OpenAI's GPT-6 adds visuals that could sideline brand links
OpenAI's new Intelligent UI renders answers as charts and interactive tools, raising the risk that brand data informs AI answers without ever earning a citation.
Key takeaways
- GPT-6's Intelligent UI generates charts and interactive widgets inside ChatGPT answers instead of plain cited text.
- Visual, model-rendered answers carry no durable link or footnote back to the original data source.
- Institutions whose authority rests on data and statistics face a growing attribution gap in AI answers.
- Citation-tracking alone now misses a growing share of how AI models actually use brand and institutional data.
- Distinctive, named methodologies and proprietary indices are harder for models to render without attribution than generic data.
Search Engine Journal reports that OpenAI is rolling out GPT-6 in ChatGPT with a feature called Intelligent UI: answers that now arrive with charts, interactive widgets and visual components built on the fly, rather than a block of text with a tidy row of citations underneath. The model decides, case by case, whether a question deserves a diagram, a comparison table or a simple paragraph. OpenAI frames this as a leap in usefulness. It is also, less publicised, a leap away from the link.
For a decade, search optimisation rested on a stable bargain: produce the authoritative page, win the click or at minimum the citation, capture the attribution. Generative answers already strained that bargain by compressing ten blue links into one paragraph with three footnotes. Intelligent UI strains it further. If ChatGPT can render its own chart of, say, comparative sovereign bond yields or industrial emissions data, it no longer needs to point a user toward the asset manager's research note or the NGO's dataset page that originally published the numbers. The interface becomes the destination. The source becomes an input, invisible to the reader, uncredited in any form a brand can screenshot and put in a board deck.
The citation was never the point, attention was
Marketers have spent two years fretting over citation counts in AI answers as a proxy for visibility, the same way they once fretted over search rankings. That proxy is now degrading. A citation at least offered a verifiable, countable unit: your domain appeared, or it did not. A generated visual offers no such unit. OpenAI's own synthesis of your data, repackaged as a chart ChatGPT drew itself, carries no watermark, no link, no footnote a brand's comms team can point to and say "this is ours." The intelligence is intelligent precisely because it absorbs the source material and renders something new. Erasure is a feature, not a bug.
This matters most for organisations whose value proposition is their data, not their prose. A multilateral like UNDRR publishes disaster-risk statistics that journalists and policymakers cite constantly; CGAP produces financial-inclusion data that shapes how central banks think about microfinance. Their authority has always been downstream of their numbers being visibly, traceably theirs. If ChatGPT can take a World Bank dataset and render its own interactive chart in response to a user's question, the provenance trail that used to run "chart, source, methodology link, institutional credibility" gets truncated to just "chart." The institution did the work. The interface gets the credit.
Industrial groups face a related but distinct problem. A company like Holcim competes partly on technical claims (carbon intensity per tonne of cement, say) that specialists verify by checking primary documentation. An Intelligent UI that visualises those claims without linking to the sustainability report strips out the verification step that made the claim credible in the first place. Financial services firms, already operating under disclosure rules that assume a traceable paper trail, have more reason than most to be uneasy about an answer layer that summarises their research without a durable, auditable link back to it.
None of this means citations vanish entirely. OpenAI has not announced the removal of source attribution from GPT-6, and SEJ's report describes Intelligent UI as additive, a new rendering layer, not a replacement for the underlying retrieval and citation mechanics that still run in the background for many queries. But the direction of travel is clear enough. Every model generation pushes more synthesis into the interface and less into the footnote. Perplexity has run visual answer formats for a year. Google's AI Overviews already favour extracted snippets over outbound clicks, per data various SEOs have tracked since the feature's 2024 rollout. GPT-6 extends the same logic: let the model build the artefact, rather than point to it.
The practical consequence for brands is that citation-tracking tools, useful as they are, measure a shrinking share of what actually happens inside an AI answer. A financial institution or UN agency that only monitors whether its domain appears in ChatGPT's source list will miss the growing number of interactions where its data informed an answer that cites nothing at all. The content that survives this transition will be the kind built for extraction and re-rendering: structured, well-labelled, machine-legible data rather than long-form narrative prose that depends on a reader clicking through to understand the context.
The institutions best placed to retain authority in an Intelligent UI world are the ones that make their data hard to detach from their name, through distinctive methodology, named indices, proprietary benchmarks, the kind of branding that survives even when the underlying numbers get redrawn as someone else's chart. Everyone else should expect their research to keep showing up in ChatGPT's answers, just with their name quietly absent from the picture.