ChatGPT Rebuilt Its Search Tool, I Read the New Language It Speaks

On August 16, when ChatGPT ran a web search for me, the tool call was JSON. On August 20, I asked the same question on the same account.

The tool call now looks like this. Somewhere in those four days, OpenAI replaced the JSON with a compact query language of its own.

Idea 1: Every Line Is A Search

A tool call is now a block of lines, one search per line, fields separated by pipes. The pattern that repeats is a call type, the query, a number, and sometimes a domain. fast is the call type. The query reads like the old fan out queries,. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

The risk is usually hidden in the execution layer. A page can look fine to a human and still fail for an automated visitor if the form, call to action, rendering path, or confirmation step is not accessible enough for the agent to complete the task.

Practical AI SEO/GEO Tip

Ask the questions your buyers ask and check the brands written by ChatGPT. The brand names in those queries are the competitive set ChatGPT already holds for your category, and the domains in the slots are the sites it intends to visit. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

The useful check is whether this improves the system behind search performance, not only the words on the page. Internal links, crawlable content, clear entities, current evidence, and a sensible page structure all help the recommendation become easier to trust.

Idea 2: The 3rd Field Is A Freshness Window

Here's the same field across five lines from the questions I asked on August 20. A stock price gets two. Football results get seven. Commercial product research gets 30. Earnings guidance gets 90. I think this number is a recency window in. The search implication is whether the section improves the evidence around the page, not simply whether it adds more wording. Clear entities, crawlable structure, internal links, and useful context are what make the topic easier to evaluate.

Practical AI SEO/GEO Tip in practice

On the commercial questions I ran, every brand probe had a 30 day window. If the days assumption is right, a pricing or comparison page that hasn't changed in a month is competing from outside the default window on exactly the queries. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

Idea 3: Search Is A Set Of Verticals

The call types I've captured so far, with the real lines. fast is the web search. It's the successor to the whole old fan out. product is a catalogue lookup. It appeared on robot vacuums and mattresses, and on none of the software. The search implication is whether the section improves the evidence around the page, not simply whether it adds more wording. Clear entities, crawlable structure, internal links, and useful context are what make the topic easier to evaluate.

How Practical AI SEO/GEO Tip changes the decision

If you sell physical products, check whether yours come back with cards, offers and merchant names when ChatGPT recommends them. Cards mean the catalogue knows you. A text mention with no card means the content game is the only one you're. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

The operational question is whether the public business data is complete enough to support the query. Hours, categories, services, reviews, photos, and page content need to reinforce each other so Google can understand the business in a specific situation, not only as a generic listing.

Idea 4: The Domain Slot

Until this week, when ChatGPT wanted to search one specific website, it wrote the classic site:intercom.com in front of the query, the same operator search we SEOs often use… The new format has a slot for it instead. The last field. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

Where Practical AI SEO/GEO Tip needs clearer evidence

Read the slot on your own captures and check the domain is one you actually run. Rebrands, migrations, and country domains are the cases that produce a wrong slot, because the model fills it from memory, and memory lags. A wasted probe. The practical read is that brand signals need to be consistent enough for both people and AI systems to form a stable view of the company, its expertise, and its trust signals.

Idea 5: Some Answers Are Becoming Widgets

The genui_run lines don't return pages. They return interface. In the finance answer, the stock chart showed up as a reference named stock_chart pointing at a widget hosted on cdn.platform.openai.com, version 2.0, hydrated from a state. The practical question is what this changes in the system: the page structure, the evidence presented, the measurement habit, or the way the topic is connected to related work.

Practical AI SEO/GEO Tip 2

Run the queries you rank for and note which come back with a widget where the answer used to be. A genui_run line in the capture is the tell. I can't give you a tactic for winning inside a widget, because on what I've seen, there's nothing. The search implication is whether the section improves the evidence around the page, not simply whether it adds more wording. Clear entities, crawlable structure, internal links, and useful context are what make the topic easier to evaluate.

What the visibility signal actually changes

What the visibility signal actually changes: chatGPT Rebuilt Its Search Tool, I Read the New Language It Speaks: the Practical Angle should be treated as a visibility signal, not a standalone headline. Introduction On August 16, when ChatGPT ran a web search for me, the tool call was JSON. On August 20, I asked the same question on the same account. The tool call now looks like this. Somewhere in those four days, OpenAI replaced the JSON with a compact. This connects with Not the Outputs) when the same signal needs a clearer operating decision. A useful companion note is Not the Answers), because it looks at a nearby part of the same system.

What the visibility signal actually changes: the practical question is whether the page, brand evidence, and surrounding content make the answer easier to trust. If that support is weak, search systems can still understand the topic but fail to connect it confidently to the brand.

What the visibility signal actually changes: that is why the response should begin with an audit of the evidence already on the site before creating a new asset. The fastest improvement is often a clearer page, a better internal link, or a stronger explanation of why the brand belongs in the answer. The same pattern also shows up in AI Agents Read Your Site & It’s Breaking, where the practical question is how the signal becomes visible.