OpenAI has shared data showing that looking things up now accounts for roughly as large a share of consumer ChatGPT messages as writing. That is a significant shift in how people are actually using the platform - and it has direct consequences for any brand trying to earn visibility in AI search.
For most of ChatGPT's existence, the dominant use case was generation: draft this email, write this code, summarise this document. Information retrieval was secondary. That balance appears to be shifting. The implication is not subtle: ChatGPT is functioning as a search engine for a growing proportion of its users, whether OpenAI positions it that way or not.
The Use Case Shift Changes Who Needs to Care
Until recently, AEO and GEO programmes could be presented as forward-looking investments - preparing for a future in which AI answers replace traditional search results. That framing no longer holds. If information-seeking in ChatGPT is already at parity with writing, then the audience asking questions in that environment is large enough to matter to most mid-market and enterprise brands right now.
This also changes the stakeholder conversation. Brands that dismissed AI search visibility as a fringe concern - on the grounds that users were mostly generating content, not researching purchases - need a revised view. Practitioners report that procurement teams, financial buyers, and consumers with high-consideration decisions are increasingly looking things up in ChatGPT, though supporting data across these segments remains limited. If your brand does not appear in those answers, a competitor's probably does.
It is also worth contextualising this against the broader AI search picture. Google AI Overviews, Perplexity, and Gemini are all reported to be handling growing volumes of information-seeking queries, though comparable usage data for those platforms has not been published in this context. If the same shift is playing out across multiple surfaces simultaneously, the ChatGPT signal may be indicative of a wider pattern rather than an isolated one.
What Information-Seeking Behaviour Actually Looks Like in ChatGPT
The queries being run in ChatGPT's information-seeking mode tend to be longer, more conversational, and more context-rich than traditional search queries. Someone might type three sentences explaining their situation before asking for a recommendation. That is fundamentally different from a four-word Google query - and it demands a different kind of content response.
AI models respond to that kind of query by drawing on sources they have been trained on or can retrieve in real time. Getting cited means your content needs to be structured in a way that the model can extract a clear, authoritative answer from. That means well-defined entities, direct answers to specific questions, and content that does not bury its point in preamble.
The conversational format also means users are more likely to follow up, refine their question, or ask for comparisons. Brands that appear in the initial answer have an advantage across the full conversation thread - not just the opening query. That compounds visibility in a way that a single organic ranking cannot.
What This Means for Content Structured for AI Extraction
If ChatGPT is handling a substantial and growing share of information-seeking queries, the mechanics of earning citations there become a practical priority, not a theoretical one. The core requirements have not changed: clear entity signals, direct answers to specific questions, structured content that a model can extract without interpretation, and credible sourcing that gives the model reason to trust your content over a competitor's.
What does change is the urgency of the content audit. Brands that have written primarily for Google's ranking signals - keyword density, link equity, structured data for rich results - may find their content poorly suited to AI extraction. Long-form content that answers one question across 2,000 words is harder for an AI model to cite than a page that answers ten related questions clearly and concisely.
Digital PR also becomes more relevant here. ChatGPT draws on content it can access and on information embedded in its training data. Mentions in credible publications, consistent brand information across sources, and genuine third-party endorsements all contribute to whether a model treats your brand as a reliable answer to a question. That is not a shortcut - it is the same trust-building that good PR has always done, applied to a new endpoint.
Attribution Is the Problem Most Brands Are Not Solving
More information-seeking in ChatGPT means more potential referral traffic - but referral traffic from AI surfaces is notoriously difficult to measure accurately. ChatGPT does drive clicks when it cites sources, but the referral string is not always clean, and GA4's default configuration misses a meaningful proportion of it. If your analytics setup has not been configured to capture AI-referred sessions specifically, you are almost certainly underreporting the channel.
The measurement problem matters because it affects investment decisions. If a brand cannot see ChatGPT-referred visits in their data, they have no basis for attributing leads or revenue to the channel - and no internal case for continuing to invest in AEO. The practical fix involves UTM tagging where possible, custom channel groupings in GA4, and - critically - asking in your CRM intake forms how prospects found you. Self-reported attribution is imperfect, but it captures what analytics cannot.
At the CRM level, the goal is to close the loop between an AI-referred visit and a converted opportunity. That means ensuring that lead source data flows through from first touch to closed deal, and that your revenue reporting can segment by channel with enough granularity to surface AI referrals as a distinct line. Without that, AI search remains invisible in the numbers even when it is delivering real commercial results.
The Competitive Timing Argument
The OpenAI data shows information-seeking is growing - but also that the increase is smaller than earlier data suggested. That is a useful calibration. This is not a vertical-line growth story. It is a steady, structural shift in how people use AI tools, playing out across a user base that is already large.
The competitive opportunity is real precisely because many brands are still treating AEO as a future concern. In categories where two or three competitors are actively building AI visibility and the rest are not, the gap in citation share is already opening. Once a model consistently associates a competitor's brand with the answer to a relevant question, displacing that association requires sustained effort over time.
The practical question is not whether to take AI information-seeking seriously - the usage data answers that. The question is whether your content, your entity signals, your PR strategy, and your measurement infrastructure are set up to earn and track visibility in that environment. For most brands, the honest answer is that at least one of those four is missing.