Jede fünfte ChatGPT-Nachricht dient der Informationssuche

Posted on Oct 27 • Originally published at blog.geosearchleads.com

OpenAI’s Signals v3.0 data indicates that ChatGPT is increasingly being used as an information-discovery interface, not only as a writing or productivity tool. In August 2026, “seeking information” accounted for 20.7% of messages from consumer ChatGPT plans, effectively one in five messages. Writing accounted for 20.6% in the same month.

That near-parity matters for search visibility. When people use an AI assistant to research facts, evaluate products or seek practical answers, the content most likely to inform a response may not receive a conventional search click. Publishers and brands therefore need to consider whether their information is clear, credible, accessible and easy for AI systems to interpret.

OpenAI released Signals v3.0 on October 26, 2026. The dataset provides monthly, privacy-preserving data about individual use of ChatGPT’s consumer plans: Free, Go, Plus and Pro. It excludes Enterprise use and Codex.

The OpenAI Signals data page covers usage from July 2024 through September 2026. Its monthly figures are based on a sample of about one million messages, classified by subject area and interaction type. OpenAI also provides downloadable data, methodology and a data dictionary explaining the classification taxonomy.

Within that taxonomy, seeking information includes requests for facts, product inquiries and similar questions. The 20.7% figure for August 2026, also summarized in Search Engine Journal’s reporting, shows that this category has become one of the largest observable uses of consumer ChatGPT.

These figures describe message share in OpenAI’s consumer dataset. They do not measure all search behavior, web traffic, clicks, conversions or the frequency with which ChatGPT cites external websites.

A user asking ChatGPT for product details, an explanation of a complex topic or help comparing options is expressing an information need that might once have started solely in a traditional search engine. The interface and downstream behavior are different: an AI response can synthesize multiple sources into one answer, and a user may continue the conversation rather than visit a list of results.

This does not mean conventional SEO has become irrelevant. Search crawling, indexing, authority, technical accessibility and helpful content remain foundational ways for websites to be discovered. But the Signals data is another indication that content strategy should account for answer-oriented journeys as well as result-page journeys.

For organisations, the practical question is not simply whether a page ranks for a keyword. It is whether the page provides reliable source material that can be understood, compared and potentially referenced when an AI system responds to a relevant question.

An earlier OpenAI and Harvard working paper, published in September 2025, found that seeking information grew from 14% to 24% of ChatGPT usage between July 2024 and June 2025. The research was based on roughly 1.1 million conversations and reported that Practical Guidance, Seeking Information and Writing together made up about 78% of conversations.

That earlier work supports the broader conclusion that information retrieval is a major ChatGPT use case. However, its figures should not be treated as a like-for-like trend line against Signals v3.0. The studies cover different populations, periods and reporting approaches, and classifications can be versioned. The latest 20.7% result should be understood specifically as a Signals v3.0 finding for consumer messages in August 2026.

The data does not establish a formula for appearing in AI answers. OpenAI Signals does not disclose which sites ChatGPT uses or how often it presents citations. Still, the nature of information-seeking queries points to practical content and technical priorities.

First, make factual claims easy to verify. Product specifications, pricing conditions, policies, authorship, publication dates and source references should be explicit on the page. Vague marketing copy is difficult for people to assess and provides less useful material for answer-focused discovery.

Second, structure important information clearly. Descriptive headings, concise explanations, tables for genuine comparisons and direct answers to common questions can help both users and machines identify the purpose of a page. A well-maintained FAQ can be valuable when it answers real customer questions, rather than when it exists only to target queries.

Third, use structured data that accurately reflects visible content. Relevant schema.org markup can clarify entities, products, organisations, authors and page attributes. It is not a guarantee of AI citations or search features, and it should never be used to mark up information users cannot see. Its value is in reducing ambiguity around material that is already useful and trustworthy.

Finally, measure AI visibility alongside established SEO measures. Organic rankings and referral traffic remain important, but teams may also need to monitor whether their brand is mentioned, cited or accurately represented across relevant AI discovery environments. A lack of direct referral traffic does not necessarily mean a brand has no role in a user’s research journey.

As AI-mediated research becomes more common, content teams should focus less on trying to reverse-engineer a single answer engine and more on maintaining information that is authoritative, current, attributable and technically accessible. Those are durable qualities whether a user arrives through a search result, an AI citation or a direct visit.

If your site contains product information, expert guidance or comparison content, create a GEO Search Leads account to begin assessing its GEO readiness and identify opportunities to make key information clearer for AI-driven discovery. This can help turn broad concerns about AI visibility into a more focused review of content credibility, entity clarity and citation-friendly page structure.

What does OpenAI Signals say about information seeking in ChatGPT?

OpenAI Signals v3.0 reported that seeking information represented 20.7% of consumer ChatGPT messages in August 2026. The category includes fact requests, product inquiries and similar information needs.

Does the 20.7% figure represent all ChatGPT use?

No. It represents a monthly sample of about one million messages from consumer ChatGPT plans. Enterprise usage and Codex are excluded, so the figure should not be generalized to all ChatGPT activity.

Does OpenAI Signals show which websites ChatGPT cites?

No. The Signals data reports usage classifications, not the websites used in responses or the rate at which ChatGPT cites external sources.

Can structured data guarantee inclusion in AI answers?

No. Accurate structured data can help clarify the meaning of visible page content, but it does not guarantee rankings, citations or inclusion in an AI-generated answer.

OpenAI’s latest Signals data shows that information seeking is now one of the largest observed uses of consumer ChatGPT. While the dataset cannot reveal which websites gain visibility from those interactions, it strengthens the case for publishing verifiable, well-structured and clearly attributed information that can serve both conventional search and AI-driven discovery.