GPT-6's Intelligent UI changes what brand visibility means

OpenAI's own description of GPT-6 is almost quaint: "faster responses with visuals and interactive experiences you can explore and use directly." Buried in that sentence is the end of an era for anyone who has spent the past two years optimising for citations in a wall of text. ChatGPT answers are no longer destined to be prose. They are becoming interfaces.

Intelligent UI, as OpenAI calls it, renders charts, product cards, comparison tables and interactive widgets inline, generated on the fly in response to a query. Ask about mortgage rates and you may get a live calculator rather than a paragraph explaining amortisation. Ask about industrial coatings and you might get a comparison grid pulling specifications from several manufacturers, rendered as a table rather than cited as a list of links. The model is still retrieving and synthesising information from the web, the mechanism that has mattered to brands since ChatGPT started citing sources. What changes is the packaging, and packaging is never neutral.

Citation is necessary but no longer sufficient

For two years, the working assumption among brand and comms teams has been that getting mentioned, with attribution, in a ChatGPT answer was the prize. Appear in the citation list, and you have a foothold in the AI answer economy. GPT-6's visual layer complicates that. A brand's data can be present in the underlying synthesis yet invisible in the rendered output if it loses the competition for which facts earn a slot in the chart, the table, or the card. OpenAI has not published the selection logic for what gets visualised versus what gets relegated to a text footnote, and it is unlikely to. That opacity is the point: it creates a new, unmeasured layer of editorial judgment sitting on top of retrieval, one that brands cannot audit the way they can check whether their domain shows up in a citation list.

This matters more for sectors where comparison is the entire point of the query. A pension fund researching ESG-linked bond issuers, a procurement officer at an industrial group comparing turbine suppliers, a donor vetting development-finance institutions: these are exactly the queries Intelligent UI is built to answer with a table, not a paragraph. If GPT-6 renders a comparison grid with five rows and your institution is the sixth fact mentioned in the surrounding prose, you have effectively lost the query. Attribution without visual placement is becoming the new version of ranking on page two.

What structured data has to do with it

The practical implication is unglamorous but concrete: the brands likely to win slots in GPT-6's rendered tables and cards are the ones whose data already exists in a form a model can lift cleanly, structured specifications, labelled comparison points, machine-readable facts on owned web pages, rather than discursive brand narrative buried in PDFs and annual reports. This has been good practice for schema markup and structured content for years. It now has a sharper commercial edge, because the cost of unstructured content is no longer "ranks lower in a list of ten links." It is "does not exist in the part of the answer the user actually reads."

Multilaterals and policy institutions have a particular exposure here. Their authority has historically rested on long-form reports, communiqués and position papers, exactly the format Intelligent UI is designed to compress or bypass. A reader asking ChatGPT to compare disaster-risk financing frameworks across the UN system is now plausibly getting a generated table, built from whatever structured data points the model found easiest to extract. An institution whose flagship findings live only in a 90-page PDF risks becoming the unvisualised source, cited in a sentence nobody reads, while a competitor with a clean public dataset or a well-tagged summary page gets the grid.

OpenAI frames Intelligent UI as a user-experience upgrade, and for users it probably is: faster, more legible answers beat scrolling through prose. For brands it is a quiet redefinition of the competitive surface. The question is no longer only "does the model know about us." It is "does the model consider us worth drawing."