The free scorecard has been the most dependable lead magnet in coaching for a decade. On October 7 it became a standard feature of a chatbot.

OpenAI released a new GPT-6 model in ChatGPT built for the more than 1.2 billion people who use the product each week, introducing a capability called Intelligent UI that lets ChatGPT answer with fully interactive interfaces and "quickly create a tool to solve a task right in the moment."

Not a text answer about a tool. The tool.

The rollout started with paid tiers and reached Free and Go users on October 8, 2026. Paid tiers run on GPT-6 Sol, Free and Go on GPT-6 Luna, both tuned for everyday conversation. So the version in a cold prospect's hands at 11pm on a Tuesday is the free one, and it has this.

What actually shipped

OpenAI says it trained GPT-6 to compose responses using text, visuals and interactive elements, choosing how they fit together based on the question: graphics, tappable buttons, forms, charts and interactive experiences used directly in the conversation, with a comparison rendered side by side, an explanation turned into an interactive diagram, or a plain text answer when that is most useful.

No special prompt is required, though users can ask ChatGPT to create a tool for a particular task.

Now read OpenAI's own demo list with a coach's eye. The company's examples include a calculator for exploring how savings could grow, a bill splitter for dinner with friends, and a game playable inside the conversation.

A savings calculator is a financial coach's entire opt-in page. A side-by-side comparison is the "which program is right for you" grid. A scored form is the readiness assessment that sits at the top of a few thousand coaching funnels right now.

Under the hood, OpenAI built a component library and a compiler that renders interfaces during generation, so users see parts of the interface before the response is finished. The model also starts answering while it is still thinking, building an answer across partial responses. It feels less like waiting for a reply and more like watching software assemble itself.

Why this specifically hits the coaching funnel

The quiz was never valuable because the questions were clever. It was valuable because of friction.

Someone had to pick a quiz platform, write the logic branches, design the scoring bands, wire it to an email provider, build a results page and a nurture sequence. That work took a week, or it took money, and most coaches never did it. The ones who did had something the others didn't.

That friction is now roughly one sentence long. A prospect who wants to know whether they're ready to hire a business coach can ask for an assessment, answer it, and get a result without ever meeting a landing page, a pop-up or a double opt-in.

The exchange that powered the whole model — your diagnostic for their email address — only works when the diagnostic is scarce on one side.

The limits, which are real

This is not total. OpenAI's own release notes put fences around it.

Intelligent UI is available from Instant through Extra High reasoning; the Pro reasoning option continues to use GPT-6 Astra and doesn't support Intelligent UI, and the models powering Work and Codex aren't changing with this release.

The company is also openly unfinished on quality. OpenAI says there's "still work ahead to improve the model's design judgment and expand what it can create." Anyone who has watched an LLM confidently produce a weighted scoring rubric with no basis for the weights knows what that looks like in practice.

And a generated tool is ephemeral. It exists inside a chat thread. It is not in a CRM, it is not tagged, it does not trigger anything, and nobody follows up. A prospect who scores badly gets a number and closes the tab.

What a generated assessment can't produce

A score is not a diagnosis. The coach's actual product was never the twelve questions.

It was the interpretation — what a 62 means for someone running a seven-person agency versus someone solo. It was the benchmark: a quiz built on the spot has no comparison set, while a coach who has run the same assessment across hundreds of clients can tell a prospect where they sit against people like them. And it was the call afterwards, which is where assessments have always converted.

The practical move is not to kill the scorecard. It's to stop treating the instrument as the giveaway and start treating the read-out as the giveaway.

Three things that survive this week intact:

Proprietary data. Your scoring bands mean something if they're built from your client base. ChatGPT can generate the form. It cannot tell someone they scored in the bottom third of 400 founders you've personally worked with.

The human verdict. Deliver results live, on a short call, or in a recorded personal response. That was always the conversion mechanism. It is now also the differentiator.

Diagnosis of the thing people can't self-report. The assessments most at risk are the ones where the user already knows the answers and just wants them organised. The durable ones surface what the person cannot see about themselves.

There's a pattern here worth watching

This isn't OpenAI's first move into interactive explanation. In March 2026 the company introduced interactive learning modules for around 70 STEM topics, from the Pythagorean theorem to compound interest, where users could adjust sliders to understand concepts; Intelligent UI takes that principle out of the educational context and applies it to all responses.

That's the trajectory. Something gets built narrowly for education, then generalised to everything. The worksheet, the template, the self-audit, the curriculum outline and the "download my framework" PDF are all the same category of object: static information a reader has to apply themselves.

The quiz was never the moat. This is the week a lot of funnels find out what else was holding them up.