How ChatGPT Finds Information About Your Financial Firm

Suppose an investment advisor asks ChatGPT what a fund charges, how its managers choose investments, and what its largest holdings are.

A useful answer needs to bring together several pieces of information. The fees must apply to the right share class, the investment process needs a clear explanation, and the holdings should reflect the latest information published by the fund provider. An older profile might list a fee that has since changed, while a third-party publisher could report portfolio holdings from several months ago.

For the asset manager behind that fund, this raises a practical question: where will ChatGPT get the information?

The short answer: ChatGPT generates responses using patterns learned during training and information available in the conversation. It also searches the web, either automatically or at the user’s request; some queries don’t require a web search. Connected apps can also provide additional sources of information, depending on the connections and permissions in place. The information ChatGPT uses depends on the request and available features.

Key takeaways

• ChatGPT uses several information sources. These include patterns learned during training, conversation context, and information retrieved through web search or connected apps.
• Your website is one potential source. Search answers may also draw on third-party publishers, making it useful to review how your firm is described elsewhere.
• Training and search crawlers have different jobs. You can block GPTBot while allowing OAI-SearchBot to access your website for ChatGPT search.
• Make your information easy to understand and verify. Publish current product details, answer customer questions, and support your firm’s strengths with evidence. Include key information in text alongside videos and charts.
• Monitor how your firm appears. Test relevant questions, review descriptions and cited sources, and identify gaps, factual errors, or outdated information.

Where does ChatGPT get its information?

The models behind ChatGPT are developed using three broad sources: publicly available internet information, material accessed through third-party partnerships, and information provided or generated by users, human trainers, and researchers.

During training, the model learns patterns by adjusting its numerical parameters. Some training processes also use synthetic data, as explained in OpenAI’s overview of model development.

Information can also come from uploaded files and connected apps, subject to the access and permissions available. Memory, when enabled, can contribute relevant context. Whether that information may also be used to improve models depends on the service and data settings.

Consider an investment advisor who uploads a fund’s fact sheet or investment deck that has an explanation of its investment process and its recent holdings. The advisor has supplied material ChatGPT can use to answer questions about the fund. Someone asking the same questions without those documents uploaded has given it a different starting point. In the latter case, ChatGPT may use search to locate relevant information.

What happens when ChatGPT searches the web?

As mentioned, ChatGPT can search the web automatically or at a user’s request. In managed workspaces, administrators can control access to search through settings and permissions.

When search is used, retrieved information helps shape the answer, which includes links to sources. The model still has to interpret the material and generate a response.

The wording entered by the user may differ from the query sent to a search provider. When working with search partners, ChatGPT typically reformulates a request into targeted queries. Initial results can lead to further searches, and relevant saved memories and location information may influence the queries.

In our hypothetical fund example, it may find the information from the fund provider’s website, or it could find it from third-party publishers. For the fund company this opens up the firm to a certain type of information risk. The third-party publisher could mislead the user with incorrect information or negative bias. That is why it is crucial for the fund provider to ensure its content is the primary source ChatGPT draws its information from when answering queries related to the firm. If it doesn’t, it loses control of the narrative.

For a financial marketer, this is why you must build content around the substance of your customers’ questions. A fund page should help someone understand what the fund costs, how it invests, and what it holds, with the relevant share class, reporting dates, and supporting documents clearly identified.

The same applies to other types of financial firms. For a bank, a business owner might ask ChatGPT which bank is the best for their transaction needs. A useful answer would depend on the monthly fee, included transactions, and additional charges. For the bank, publishing a clear comparison of its accounts would help ChatGPT understand the differences and check the details, increasing the chances it shows up in ChatGPT answers.

How does ChatGPT choose a source?

ChatGPT ranks search results using several factors intended to find relevant, reliable information. The public guidance does not disclose the full ranking formula or say that one type of website always takes priority.

The source needs to fit the question. A fintech firm’s current documents and website can help verify its services and fees, while a regulator’s website is a sensible reference for a query about a rule.

For financial firms, our recommendation is to publish specific information about the subjects they know directly. Explain how your products and services work, who they are intended for, and what the relevant conditions are. Give readers supporting material they can check.

GPTBot and OAI-SearchBot: What are they and what do they do?

OpenAI’s crawler documentation explains the roles of three relevant agents:

Your firm can allow ChatGPT’s search crawler while blocking its training crawler. These choices are managed separately through robots.txt, a file that gives crawlers instructions about accessing your website.

