An Tran Solutions
VI
An Tran Solutions
Back to Blog
newsAIOpenAIChatGPTSEOweb design

GPT-6 Turns Answers Into Interfaces. Websites Built on Generic Tools Will Lose Visitors First

October 8, 20266 min readby An Tran
On this page

Read the GPT-6 coverage from the past two days and you'll find one story: ChatGPT got prettier, faster, and now has charts and buttons. A nice UX upgrade.

I read it differently. For the first time, a chat window used by 1.2 billion people a week is building the exact things your website exists to provide: comparison tables, cost calculators, checklists, forms that help people pick an option. No designer involved. No link clicked.

For a decade, websites competed on content. As of this week, the generic interface layer of a website has a competitor too, and that competitor is free.

What happened

On October 7, 2026, OpenAI announced that GPT-6 is coming to the Chat tab in ChatGPT. According to MacRumors and The Decoder, the rollout looks like this:

  • From October 7: Plus, Pro, Business and Enterprise users get GPT-6 Sol.
  • From October 8: Free and Go users get GPT-6 Luna, the cheaper, speed-optimized model. That wave starts today, per OpenAI's schedule.
  • Models in ChatGPT Work and Codex don't change. Sol and Luna have been on the API since September 22, per Search Engine Journal.

The headline feature isn't the model. It's Intelligent UI. Answers can now include buttons, forms, charts, diagrams, maps and checklists, and users can calculate, compare and adjust data without leaving the chat. OpenAI's own examples (via iClarified and MacRumors): a savings calculator, a dinner bill splitter, a recipe that rescales ingredients by guest count, even a small game.

One more technical detail: GPT-6 can start answering while it's still reasoning. OpenAI says that for questions needing a web search, GPT-6 Instant starts responding 44% sooner on average than GPT-5.6 Instant.

There are limits. Per Search Engine Journal, Intelligent UI is rolling out gradually on the web and in updated apps. The Pro thinking level (which runs GPT-6 Astra) and older macOS/Windows desktop apps don't support it. Web users can switch off "Layout and visuals", though some visuals may still appear.

Why this is more than "ChatGPT looks nicer"

Look again at the examples OpenAI chose: savings calculator, bill splitter, comparison table. Those aren't fun extras. They're a list of the page types that many businesses use to pull visitors in from Google.

"Estimate your renovation cost." "Compare plans A, B and C." "Loan payment calculator." "Pre-launch checklist." They share a formula: generic inputs, public math, a result anyone could produce. That's precisely what a model can rebuild in seconds.

Search Engine Journal makes the same point: OpenAI's examples rely on generic inputs and public math, the kind of tool that's easiest to replicate. Interactive answers that start appearing before the model finishes thinking give people fewer reasons to visit a site for simple tasks. And how sources get credited inside a chart or a tool, OpenAI hasn't explained. For now, citations sit behind a "Sources" button when they're available.

This isn't just OpenAI, either. Google got there first:

  • Generative UI debuted in AI Mode in November 2025, alongside Gemini 3.
  • On August 19, 2026, Google began bringing it into AI Overviews, the AI summary at the top of search results, per Search Engine Journal. Google says AI Overviews reach more than 2.5 billion monthly users.

So both of the biggest doors into your website, Google and ChatGPT, can now generate a tool instead of sending people to yours.

We've seen this film before

Lost clicks aren't new. According to Pew Research Center (March 2025 data, published July 22, 2025):

  • When Google showed an AI summary, users clicked a traditional result on 8% of visits, versus 15% without one.
  • Only 1% of visits included a click on a link inside the summary.
  • 26% of visits to pages with an AI summary ended the browsing session, versus 16% without.

That was when AI only wrote text. Now it builds tools, so "visit the site to use the calculator" is no longer a safe reason for anyone to click.

The part almost nobody noticed

If the story stopped there, this would be a lament. But there's a number worth more than every headline this week, and it comes from Google itself.

In Google's research paper on generative UI, summarized by Search Engine Journal, AI-generated interfaces won 90% of head-to-head comparisons against the top search result. Scary. But against websites built by people for each specific request, the result flips: the custom sites won 50%, generated interfaces just 35.3%.

