Information-seeking rivals writing as top ChatGPT use

Search Engine Journal's reading of OpenAI's own usage data contains a quieter number than the headline suggests. Writing assistance, the task everyone associates with ChatGPT, now sits roughly level with information-seeking among consumer messages: both near one in five. The growth in look-up queries is real but has decelerated from earlier readings. ChatGPT, in other words, has finished its transition from novelty writing tool to something closer to a reference desk, and the pace of that shift is now settling rather than accelerating.

That plateau matters more than the raw share. A fast-growing category invites speculation about where it tops out. A category that has started to level off tells you something more useful: this is roughly the steady-state shape of consumer demand on the platform. One in five messages asking "what is," "how does," "who is," or "compare X to Y" is not a transient spike tied to a product launch or a news cycle. It is the baseline. For any organisation whose name, data, or claims might plausibly surface in that fifth of all traffic, the question is no longer whether ChatGPT does lookups. It is whether your material is positioned to answer them.

The reference-desk economy has a citation problem

Search engines built an entire commercial infrastructure around the assumption that information-seeking queries are valuable and contestable: that is what search-engine optimisation has always been. ChatGPT's information-seeking share arriving at parity with its writing-assistance share means a comparable volume of intent is now flowing through a channel with none of that infrastructure. There is no bid system, no meta-description competition, no featured-snippet scramble. There is a model deciding, largely opaquely, which sources it trusts enough to synthesise or cite when a user asks a factual question.

For financial-services firms, multilateral institutions, and large industrial groups, that opacity is the real story. A bank's standard-setting white paper, a UN agency's statistical release, an industrial group's safety data: these are exactly the kind of material that answers "what is" and "how does" queries well. But being authoritative on a .org or .com domain does not automatically translate into being the source ChatGPT reaches for. Citation in these systems correlates with structured, unambiguous, frequently-referenced content, not with institutional prestige alone. A well-regarded but poorly structured policy brief can lose a citation fight to a tertiary explainer site that happens to answer the question in a format the model parses easily.

Why deceleration is the more useful signal than growth

Commentators chasing the growth number will miss the point. A still-rising category is one brands can afford to watch for another quarter before acting. A category approaching equilibrium is one where today's citation patterns are a reasonable proxy for tomorrow's, which means the window to audit and correct how your institution appears in ChatGPT's answers is now, not after the next earnings cycle. If one-fifth of consumer queries are look-ups and that share has stopped climbing, the compositional mix of ChatGPT traffic has essentially arrived. Brands still treating generative search visibility as an emerging, speculative concern are behind a trend that has already matured.