To make your content eligible for ChatGPT search, ask your web team to allow OAI-SearchBot and check that your hosting and security settings permit traffic from OpenAI’s published searchbot IP addresses.

Blocking OAI-SearchBot excludes your website’s content from ChatGPT search answers. However, ChatGPT may still display a link to help someone find your website, such as when they ask for your firm’s web address.

ChatGPT-User works differently. It visits pages in response to certain user requests, and robots.txt rules may not apply to those visits. For example, someone might ask:

“Open this RIA’s website [URL] and explain how it manages client portfolios, what financial planning services it offers, and how it charges clients.”

ChatGPT-User may be used for the page visit involved in answering that request.

Can ChatGPT use PDFs, images, and video?

It depends on how the content is provided and which features are available.

OpenAI’s training-data summary lists text, images, audio, and video among its training materials. For uploaded files, support for images inside documents depends on the plan and file type.

A web search works differently. Finding a page with a video does not mean ChatGPT has watched it. You also should not assume it has interpreted every chart inside a PDF.

For financial firms, we recommend publishing the key information in readable text alongside the media. For example:

• Add an accurate summary or transcript to a portfolio manager’s interview about the investment process.
• Explain what a chart shows, including its reporting period and source.
• Add a written walkthrough alongside a fintech app demo, explaining the features and steps shown in the video.

How can financial firms optimize for ChatGPT?

Your firm can improve the information available about it and make that information easier to access. The practical aim is to help ChatGPT find relevant material and describe your firm’s offerings and strengths accurately.

We recommend starting with an important product or service page. Check whether a reader can understand the offering, decide whether it fits their needs, and verify the information. Then work through the following areas.

Make product and service information clear

For an asset manager, identify the fund and relevant share classes. Explain its fees and investment process, date of the recorded holdings, and link to supporting documents.

For an RIA or wealth-management firm, explain who the firm serves, how it manages portfolios, and what the relationship includes. Make fees, account minimums, and the scope of any financial planning services clear.

For a bank or credit union, present account rates alongside the fees and conditions that apply. Explain membership or account eligibility, how long promotional offers last, and what happens afterward.

For a fintech firm, explain what the service does, who can use it, and how pricing works. Describe relevant limits, such as supported countries, account types, or transaction restrictions.

Answer the questions people actually ask

Use questions from your distribution, advisory, and customer-service teams to plan your financial content.

Answer the main question early, then explain the conditions and supporting details. Use descriptive headings so readers can find what matters to them.

Keep the reader’s decision in mind. Someone comparing funds needs different information from someone deciding whether to hire an RIA or open an account.

Back up your firm’s strengths

Make the reasons to consider your firm specific and verifiable.

If your RIA specializes in serving business owners, explain the planning services you provide and the team’s relevant experience. If your fintech platform simplifies a process, show the steps involved and explain what the user needs to do.

Publish accurate team biographies, explain relevant qualifications, and support factual claims with documents or reliable references. This gives readers evidence for the qualities you want associated with your firm.

Keep public information up to date

Assign someone responsibility for maintaining each important page. When a fee, rate, product feature, or service changes, update the other pages that mention it too.

Put dates beside information that changes over time. Clearly label historical material so readers can distinguish an older article from a current product description.

Review relevant company profiles and directory listings as well. Correct outdated information wherever you can update it. If another publisher has made a factual error about your firm, request a correction.

This work should cover the information a prospective client might encounter beyond your own website.

Check website access and track results

Ask your web team to check crawler permissions and hosting restrictions. Make important public pages easy to reach through internal links, and keep confidential information behind appropriate access controls.

When testing ChatGPT, review whether your firm appears, how it is described, and which sources support the answer. Record the question, date, and sources shown. Look for recurring factual errors and review the pages or profiles being cited. Finpact’s Answer Engine Optimization (AEO) and Generative Engine Optimization services can help with that. Our proprietary Answer Share provides a detailed breakdown of how different AI tools source information and talk about your firm.

Keep SEO foundations in place

Search engine optimization (SEO), answer engine optimization (AEO), and generative engine optimization (GEO) involve different search experiences. Our approach to each starts with information that readers can find, understand, and verify.

For Google’s AI Overviews and AI Mode, established SEO practices still apply. No special AI markup or schema is required. Important content should be available in text, and structured data should match the information visible on the page.

That guidance applies to Google’s features. For ChatGPT, the crawler settings explained earlier provide a practical starting point, alongside the publishing recommendations above.

Choose the pages most likely to influence a decision. Make the case for your firm clearly, support it with evidence, and assign responsibility for keeping it accurate.

Frequently asked questions