of information a model learns from before release. It gives ChatGPT its general knowledge. Your site can end up in it if it's public, but the effect is slow and you can't see what the model learned. It matters less for a plumber than live search.

According to OpenAI, as reported by Ahrefs, its models are built from three kinds of data: information publicly available on the internet, information accessed through partnerships with third parties, and information that users, human trainers and researchers provide or generate.

What exactly is in the mix? Nobody outside OpenAI knows today. Ahrefs says the last detailed breakdown was for GPT-3 back in 2020. Filtered Common Crawl, a huge public scrape of the web, made up about 60% of that mix. Pages linked from Reddit posts made up about 22%, two book collections about 16%, and Wikipedia about 3%. Since then, OpenAI has said only that its models were trained on "diverse datasets."

Here is my take. If your plumbing site is public and you haven't blocked OpenAI's training crawler, your pages may end up in what future models learn. That is a nice bonus. But it is slow, and you cannot plan a lead pipeline around it. Your content only shows up in the next model, months later.

How does ChatGPT find a plumber with live web search?

When a question would benefit from current information, ChatGPT writes searches and sends them to search providers. Ahrefs says a website can influence this source in two ways: by ranking for the topic in the results ChatGPT pulls from, and by letting OpenAI's search crawler access the site.

Ahrefs quotes OpenAI's wording that live search happens when a question "would benefit from current information." A homeowner with water coming through the ceiling wants a plumber who is open and nearby. That is exactly that kind of question.

So the practical bet is simple. ChatGPT searches the same open web that people search. If your company doesn't show up for "emergency plumber" and "water heater replacement" in your service area, you are not giving ChatGPT much to pull in.

The source doesn't say which search providers ChatGPT uses for local questions, and I won't pretend to know. What I can say from experience in local search is that the basics keep paying off. Clear service pages, a complete and active Google Business Profile, and consistent business details are what make you easy to find and easy to describe. If you want to see where you stand on those, our free Local Visibility Score is a quick start.

What is OAI-SearchBot, and should I let it on my site?

OAI-SearchBot is OpenAI's search crawler, separate from GPTBot, which collects training data. Ahrefs says blocking GPTBot does not remove you from ChatGPT's search results, because the two are independent settings. If you want to be found when homeowners ask ChatGPT for help, let OAI-SearchBot in.

This is the cheapest check in the whole article. Many small business sites are built by a web designer or an old agency, and nobody has looked at the robots.txt file in years. Sometimes a plugin or a security setting blocks crawlers without anyone deciding to.

Ahrefs also reports that GPTBot is the most-blocked AI crawler, based on its study of about 140 million websites. Big publishers often block it to keep leverage in licensing deals. That logic makes sense for a newspaper. It makes much less sense for a plumber who wants to be known.

Ahrefs puts the trade-off plainly. If you block training, future models learn less about you from your own site. If you want ChatGPT to know your brand, you probably want to leave it open.

Want to go deeper on the plumbing angle? This video covers how a plumbing business can get recommended by ChatGPT.

How to Get ChatGPT to Recommend Your Plumbing Business | AI SEO for Contractors

What about licensed publisher content and the user's own context?

Licensed publisher content is summaries of articles from publishers that have deals with OpenAI. It shows up alongside search results for news, reviews and similar topics. The user's own context is their prompt, files, memory or connected apps. A local plumber influences neither directly.

Ahrefs says licensed content can only be influenced through coverage in those publications. For a local plumbing company, chasing that is not a good use of your time. A local newspaper feature is nice, but it isn't a plan.

The user's own context matters in one small way. If a homeowner pastes your link into a chat, or uploads a quote you sent and asks "is this price fair?", ChatGPT reads your content directly. Clear, plain pricing explanations on your site help in that moment. A quote with vague line items doesn't.

What should a plumbing company do this week?

Focus on the live search source, because it is the only one you can change this month. Check that crawlers can reach your site, make each service page clear, and keep your Google Business Profile accurate. None of this is exotic. It is basic local search, done properly and kept up.

Here is the order I would work in:

  1. Open your robots.txt file. Go to yourwebsite.com/robots.txt. Make sure OAI-SearchBot is not blocked. Decide on GPTBot on purpose, not by accident.
  2. Give every service its own page. Drain cleaning, water heater replacement, sewer line repair, and so on. Each page should say what you do, where you do it, and what a homeowner should expect.
  3. Answer real questions in plain words. Put the questions customers ask on the phone onto the page, with short, direct answers.
  4. Fix your Google Business Profile. Correct categories, hours, service areas, photos and recent reviews. Our free Google Profile audit shows what's missing.
  5. Keep your details identical everywhere. Name, phone and address should match across your site and your listings.
  6. Check what AI says about you. Ask ChatGPT the way a homeowner would, and note whether you appear. Our AI visibility report does this for you.

None of this guarantees a mention. Nobody can promise that, and be careful of anyone who does. What it does is remove the reasons you'd be left out.

Why does this matter more for home-services companies?

Home-services customers often search in a hurry, when something is broken. They want a name and a phone number, not a lecture. If an AI assistant gives them a short list of companies, being on that list is worth a lot, and being missing costs you jobs you never knew about.

The good news is that the fixes overlap with what already works in local search. A site that is easy to crawl, a profile that is complete, and service pages that say what you do help you in Google and in AI answers alike.

The bad news is that it takes steady upkeep. A profile that was set up in 2021 and never touched looks stale next to a competitor who posts photos and answers reviews every week. If you would rather hand that work off, that is what our local search service is for.

Common questions

Can I stop ChatGPT from training on my website?

Yes. Block GPTBot in your robots.txt file. GPTBot is the crawler OpenAI uses to collect training data. Blocking it does not remove you from ChatGPT's search results, which are controlled by a different crawler, OAI-SearchBot. For a local business that wants to be found, I'd usually leave training open.

Does ChatGPT already know my plumbing company?

Maybe. Ahrefs says that if your site is public and doesn't block OpenAI's training crawler, there is a good chance your pages end up in the data future models learn from. But it is slow and indirect. Your content only shows up in the next model, months later, and you can't see what it learned.

Do we know exactly what is in ChatGPT's training data?

No. The last detailed breakdown OpenAI published was for GPT-3 in 2020. Since then it has shared less. Its latest system card says only that the models were trained on diverse datasets. Ahrefs notes that nobody outside OpenAI knows exactly what is in the training data today.

Do I need coverage in big publications to show up in ChatGPT?

Not for local searches. Licensed publisher content is real, but Ahrefs says you can only influence it through coverage in those publications. For a plumber, a crawlable site, strong service pages and a well-kept Google Business Profile are a far better use of your time and money.

Free Score · 60 Seconds

Ready to become the answer?