And those custom sites weren't premium builds. Each cost roughly $100–130 and three to five hours of work. It's a lab test designed by Google, so read it carefully. But the direction is clear:

AI beats the generic web page. It hasn't beaten a page carefully built for one specific need.

That's my thesis. GPT-6 doesn't kill websites. It kills the average part of websites. Pages that only repeat public knowledge and public math lose visitors first. Pages holding something AI can't generate become more valuable, because the rest of the web is being flattened.

Honestly, I've built "cost estimator" lead magnets for client projects myself. They used to work. If I were starting one today, that's not where I'd begin.

What AI can't build for itself

Search Engine Journal draws the line clearly: tools built on non-public data, tied to logged-in workflows, or relying on integrations beyond the page are much harder to copy. For a small or mid-sized business, I'd split that into four things:

  1. Your own data. Real prices, real stock, real availability, real delivery times by region. ChatGPT can do generic loan math. It doesn't know how many units you have left on the 12th floor.
  2. Proof that can't be faked. Shipped projects, real photos, client results, named reviews. AI can summarize praise. It can't produce a track record.
  3. Action. Booking, deposits, quote requests, payment. The chat may be where people research, but the transaction still has to land on your site, and that site has to be fast, clear and trustworthy.
  4. Opinion and experience. "Which plan suits a three-location store?" gets an average answer from AI. "We've built 40 stores like that, and here's the most common mistake" does not.

What to do this quarter

A 30-minute check: open the 20 pages that bring you the most Google traffic. For each one, ask a single question: "If a customer asked ChatGPT this, could it build something equivalent in ten seconds?" If yes, that page is standing on borrowed ground.

Then sort your pages into three groups and treat them differently:

  • "AI can do this": generic calculators, basic explainers, standard checklists. Don't pour more ad spend into them or pad them out. Either tie them to your own data (a calculator that uses your real price list) or accept that their traffic will shrink.
  • "Only you have this": pricing, projects, capabilities, policies. Make them readable by humans and machines: structured data (schema), current prices and details, nothing hidden inside images or PDFs. If you want ChatGPT to cite you, don't block its search crawler (OAI-SearchBot).
  • "Closing the deal": contact, booking, quote pages. This is where visitors from the chat window will finally land. Load speed, short forms and transparent information matter more than ever, because people arriving here have already been filtered by AI. They're coming to decide, not to read.

Finally: measure it properly. Separate traffic from ChatGPT and other AI tools in Google Analytics, or you won't know whether you're losing visitors or just changing where they come from. If your tagging isn't solid, our post on common GA4 and GTM mistakes is the place to start.

One more signal: whoever owns the interface sells ads on it

Two days before GPT-6, on October 5, 2026, OpenAI announced a new visual ad format shown while ChatGPT generates images, to be tested in the US later in October for Free and Go users, alongside a set of new measurement and attribution partners, per Complete AI Training.

Put the two announcements together and the picture is familiar: whoever owns the interface owns attention, and whoever owns attention sells it. Google has done it for 25 years. ChatGPT is doing the same thing, faster.

For a business, the lesson isn't "go buy ChatGPT ads". It's this: your website is the only part of the customer journey you actually own. Every chat window and results page is rented land. That land keeps getting more expensive and keeps getting fewer windows to the outside.

The bottom line

This week's headlines will talk about pretty charts and answers that arrive 44% sooner. The business consequence is elsewhere: the box-ticking "utility" pages that used to bring in cheap traffic are losing value fast. Not because they got worse, but because a free version of them now sits in the pocket of 1.2 billion people.

There's no need to panic. There is a need to re-sort: spend less on pages AI can replace, make the pages only you can offer excellent and machine-readable, and make the pages that close the deal fast.

If you'd like a second pair of eyes to sort your site into those three groups, see our SEO and performance service or get in touch.

Sources

Related articles

An Tran Solutions
September 30, 20268 min read

Spec-Driven Development Isn't Documentation. It's a Governance System for AI-Written Code

From vibe coding to spec-driven development: why longer specs won't save you, and a four-layer governance framework (constitution, risk-tiered specs, executable acceptance, control gates) that keeps AI-written code under your control. With data from Veracode, Thoughtworks and Martin Fowler.

spec-driven developmentvibe codingAIengineering managementsoftware engineering
Read article